diff options
Diffstat (limited to 'drivers/net/ethernet/intel/ice/ice_txrx.c')
| -rw-r--r-- | drivers/net/ethernet/intel/ice/ice_txrx.c | 1871 |
1 files changed, 867 insertions, 1004 deletions
diff --git a/drivers/net/ethernet/intel/ice/ice_txrx.c b/drivers/net/ethernet/intel/ice/ice_txrx.c index abdb137c8bb7..ad76768a4232 100644 --- a/drivers/net/ethernet/intel/ice/ice_txrx.c +++ b/drivers/net/ethernet/intel/ice/ice_txrx.c @@ -3,19 +3,25 @@ /* The driver transmit and receive code */ -#include <linux/prefetch.h> #include <linux/mm.h> +#include <linux/netdevice.h> +#include <linux/prefetch.h> #include <linux/bpf_trace.h> +#include <linux/net/intel/libie/rx.h> +#include <net/libeth/xdp.h> +#include <net/dsfield.h> +#include <net/mpls.h> #include <net/xdp.h> #include "ice_txrx_lib.h" #include "ice_lib.h" #include "ice.h" +#include "ice_trace.h" #include "ice_dcb_lib.h" #include "ice_xsk.h" +#include "ice_eswitch.h" #define ICE_RX_HDR_SIZE 256 -#define FDIR_DESC_RXDID 0x40 #define ICE_FDIR_CLEAN_DELAY 10 /** @@ -31,7 +37,7 @@ ice_prgm_fdir_fltr(struct ice_vsi *vsi, struct ice_fltr_desc *fdir_desc, struct ice_tx_buf *tx_buf, *first; struct ice_fltr_desc *f_desc; struct ice_tx_desc *tx_desc; - struct ice_ring *tx_ring; + struct ice_tx_ring *tx_ring; struct device *dev; dma_addr_t dma; u32 td_cmd; @@ -80,7 +86,7 @@ ice_prgm_fdir_fltr(struct ice_vsi *vsi, struct ice_fltr_desc *fdir_desc, td_cmd = ICE_TXD_LAST_DESC_CMD | ICE_TX_DESC_CMD_DUMMY | ICE_TX_DESC_CMD_RE; - tx_buf->tx_flags = ICE_TX_FLAGS_DUMMY_PKT; + tx_buf->type = ICE_TX_BUF_DUMMY; tx_buf->raw_buf = raw_packet; tx_desc->cmd_type_offset_bsz = @@ -105,47 +111,100 @@ ice_prgm_fdir_fltr(struct ice_vsi *vsi, struct ice_fltr_desc *fdir_desc, * @tx_buf: the buffer to free */ static void -ice_unmap_and_free_tx_buf(struct ice_ring *ring, struct ice_tx_buf *tx_buf) +ice_unmap_and_free_tx_buf(struct ice_tx_ring *ring, struct ice_tx_buf *tx_buf) { - if (tx_buf->skb) { - if (tx_buf->tx_flags & ICE_TX_FLAGS_DUMMY_PKT) - devm_kfree(ring->dev, tx_buf->raw_buf); - else if (ice_ring_is_xdp(ring)) - page_frag_free(tx_buf->raw_buf); - else - dev_kfree_skb_any(tx_buf->skb); - if (dma_unmap_len(tx_buf, len)) - dma_unmap_single(ring->dev, - dma_unmap_addr(tx_buf, dma), - dma_unmap_len(tx_buf, len), - DMA_TO_DEVICE); - } else if (dma_unmap_len(tx_buf, len)) { + if (tx_buf->type != ICE_TX_BUF_XDP_TX && dma_unmap_len(tx_buf, len)) dma_unmap_page(ring->dev, dma_unmap_addr(tx_buf, dma), dma_unmap_len(tx_buf, len), DMA_TO_DEVICE); + + switch (tx_buf->type) { + case ICE_TX_BUF_DUMMY: + devm_kfree(ring->dev, tx_buf->raw_buf); + break; + case ICE_TX_BUF_SKB: + dev_kfree_skb_any(tx_buf->skb); + break; + case ICE_TX_BUF_XDP_TX: + libeth_xdp_return_va(tx_buf->raw_buf, false); + break; + case ICE_TX_BUF_XDP_XMIT: + xdp_return_frame(tx_buf->xdpf); + break; } tx_buf->next_to_watch = NULL; - tx_buf->skb = NULL; + tx_buf->type = ICE_TX_BUF_EMPTY; dma_unmap_len_set(tx_buf, len, 0); /* tx_buf must be completely set up in the transmit path */ } -static struct netdev_queue *txring_txq(const struct ice_ring *ring) +static struct netdev_queue *txring_txq(const struct ice_tx_ring *ring) { return netdev_get_tx_queue(ring->netdev, ring->q_index); } /** + * ice_clean_tstamp_ring - clean time stamp ring + * @tx_ring: Tx ring to clean the Time Stamp ring for + */ +static void ice_clean_tstamp_ring(struct ice_tx_ring *tx_ring) +{ + struct ice_tstamp_ring *tstamp_ring = tx_ring->tstamp_ring; + u32 size; + + if (!tstamp_ring->desc) + return; + + size = ALIGN(tstamp_ring->count * sizeof(struct ice_ts_desc), + PAGE_SIZE); + memset(tstamp_ring->desc, 0, size); + tstamp_ring->next_to_use = 0; +} + +/** + * ice_free_tstamp_ring - free time stamp resources per queue + * @tx_ring: Tx ring to free the Time Stamp ring for + */ +void ice_free_tstamp_ring(struct ice_tx_ring *tx_ring) +{ + struct ice_tstamp_ring *tstamp_ring = tx_ring->tstamp_ring; + u32 size; + + if (!tstamp_ring->desc) + return; + + ice_clean_tstamp_ring(tx_ring); + size = ALIGN(tstamp_ring->count * sizeof(struct ice_ts_desc), + PAGE_SIZE); + dmam_free_coherent(tx_ring->dev, size, tstamp_ring->desc, + tstamp_ring->dma); + tstamp_ring->desc = NULL; +} + +/** + * ice_free_tx_tstamp_ring - free time stamp resources per Tx ring + * @tx_ring: Tx ring to free the Time Stamp ring for + */ +void ice_free_tx_tstamp_ring(struct ice_tx_ring *tx_ring) +{ + ice_free_tstamp_ring(tx_ring); + kfree_rcu(tx_ring->tstamp_ring, rcu); + tx_ring->tstamp_ring = NULL; + tx_ring->flags &= ~ICE_TX_FLAGS_TXTIME; +} + +/** * ice_clean_tx_ring - Free any empty Tx buffers * @tx_ring: ring to be cleaned */ -void ice_clean_tx_ring(struct ice_ring *tx_ring) +void ice_clean_tx_ring(struct ice_tx_ring *tx_ring) { + u32 size; u16 i; - if (ice_ring_is_xdp(tx_ring) && tx_ring->xsk_umem) { + if (ice_ring_is_xdp(tx_ring) && tx_ring->xsk_pool) { ice_xsk_clean_xdp_ring(tx_ring); goto tx_skip_free; } @@ -161,8 +220,10 @@ void ice_clean_tx_ring(struct ice_ring *tx_ring) tx_skip_free: memset(tx_ring->tx_buf, 0, sizeof(*tx_ring->tx_buf) * tx_ring->count); + size = ALIGN(tx_ring->count * sizeof(struct ice_tx_desc), + PAGE_SIZE); /* Zero out the descriptor ring */ - memset(tx_ring->desc, 0, tx_ring->size); + memset(tx_ring->desc, 0, size); tx_ring->next_to_use = 0; tx_ring->next_to_clean = 0; @@ -172,6 +233,9 @@ tx_skip_free: /* cleanup Tx queue statistics */ netdev_tx_reset_queue(txring_txq(tx_ring)); + + if (ice_is_txtime_cfg(tx_ring)) + ice_free_tx_tstamp_ring(tx_ring); } /** @@ -180,14 +244,18 @@ tx_skip_free: * * Free all transmit software resources */ -void ice_free_tx_ring(struct ice_ring *tx_ring) +void ice_free_tx_ring(struct ice_tx_ring *tx_ring) { + u32 size; + ice_clean_tx_ring(tx_ring); devm_kfree(tx_ring->dev, tx_ring->tx_buf); tx_ring->tx_buf = NULL; if (tx_ring->desc) { - dmam_free_coherent(tx_ring->dev, tx_ring->size, + size = ALIGN(tx_ring->count * sizeof(struct ice_tx_desc), + PAGE_SIZE); + dmam_free_coherent(tx_ring->dev, size, tx_ring->desc, tx_ring->dma); tx_ring->desc = NULL; } @@ -200,7 +268,7 @@ void ice_free_tx_ring(struct ice_ring *tx_ring) * * Returns true if there's any budget left (e.g. the clean is finished) */ -static bool ice_clean_tx_irq(struct ice_ring *tx_ring, int napi_budget) +static bool ice_clean_tx_irq(struct ice_tx_ring *tx_ring, int napi_budget) { unsigned int total_bytes = 0, total_pkts = 0; unsigned int budget = ICE_DFLT_IRQ_WORK; @@ -209,6 +277,9 @@ static bool ice_clean_tx_irq(struct ice_ring *tx_ring, int napi_budget) struct ice_tx_desc *tx_desc; struct ice_tx_buf *tx_buf; + /* get the bql data ready */ + netdev_txq_bql_complete_prefetchw(txring_txq(tx_ring)); + tx_buf = &tx_ring->tx_buf[i]; tx_desc = ICE_TX_DESC(tx_ring, i); i -= tx_ring->count; @@ -222,8 +293,12 @@ static bool ice_clean_tx_irq(struct ice_ring *tx_ring, int napi_budget) if (!eop_desc) break; + /* follow the guidelines of other drivers */ + prefetchw(&tx_buf->skb->users); + smp_rmb(); /* prevent any other reads prior to eop_desc */ + ice_trace(clean_tx_irq, tx_ring, tx_desc, tx_buf); /* if the descriptor isn't done, no work yet to do */ if (!(eop_desc->cmd_type_offset_bsz & cpu_to_le64(ICE_TX_DESC_DTYPE_DESC_DONE))) @@ -236,11 +311,8 @@ static bool ice_clean_tx_irq(struct ice_ring *tx_ring, int napi_budget) total_bytes += tx_buf->bytecount; total_pkts += tx_buf->gso_segs; - if (ice_ring_is_xdp(tx_ring)) - page_frag_free(tx_buf->raw_buf); - else - /* free the skb */ - napi_consume_skb(tx_buf->skb, napi_budget); + /* free the skb */ + napi_consume_skb(tx_buf->skb, napi_budget); /* unmap skb header data */ dma_unmap_single(tx_ring->dev, @@ -249,11 +321,12 @@ static bool ice_clean_tx_irq(struct ice_ring *tx_ring, int napi_budget) DMA_TO_DEVICE); /* clear tx_buf data */ - tx_buf->skb = NULL; + tx_buf->type = ICE_TX_BUF_EMPTY; dma_unmap_len_set(tx_buf, len, 0); /* unmap remaining buffers */ while (tx_desc != eop_desc) { + ice_trace(clean_tx_irq_unmap, tx_ring, tx_desc, tx_buf); tx_buf++; tx_desc++; i++; @@ -272,6 +345,7 @@ static bool ice_clean_tx_irq(struct ice_ring *tx_ring, int napi_budget) dma_unmap_len_set(tx_buf, len, 0); } } + ice_trace(clean_tx_irq_unmap_eop, tx_ring, tx_desc, tx_buf); /* move us one more past the eop_desc for start of next pkt */ tx_buf++; @@ -293,12 +367,7 @@ static bool ice_clean_tx_irq(struct ice_ring *tx_ring, int napi_budget) tx_ring->next_to_clean = i; ice_update_tx_ring_stats(tx_ring, total_pkts, total_bytes); - - if (ice_ring_is_xdp(tx_ring)) - return !!budget; - - netdev_tx_completed_queue(txring_txq(tx_ring), total_pkts, - total_bytes); + netdev_tx_completed_queue(txring_txq(tx_ring), total_pkts, total_bytes); #define TX_WAKE_THRESHOLD ((s16)(DESC_NEEDED * 2)) if (unlikely(total_pkts && netif_carrier_ok(tx_ring->netdev) && @@ -307,12 +376,10 @@ static bool ice_clean_tx_irq(struct ice_ring *tx_ring, int napi_budget) * sees the new next_to_clean. */ smp_mb(); - if (__netif_subqueue_stopped(tx_ring->netdev, - tx_ring->q_index) && - !test_bit(__ICE_DOWN, vsi->state)) { - netif_wake_subqueue(tx_ring->netdev, - tx_ring->q_index); - ++tx_ring->tx_stats.restart_q; + if (netif_tx_queue_stopped(txring_txq(tx_ring)) && + !test_bit(ICE_VSI_DOWN, vsi->state)) { + netif_tx_wake_queue(txring_txq(tx_ring)); + ++tx_ring->ring_stats->tx_stats.restart_q; } } @@ -320,14 +387,93 @@ static bool ice_clean_tx_irq(struct ice_ring *tx_ring, int napi_budget) } /** + * ice_alloc_tstamp_ring - allocate the Time Stamp ring + * @tx_ring: Tx ring to allocate the Time Stamp ring for + * + * Return: 0 on success, negative on error + */ +static int ice_alloc_tstamp_ring(struct ice_tx_ring *tx_ring) +{ + struct ice_tstamp_ring *tstamp_ring; + + /* allocate with kzalloc(), free with kfree_rcu() */ + tstamp_ring = kzalloc(sizeof(*tstamp_ring), GFP_KERNEL); + if (!tstamp_ring) + return -ENOMEM; + + tstamp_ring->tx_ring = tx_ring; + tx_ring->tstamp_ring = tstamp_ring; + tstamp_ring->desc = NULL; + tstamp_ring->count = ice_calc_ts_ring_count(tx_ring); + tx_ring->flags |= ICE_TX_FLAGS_TXTIME; + return 0; +} + +/** + * ice_setup_tstamp_ring - allocate the Time Stamp ring + * @tx_ring: Tx ring to set up the Time Stamp ring for + * + * Return: 0 on success, negative on error + */ +static int ice_setup_tstamp_ring(struct ice_tx_ring *tx_ring) +{ + struct ice_tstamp_ring *tstamp_ring = tx_ring->tstamp_ring; + struct device *dev = tx_ring->dev; + u32 size; + + /* round up to nearest page */ + size = ALIGN(tstamp_ring->count * sizeof(struct ice_ts_desc), + PAGE_SIZE); + tstamp_ring->desc = dmam_alloc_coherent(dev, size, &tstamp_ring->dma, + GFP_KERNEL); + if (!tstamp_ring->desc) { + dev_err(dev, "Unable to allocate memory for Time stamp Ring, size=%d\n", + size); + return -ENOMEM; + } + + tstamp_ring->next_to_use = 0; + return 0; +} + +/** + * ice_alloc_setup_tstamp_ring - Allocate and setup the Time Stamp ring + * @tx_ring: Tx ring to allocate and setup the Time Stamp ring for + * + * Return: 0 on success, negative on error + */ +int ice_alloc_setup_tstamp_ring(struct ice_tx_ring *tx_ring) +{ + struct device *dev = tx_ring->dev; + int err; + + err = ice_alloc_tstamp_ring(tx_ring); + if (err) { + dev_err(dev, "Unable to allocate Time stamp ring for Tx ring %d\n", + tx_ring->q_index); + return err; + } + + err = ice_setup_tstamp_ring(tx_ring); + if (err) { + dev_err(dev, "Unable to setup Time stamp ring for Tx ring %d\n", + tx_ring->q_index); + ice_free_tx_tstamp_ring(tx_ring); + return err; + } + return 0; +} + +/** * ice_setup_tx_ring - Allocate the Tx descriptors * @tx_ring: the Tx ring to set up * * Return 0 on success, negative on error */ -int ice_setup_tx_ring(struct ice_ring *tx_ring) +int ice_setup_tx_ring(struct ice_tx_ring *tx_ring) { struct device *dev = tx_ring->dev; + u32 size; if (!dev) return -ENOMEM; @@ -335,25 +481,25 @@ int ice_setup_tx_ring(struct ice_ring *tx_ring) /* warn if we are about to overwrite the pointer */ WARN_ON(tx_ring->tx_buf); tx_ring->tx_buf = - devm_kzalloc(dev, sizeof(*tx_ring->tx_buf) * tx_ring->count, + devm_kcalloc(dev, sizeof(*tx_ring->tx_buf), tx_ring->count, GFP_KERNEL); if (!tx_ring->tx_buf) return -ENOMEM; /* round up to nearest page */ - tx_ring->size = ALIGN(tx_ring->count * sizeof(struct ice_tx_desc), - PAGE_SIZE); - tx_ring->desc = dmam_alloc_coherent(dev, tx_ring->size, &tx_ring->dma, + size = ALIGN(tx_ring->count * sizeof(struct ice_tx_desc), + PAGE_SIZE); + tx_ring->desc = dmam_alloc_coherent(dev, size, &tx_ring->dma, GFP_KERNEL); if (!tx_ring->desc) { dev_err(dev, "Unable to allocate memory for the Tx descriptor ring, size=%d\n", - tx_ring->size); + size); goto err; } tx_ring->next_to_use = 0; tx_ring->next_to_clean = 0; - tx_ring->tx_stats.prev_pkt = -1; + tx_ring->ring_stats->tx_stats.prev_pkt = -1; return 0; err: @@ -362,57 +508,71 @@ err: return -ENOMEM; } +void ice_rxq_pp_destroy(struct ice_rx_ring *rq) +{ + struct libeth_fq fq = { + .fqes = rq->rx_fqes, + .pp = rq->pp, + }; + + libeth_rx_fq_destroy(&fq); + rq->rx_fqes = NULL; + rq->pp = NULL; + + if (!rq->hdr_pp) + return; + + fq.fqes = rq->hdr_fqes; + fq.pp = rq->hdr_pp; + + libeth_rx_fq_destroy(&fq); + rq->hdr_fqes = NULL; + rq->hdr_pp = NULL; +} + /** * ice_clean_rx_ring - Free Rx buffers * @rx_ring: ring to be cleaned */ -void ice_clean_rx_ring(struct ice_ring *rx_ring) +void ice_clean_rx_ring(struct ice_rx_ring *rx_ring) { - struct device *dev = rx_ring->dev; - u16 i; + u32 size; - /* ring already cleared, nothing to do */ - if (!rx_ring->rx_buf) - return; - - if (rx_ring->xsk_umem) { + if (rx_ring->xsk_pool) { ice_xsk_clean_rx_ring(rx_ring); goto rx_skip_free; } + /* ring already cleared, nothing to do */ + if (!rx_ring->rx_fqes) + return; + + libeth_xdp_return_stash(&rx_ring->xdp); + /* Free all the Rx ring sk_buffs */ - for (i = 0; i < rx_ring->count; i++) { - struct ice_rx_buf *rx_buf = &rx_ring->rx_buf[i]; + for (u32 i = rx_ring->next_to_clean; i != rx_ring->next_to_use; ) { + libeth_rx_recycle_slow(rx_ring->rx_fqes[i].netmem); - if (rx_buf->skb) { - dev_kfree_skb(rx_buf->skb); - rx_buf->skb = NULL; - } - if (!rx_buf->page) - continue; + if (rx_ring->hdr_pp) + libeth_rx_recycle_slow(rx_ring->hdr_fqes[i].netmem); - /* Invalidate cache lines that may have been written to by - * device so that we avoid corrupting memory. - */ - dma_sync_single_range_for_cpu(dev, rx_buf->dma, - rx_buf->page_offset, - rx_ring->rx_buf_len, - DMA_FROM_DEVICE); - - /* free resources associated with mapping */ - dma_unmap_page_attrs(dev, rx_buf->dma, ice_rx_pg_size(rx_ring), - DMA_FROM_DEVICE, ICE_RX_DMA_ATTR); - __page_frag_cache_drain(rx_buf->page, rx_buf->pagecnt_bias); - - rx_buf->page = NULL; - rx_buf->page_offset = 0; + if (unlikely(++i == rx_ring->count)) + i = 0; } -rx_skip_free: - memset(rx_ring->rx_buf, 0, sizeof(*rx_ring->rx_buf) * rx_ring->count); + if (rx_ring->vsi->type == ICE_VSI_PF && + xdp_rxq_info_is_reg(&rx_ring->xdp_rxq)) { + xdp_rxq_info_detach_mem_model(&rx_ring->xdp_rxq); + xdp_rxq_info_unreg(&rx_ring->xdp_rxq); + } + + ice_rxq_pp_destroy(rx_ring); +rx_skip_free: /* Zero out the descriptor ring */ - memset(rx_ring->desc, 0, rx_ring->size); + size = ALIGN(rx_ring->count * sizeof(union ice_32byte_rx_desc), + PAGE_SIZE); + memset(rx_ring->desc, 0, size); rx_ring->next_to_alloc = 0; rx_ring->next_to_clean = 0; @@ -425,19 +585,22 @@ rx_skip_free: * * Free all receive software resources */ -void ice_free_rx_ring(struct ice_ring *rx_ring) +void ice_free_rx_ring(struct ice_rx_ring *rx_ring) { + struct device *dev = ice_pf_to_dev(rx_ring->vsi->back); + u32 size; + ice_clean_rx_ring(rx_ring); - if (rx_ring->vsi->type == ICE_VSI_PF) - if (xdp_rxq_info_is_reg(&rx_ring->xdp_rxq)) - xdp_rxq_info_unreg(&rx_ring->xdp_rxq); - rx_ring->xdp_prog = NULL; - devm_kfree(rx_ring->dev, rx_ring->rx_buf); - rx_ring->rx_buf = NULL; + WRITE_ONCE(rx_ring->xdp_prog, NULL); + if (rx_ring->xsk_pool) { + kfree(rx_ring->xdp_buf); + rx_ring->xdp_buf = NULL; + } if (rx_ring->desc) { - dmam_free_coherent(rx_ring->dev, rx_ring->size, - rx_ring->desc, rx_ring->dma); + size = ALIGN(rx_ring->count * sizeof(union ice_32byte_rx_desc), + PAGE_SIZE); + dmam_free_coherent(dev, size, rx_ring->desc, rx_ring->dma); rx_ring->desc = NULL; } } @@ -448,30 +611,20 @@ void ice_free_rx_ring(struct ice_ring *rx_ring) * * Return 0 on success, negative on error */ -int ice_setup_rx_ring(struct ice_ring *rx_ring) +int ice_setup_rx_ring(struct ice_rx_ring *rx_ring) { - struct device *dev = rx_ring->dev; - - if (!dev) - return -ENOMEM; - - /* warn if we are about to overwrite the pointer */ - WARN_ON(rx_ring->rx_buf); - rx_ring->rx_buf = - devm_kzalloc(dev, sizeof(*rx_ring->rx_buf) * rx_ring->count, - GFP_KERNEL); - if (!rx_ring->rx_buf) - return -ENOMEM; + struct device *dev = ice_pf_to_dev(rx_ring->vsi->back); + u32 size; /* round up to nearest page */ - rx_ring->size = ALIGN(rx_ring->count * sizeof(union ice_32byte_rx_desc), - PAGE_SIZE); - rx_ring->desc = dmam_alloc_coherent(dev, rx_ring->size, &rx_ring->dma, + size = ALIGN(rx_ring->count * sizeof(union ice_32byte_rx_desc), + PAGE_SIZE); + rx_ring->desc = dmam_alloc_coherent(dev, size, &rx_ring->dma, GFP_KERNEL); if (!rx_ring->desc) { dev_err(dev, "Unable to allocate memory for the Rx descriptor ring, size=%d\n", - rx_ring->size); - goto err; + size); + return -ENOMEM; } rx_ring->next_to_use = 0; @@ -480,49 +633,7 @@ int ice_setup_rx_ring(struct ice_ring *rx_ring) if (ice_is_xdp_ena_vsi(rx_ring->vsi)) WRITE_ONCE(rx_ring->xdp_prog, rx_ring->vsi->xdp_prog); - if (rx_ring->vsi->type == ICE_VSI_PF && - !xdp_rxq_info_is_reg(&rx_ring->xdp_rxq)) - if (xdp_rxq_info_reg(&rx_ring->xdp_rxq, rx_ring->netdev, - rx_ring->q_index)) - goto err; return 0; - -err: - devm_kfree(dev, rx_ring->rx_buf); - rx_ring->rx_buf = NULL; - return -ENOMEM; -} - -/** - * ice_rx_offset - Return expected offset into page to access data - * @rx_ring: Ring we are requesting offset of - * - * Returns the offset value for ring into the data buffer. - */ -static unsigned int ice_rx_offset(struct ice_ring *rx_ring) -{ - if (ice_ring_uses_build_skb(rx_ring)) - return ICE_SKB_PAD; - else if (ice_is_xdp_ena_vsi(rx_ring->vsi)) - return XDP_PACKET_HEADROOM; - - return 0; -} - -static unsigned int ice_rx_frame_truesize(struct ice_ring *rx_ring, - unsigned int size) -{ - unsigned int truesize; - -#if (PAGE_SIZE < 8192) - truesize = ice_rx_pg_size(rx_ring) / 2; /* Must be power-of-2 */ -#else - truesize = ice_rx_offset(rx_ring) ? - SKB_DATA_ALIGN(ice_rx_offset(rx_ring) + size) + - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) : - SKB_DATA_ALIGN(size); -#endif - return truesize; } /** @@ -530,41 +641,75 @@ static unsigned int ice_rx_frame_truesize(struct ice_ring *rx_ring, * @rx_ring: Rx ring * @xdp: xdp_buff used as input to the XDP program * @xdp_prog: XDP program to run + * @xdp_ring: ring to be used for XDP_TX action + * @eop_desc: Last descriptor in packet to read metadata from * * Returns any of ICE_XDP_{PASS, CONSUMED, TX, REDIR} */ -static int -ice_run_xdp(struct ice_ring *rx_ring, struct xdp_buff *xdp, - struct bpf_prog *xdp_prog) +static u32 +ice_run_xdp(struct ice_rx_ring *rx_ring, struct libeth_xdp_buff *xdp, + struct bpf_prog *xdp_prog, struct ice_tx_ring *xdp_ring, + union ice_32b_rx_flex_desc *eop_desc) { - int err, result = ICE_XDP_PASS; - struct ice_ring *xdp_ring; + unsigned int ret = ICE_XDP_PASS; u32 act; - act = bpf_prog_run_xdp(xdp_prog, xdp); + if (!xdp_prog) + goto exit; + + xdp->desc = eop_desc; + + act = bpf_prog_run_xdp(xdp_prog, &xdp->base); switch (act) { case XDP_PASS: break; case XDP_TX: - xdp_ring = rx_ring->vsi->xdp_rings[smp_processor_id()]; - result = ice_xmit_xdp_buff(xdp, xdp_ring); + if (static_branch_unlikely(&ice_xdp_locking_key)) + spin_lock(&xdp_ring->tx_lock); + ret = __ice_xmit_xdp_ring(&xdp->base, xdp_ring, false); + if (static_branch_unlikely(&ice_xdp_locking_key)) + spin_unlock(&xdp_ring->tx_lock); + if (ret == ICE_XDP_CONSUMED) + goto out_failure; break; case XDP_REDIRECT: - err = xdp_do_redirect(rx_ring->netdev, xdp, xdp_prog); - result = !err ? ICE_XDP_REDIR : ICE_XDP_CONSUMED; + if (xdp_do_redirect(rx_ring->netdev, &xdp->base, xdp_prog)) + goto out_failure; + ret = ICE_XDP_REDIR; break; default: - bpf_warn_invalid_xdp_action(act); + bpf_warn_invalid_xdp_action(rx_ring->netdev, xdp_prog, act); fallthrough; case XDP_ABORTED: +out_failure: trace_xdp_exception(rx_ring->netdev, xdp_prog, act); fallthrough; case XDP_DROP: - result = ICE_XDP_CONSUMED; - break; + libeth_xdp_return_buff(xdp); + ret = ICE_XDP_CONSUMED; } - return result; +exit: + return ret; +} + +/** + * ice_xmit_xdp_ring - submit frame to XDP ring for transmission + * @xdpf: XDP frame that will be converted to XDP buff + * @xdp_ring: XDP ring for transmission + */ +static int ice_xmit_xdp_ring(const struct xdp_frame *xdpf, + struct ice_tx_ring *xdp_ring) +{ + struct xdp_buff xdp; + + xdp.data_hard_start = (void *)xdpf; + xdp.data = xdpf->data; + xdp.data_end = xdp.data + xdpf->len; + xdp.frame_sz = xdpf->frame_sz; + xdp.flags = xdpf->flags; + + return __ice_xmit_xdp_ring(&xdp, xdp_ring, true); } /** @@ -574,8 +719,8 @@ ice_run_xdp(struct ice_ring *rx_ring, struct xdp_buff *xdp, * @frames: XDP frames to be transmitted * @flags: transmit flags * - * Returns number of frames successfully sent. Frames that fail are - * free'ed via XDP return API. + * Returns number of frames successfully sent. Failed frames + * will be free'ed by XDP core. * For error cases, a negative errno code is returned and no-frames * are transmitted (caller must handle freeing frames). */ @@ -586,83 +731,80 @@ ice_xdp_xmit(struct net_device *dev, int n, struct xdp_frame **frames, struct ice_netdev_priv *np = netdev_priv(dev); unsigned int queue_index = smp_processor_id(); struct ice_vsi *vsi = np->vsi; - struct ice_ring *xdp_ring; - int drops = 0, i; + struct ice_tx_ring *xdp_ring; + struct ice_tx_buf *tx_buf; + int nxmit = 0, i; - if (test_bit(__ICE_DOWN, vsi->state)) + if (test_bit(ICE_VSI_DOWN, vsi->state)) return -ENETDOWN; - if (!ice_is_xdp_ena_vsi(vsi) || queue_index >= vsi->num_xdp_txq) + if (!ice_is_xdp_ena_vsi(vsi)) return -ENXIO; if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK)) return -EINVAL; - xdp_ring = vsi->xdp_rings[queue_index]; + if (static_branch_unlikely(&ice_xdp_locking_key)) { + queue_index %= vsi->num_xdp_txq; + xdp_ring = vsi->xdp_rings[queue_index]; + spin_lock(&xdp_ring->tx_lock); + } else { + /* Generally, should not happen */ + if (unlikely(queue_index >= vsi->num_xdp_txq)) + return -ENXIO; + xdp_ring = vsi->xdp_rings[queue_index]; + } + + tx_buf = &xdp_ring->tx_buf[xdp_ring->next_to_use]; for (i = 0; i < n; i++) { - struct xdp_frame *xdpf = frames[i]; + const struct xdp_frame *xdpf = frames[i]; int err; - err = ice_xmit_xdp_ring(xdpf->data, xdpf->len, xdp_ring); - if (err != ICE_XDP_TX) { - xdp_return_frame_rx_napi(xdpf); - drops++; - } + err = ice_xmit_xdp_ring(xdpf, xdp_ring); + if (err != ICE_XDP_TX) + break; + nxmit++; } + tx_buf->rs_idx = ice_set_rs_bit(xdp_ring); if (unlikely(flags & XDP_XMIT_FLUSH)) ice_xdp_ring_update_tail(xdp_ring); - return n - drops; + if (static_branch_unlikely(&ice_xdp_locking_key)) + spin_unlock(&xdp_ring->tx_lock); + + return nxmit; } /** - * ice_alloc_mapped_page - recycle or make a new page - * @rx_ring: ring to use - * @bi: rx_buf struct to modify - * - * Returns true if the page was successfully allocated or - * reused. + * ice_init_ctrl_rx_descs - Initialize Rx descriptors for control vsi. + * @rx_ring: ring to init descriptors on + * @count: number of descriptors to initialize */ -static bool -ice_alloc_mapped_page(struct ice_ring *rx_ring, struct ice_rx_buf *bi) +void ice_init_ctrl_rx_descs(struct ice_rx_ring *rx_ring, u32 count) { - struct page *page = bi->page; - dma_addr_t dma; - - /* since we are recycling buffers we should seldom need to alloc */ - if (likely(page)) { - rx_ring->rx_stats.page_reuse_count++; - return true; - } + union ice_32b_rx_flex_desc *rx_desc; + u32 ntu = rx_ring->next_to_use; - /* alloc new page for storage */ - page = dev_alloc_pages(ice_rx_pg_order(rx_ring)); - if (unlikely(!page)) { - rx_ring->rx_stats.alloc_page_failed++; - return false; - } + if (!count) + return; - /* map page for use */ - dma = dma_map_page_attrs(rx_ring->dev, page, 0, ice_rx_pg_size(rx_ring), - DMA_FROM_DEVICE, ICE_RX_DMA_ATTR); + rx_desc = ICE_RX_DESC(rx_ring, ntu); - /* 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_pages(page, ice_rx_pg_order(rx_ring)); - rx_ring->rx_stats.alloc_page_failed++; - return false; - } + do { + rx_desc++; + ntu++; + if (unlikely(ntu == rx_ring->count)) { + rx_desc = ICE_RX_DESC(rx_ring, 0); + ntu = 0; + } - bi->dma = dma; - bi->page = page; - bi->page_offset = ice_rx_offset(rx_ring); - page_ref_add(page, USHRT_MAX - 1); - bi->pagecnt_bias = USHRT_MAX; + rx_desc->wb.status_error0 = 0; + count--; + } while (count); - return true; + if (rx_ring->next_to_use != ntu) + ice_release_rx_desc(rx_ring, ntu); } /** @@ -678,43 +820,62 @@ ice_alloc_mapped_page(struct ice_ring *rx_ring, struct ice_rx_buf *bi) * buffers. Then bump tail at most one time. Grouping like this lets us avoid * multiple tail writes per call. */ -bool ice_alloc_rx_bufs(struct ice_ring *rx_ring, u16 cleaned_count) +bool ice_alloc_rx_bufs(struct ice_rx_ring *rx_ring, unsigned int cleaned_count) { + const struct libeth_fq_fp hdr_fq = { + .pp = rx_ring->hdr_pp, + .fqes = rx_ring->hdr_fqes, + .truesize = rx_ring->hdr_truesize, + .count = rx_ring->count, + }; + const struct libeth_fq_fp fq = { + .pp = rx_ring->pp, + .fqes = rx_ring->rx_fqes, + .truesize = rx_ring->truesize, + .count = rx_ring->count, + }; union ice_32b_rx_flex_desc *rx_desc; u16 ntu = rx_ring->next_to_use; - struct ice_rx_buf *bi; /* do nothing if no valid netdev defined */ - if ((!rx_ring->netdev && rx_ring->vsi->type != ICE_VSI_CTRL) || - !cleaned_count) + if (!rx_ring->netdev || !cleaned_count) return false; /* get the Rx descriptor and buffer based on next_to_use */ rx_desc = ICE_RX_DESC(rx_ring, ntu); - bi = &rx_ring->rx_buf[ntu]; do { - /* if we fail here, we have work remaining */ - if (!ice_alloc_mapped_page(rx_ring, bi)) - break; + dma_addr_t addr; - /* sync the buffer for use by the device */ - dma_sync_single_range_for_device(rx_ring->dev, bi->dma, - bi->page_offset, - rx_ring->rx_buf_len, - DMA_FROM_DEVICE); + addr = libeth_rx_alloc(&fq, ntu); + if (addr == DMA_MAPPING_ERROR) { + rx_ring->ring_stats->rx_stats.alloc_page_failed++; + break; + } /* Refresh the desc even if buffer_addrs didn't change * because each write-back erases this info. */ - rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset); + rx_desc->read.pkt_addr = cpu_to_le64(addr); + + if (!hdr_fq.pp) + goto next; + + addr = libeth_rx_alloc(&hdr_fq, ntu); + if (addr == DMA_MAPPING_ERROR) { + rx_ring->ring_stats->rx_stats.alloc_page_failed++; + libeth_rx_recycle_slow(fq.fqes[ntu].netmem); + break; + } + + rx_desc->read.hdr_addr = cpu_to_le64(addr); + +next: rx_desc++; - bi++; ntu++; if (unlikely(ntu == rx_ring->count)) { rx_desc = ICE_RX_DESC(rx_ring, 0); - bi = rx_ring->rx_buf; ntu = 0; } @@ -731,345 +892,41 @@ bool ice_alloc_rx_bufs(struct ice_ring *rx_ring, u16 cleaned_count) } /** - * ice_page_is_reserved - check if reuse is possible - * @page: page struct to check - */ -static bool ice_page_is_reserved(struct page *page) -{ - return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page); -} - -/** - * ice_rx_buf_adjust_pg_offset - Prepare Rx buffer for reuse - * @rx_buf: Rx buffer to adjust - * @size: Size of adjustment + * ice_clean_ctrl_rx_irq - Clean descriptors from flow director Rx ring + * @rx_ring: Rx descriptor ring for ctrl_vsi to transact packets on * - * Update the offset within page so that Rx buf will be ready to be reused. - * For systems with PAGE_SIZE < 8192 this function will flip the page offset - * so the second half of page assigned to Rx buffer will be used, otherwise - * the offset is moved by "size" bytes + * This function cleans Rx descriptors from the ctrl_vsi Rx ring used + * to set flow director rules on VFs. */ -static void -ice_rx_buf_adjust_pg_offset(struct ice_rx_buf *rx_buf, unsigned int size) +void ice_clean_ctrl_rx_irq(struct ice_rx_ring *rx_ring) { -#if (PAGE_SIZE < 8192) - /* flip page offset to other buffer */ - rx_buf->page_offset ^= size; -#else - /* move offset up to the next cache line */ - rx_buf->page_offset += size; -#endif -} - -/** - * ice_can_reuse_rx_page - Determine if page can be reused for another Rx - * @rx_buf: buffer containing the page - * - * If page is reusable, we have a green light for calling ice_reuse_rx_page, - * which will assign the current buffer to the buffer that next_to_alloc is - * pointing to; otherwise, the DMA mapping needs to be destroyed and - * page freed - */ -static bool ice_can_reuse_rx_page(struct ice_rx_buf *rx_buf) -{ - unsigned int pagecnt_bias = rx_buf->pagecnt_bias; - struct page *page = rx_buf->page; - - /* avoid re-using remote pages */ - if (unlikely(ice_page_is_reserved(page))) - return false; - -#if (PAGE_SIZE < 8192) - /* if we are only owner of page we can reuse it */ - if (unlikely((page_count(page) - pagecnt_bias) > 1)) - return false; -#else -#define ICE_LAST_OFFSET \ - (SKB_WITH_OVERHEAD(PAGE_SIZE) - ICE_RXBUF_2048) - if (rx_buf->page_offset > ICE_LAST_OFFSET) - return false; -#endif /* PAGE_SIZE < 8192) */ - - /* If we have drained the page fragment pool we need to update - * the pagecnt_bias and page count so that we fully restock the - * number of references the driver holds. - */ - if (unlikely(pagecnt_bias == 1)) { - page_ref_add(page, USHRT_MAX - 1); - rx_buf->pagecnt_bias = USHRT_MAX; - } - - return true; -} - -/** - * ice_add_rx_frag - Add contents of Rx buffer to sk_buff as a frag - * @rx_ring: Rx descriptor ring to transact packets on - * @rx_buf: buffer containing page to add - * @skb: sk_buff to place the data into - * @size: packet length from rx_desc - * - * This function will add the data contained in rx_buf->page to the skb. - * It will just attach the page as a frag to the skb. - * The function will then update the page offset. - */ -static void -ice_add_rx_frag(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf, - struct sk_buff *skb, unsigned int size) -{ -#if (PAGE_SIZE >= 8192) - unsigned int truesize = SKB_DATA_ALIGN(size + ice_rx_offset(rx_ring)); -#else - unsigned int truesize = ice_rx_pg_size(rx_ring) / 2; -#endif - - if (!size) - return; - skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buf->page, - rx_buf->page_offset, size, truesize); - - /* page is being used so we must update the page offset */ - ice_rx_buf_adjust_pg_offset(rx_buf, truesize); -} - -/** - * ice_reuse_rx_page - page flip buffer and store it back on the ring - * @rx_ring: Rx descriptor ring to store buffers on - * @old_buf: donor buffer to have page reused - * - * Synchronizes page for reuse by the adapter - */ -static void -ice_reuse_rx_page(struct ice_ring *rx_ring, struct ice_rx_buf *old_buf) -{ - u16 nta = rx_ring->next_to_alloc; - struct ice_rx_buf *new_buf; - - new_buf = &rx_ring->rx_buf[nta]; - - /* update, and store next to alloc */ - nta++; - rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0; - - /* Transfer page from old buffer to new buffer. - * Move each member individually to avoid possible store - * forwarding stalls and unnecessary copy of skb. - */ - new_buf->dma = old_buf->dma; - new_buf->page = old_buf->page; - new_buf->page_offset = old_buf->page_offset; - new_buf->pagecnt_bias = old_buf->pagecnt_bias; -} - -/** - * ice_get_rx_buf - Fetch Rx buffer and synchronize data for use - * @rx_ring: Rx descriptor ring to transact packets on - * @skb: skb to be used - * @size: size of buffer to add to skb - * - * This function will pull an Rx buffer from the ring and synchronize it - * for use by the CPU. - */ -static struct ice_rx_buf * -ice_get_rx_buf(struct ice_ring *rx_ring, struct sk_buff **skb, - const unsigned int size) -{ - struct ice_rx_buf *rx_buf; - - rx_buf = &rx_ring->rx_buf[rx_ring->next_to_clean]; - prefetchw(rx_buf->page); - *skb = rx_buf->skb; - - if (!size) - return rx_buf; - /* we are reusing so sync this buffer for CPU use */ - dma_sync_single_range_for_cpu(rx_ring->dev, rx_buf->dma, - rx_buf->page_offset, size, - DMA_FROM_DEVICE); - - /* We have pulled a buffer for use, so decrement pagecnt_bias */ - rx_buf->pagecnt_bias--; + u32 ntc = rx_ring->next_to_clean; + unsigned int total_rx_pkts = 0; + u32 cnt = rx_ring->count; - return rx_buf; -} - -/** - * ice_build_skb - Build skb around an existing buffer - * @rx_ring: Rx descriptor ring to transact packets on - * @rx_buf: Rx buffer to pull data from - * @xdp: xdp_buff pointing to the data - * - * This function builds an skb around an existing Rx buffer, taking care - * to set up the skb correctly and avoid any memcpy overhead. - */ -static struct sk_buff * -ice_build_skb(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf, - struct xdp_buff *xdp) -{ - u8 metasize = xdp->data - xdp->data_meta; -#if (PAGE_SIZE < 8192) - unsigned int truesize = ice_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; - - /* Prefetch first cache line of first page. If xdp->data_meta - * is unused, this points exactly as 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((void *)(xdp->data + L1_CACHE_BYTES)); -#endif - /* build an skb around the page buffer */ - skb = build_skb(xdp->data_hard_start, truesize); - if (unlikely(!skb)) - return NULL; - - /* must to record Rx queue, otherwise OS features such as - * symmetric queue won't work - */ - skb_record_rx_queue(skb, rx_ring->q_index); + while (likely(total_rx_pkts < ICE_DFLT_IRQ_WORK)) { + struct ice_vsi *ctrl_vsi = rx_ring->vsi; + union ice_32b_rx_flex_desc *rx_desc; + u16 stat_err_bits; - /* 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); + rx_desc = ICE_RX_DESC(rx_ring, ntc); - /* buffer is used by skb, update page_offset */ - ice_rx_buf_adjust_pg_offset(rx_buf, truesize); + stat_err_bits = BIT(ICE_RX_FLEX_DESC_STATUS0_DD_S); + if (!ice_test_staterr(rx_desc->wb.status_error0, stat_err_bits)) + break; - return skb; -} + dma_rmb(); -/** - * ice_construct_skb - Allocate skb and populate it - * @rx_ring: Rx descriptor ring to transact packets on - * @rx_buf: Rx buffer to pull data from - * @xdp: xdp_buff pointing to the data - * - * This function allocates an skb. It then populates it with the page - * data from the current receive descriptor, taking care to set up the - * skb correctly. - */ -static struct sk_buff * -ice_construct_skb(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf, - struct xdp_buff *xdp) -{ - unsigned int size = xdp->data_end - xdp->data; - unsigned int headlen; - struct sk_buff *skb; + if (ctrl_vsi->vf) + ice_vc_fdir_irq_handler(ctrl_vsi, rx_desc); - /* prefetch first cache line of first page */ - prefetch(xdp->data); -#if L1_CACHE_BYTES < 128 - prefetch((void *)(xdp->data + L1_CACHE_BYTES)); -#endif /* L1_CACHE_BYTES */ - - /* allocate a skb to store the frags */ - skb = __napi_alloc_skb(&rx_ring->q_vector->napi, ICE_RX_HDR_SIZE, - GFP_ATOMIC | __GFP_NOWARN); - if (unlikely(!skb)) - return NULL; - - skb_record_rx_queue(skb, rx_ring->q_index); - /* Determine available headroom for copy */ - headlen = size; - if (headlen > ICE_RX_HDR_SIZE) - headlen = eth_get_headlen(skb->dev, xdp->data, ICE_RX_HDR_SIZE); - - /* align pull length to size of long to optimize memcpy performance */ - memcpy(__skb_put(skb, headlen), xdp->data, ALIGN(headlen, - sizeof(long))); - - /* if we exhaust the linear part then add what is left as a frag */ - size -= headlen; - if (size) { -#if (PAGE_SIZE >= 8192) - unsigned int truesize = SKB_DATA_ALIGN(size); -#else - unsigned int truesize = ice_rx_pg_size(rx_ring) / 2; -#endif - skb_add_rx_frag(skb, 0, rx_buf->page, - rx_buf->page_offset + headlen, size, truesize); - /* buffer is used by skb, update page_offset */ - ice_rx_buf_adjust_pg_offset(rx_buf, truesize); - } else { - /* buffer is unused, reset bias back to rx_buf; data was copied - * onto skb's linear part so there's no need for adjusting - * page offset and we can reuse this buffer as-is - */ - rx_buf->pagecnt_bias++; + if (++ntc == cnt) + ntc = 0; + total_rx_pkts++; } - return skb; -} - -/** - * ice_put_rx_buf - Clean up used buffer and either recycle or free - * @rx_ring: Rx descriptor ring to transact packets on - * @rx_buf: Rx buffer to pull data from - * - * This function will update next_to_clean and then clean up the contents - * of the rx_buf. It will either recycle the buffer or unmap it and free - * the associated resources. - */ -static void ice_put_rx_buf(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf) -{ - u16 ntc = rx_ring->next_to_clean + 1; - - /* fetch, update, and store next to clean */ - ntc = (ntc < rx_ring->count) ? ntc : 0; rx_ring->next_to_clean = ntc; - - if (!rx_buf) - return; - - if (ice_can_reuse_rx_page(rx_buf)) { - /* hand second half of page back to the ring */ - ice_reuse_rx_page(rx_ring, rx_buf); - rx_ring->rx_stats.page_reuse_count++; - } else { - /* we are not reusing the buffer so unmap it */ - dma_unmap_page_attrs(rx_ring->dev, rx_buf->dma, - ice_rx_pg_size(rx_ring), DMA_FROM_DEVICE, - ICE_RX_DMA_ATTR); - __page_frag_cache_drain(rx_buf->page, rx_buf->pagecnt_bias); - } - - /* clear contents of buffer_info */ - rx_buf->page = NULL; - rx_buf->skb = NULL; -} - -/** - * ice_is_non_eop - process handling of non-EOP buffers - * @rx_ring: Rx ring being processed - * @rx_desc: Rx descriptor for current buffer - * @skb: Current socket buffer containing buffer in progress - * - * If the buffer is an EOP buffer, this function exits returning false, - * otherwise return true indicating that this is in fact a non-EOP buffer. - */ -static bool -ice_is_non_eop(struct ice_ring *rx_ring, union ice_32b_rx_flex_desc *rx_desc, - struct sk_buff *skb) -{ - /* if we are the last buffer then there is nothing else to do */ -#define ICE_RXD_EOF BIT(ICE_RX_FLEX_DESC_STATUS0_EOF_S) - if (likely(ice_test_staterr(rx_desc, ICE_RXD_EOF))) - return false; - - /* place skb in next buffer to be received */ - rx_ring->rx_buf[rx_ring->next_to_clean].skb = skb; - rx_ring->rx_stats.non_eop_descs++; - - return true; + ice_init_ctrl_rx_descs(rx_ring, ICE_DESC_UNUSED(rx_ring)); } /** @@ -1084,41 +941,47 @@ ice_is_non_eop(struct ice_ring *rx_ring, union ice_32b_rx_flex_desc *rx_desc, * * Returns amount of work completed */ -int ice_clean_rx_irq(struct ice_ring *rx_ring, int budget) +static int ice_clean_rx_irq(struct ice_rx_ring *rx_ring, int budget) { unsigned int total_rx_bytes = 0, total_rx_pkts = 0; - u16 cleaned_count = ICE_DESC_UNUSED(rx_ring); - unsigned int xdp_res, xdp_xmit = 0; + struct ice_tx_ring *xdp_ring = NULL; struct bpf_prog *xdp_prog = NULL; - struct xdp_buff xdp; + u32 ntc = rx_ring->next_to_clean; + LIBETH_XDP_ONSTACK_BUFF(xdp); + u32 cached_ntu, xdp_verdict; + u32 cnt = rx_ring->count; + u32 xdp_xmit = 0; bool failure; - xdp.rxq = &rx_ring->xdp_rxq; - /* Frame size depend on rx_ring setup when PAGE_SIZE=4K */ -#if (PAGE_SIZE < 8192) - xdp.frame_sz = ice_rx_frame_truesize(rx_ring, 0); -#endif + libeth_xdp_init_buff(xdp, &rx_ring->xdp, &rx_ring->xdp_rxq); + + xdp_prog = READ_ONCE(rx_ring->xdp_prog); + if (xdp_prog) { + xdp_ring = rx_ring->xdp_ring; + cached_ntu = xdp_ring->next_to_use; + } /* start the loop to process Rx packets bounded by 'budget' */ while (likely(total_rx_pkts < (unsigned int)budget)) { union ice_32b_rx_flex_desc *rx_desc; - struct ice_rx_buf *rx_buf; + struct libeth_fqe *rx_buf; struct sk_buff *skb; unsigned int size; u16 stat_err_bits; - u16 vlan_tag = 0; - u8 rx_ptype; + u16 vlan_tci; + bool rxe; /* get the Rx desc from Rx ring based on 'next_to_clean' */ - rx_desc = ICE_RX_DESC(rx_ring, rx_ring->next_to_clean); + rx_desc = ICE_RX_DESC(rx_ring, ntc); - /* status_error_len will always be zero for unused descriptors - * because it's cleared in cleanup, and overlaps with hdr_addr - * which is always zero because packet split isn't used, if the - * hardware wrote DD then it will be non-zero + /* + * The DD bit will always be zero for unused descriptors + * because it's cleared in cleanup or when setting the DMA + * address of the header buffer, which never uses the DD bit. + * If the hardware wrote the descriptor, it will be non-zero. */ stat_err_bits = BIT(ICE_RX_FLEX_DESC_STATUS0_DD_S); - if (!ice_test_staterr(rx_desc, stat_err_bits)) + if (!ice_test_staterr(rx_desc->wb.status_error0, stat_err_bits)) break; /* This memory barrier is needed to keep us from reading @@ -1127,337 +990,170 @@ int ice_clean_rx_irq(struct ice_ring *rx_ring, int budget) */ dma_rmb(); - if (rx_desc->wb.rxdid == FDIR_DESC_RXDID || !rx_ring->netdev) { - ice_put_rx_buf(rx_ring, NULL); - cleaned_count++; - continue; - } + ice_trace(clean_rx_irq, rx_ring, rx_desc); + + stat_err_bits = BIT(ICE_RX_FLEX_DESC_STATUS0_HBO_S) | + BIT(ICE_RX_FLEX_DESC_STATUS0_RXE_S); + rxe = ice_test_staterr(rx_desc->wb.status_error0, + stat_err_bits); + + if (!rx_ring->hdr_pp) + goto payload; + + size = le16_get_bits(rx_desc->wb.hdr_len_sph_flex_flags1, + ICE_RX_FLEX_DESC_HDR_LEN_M); + if (unlikely(rxe)) + size = 0; + + rx_buf = &rx_ring->hdr_fqes[ntc]; + libeth_xdp_process_buff(xdp, rx_buf, size); + rx_buf->netmem = 0; +payload: size = le16_to_cpu(rx_desc->wb.pkt_len) & ICE_RX_FLX_DESC_PKT_LEN_M; + if (unlikely(rxe)) + size = 0; /* retrieve a buffer from the ring */ - rx_buf = ice_get_rx_buf(rx_ring, &skb, size); + rx_buf = &rx_ring->rx_fqes[ntc]; + libeth_xdp_process_buff(xdp, rx_buf, size); - if (!size) { - xdp.data = NULL; - xdp.data_end = NULL; - xdp.data_hard_start = NULL; - xdp.data_meta = NULL; - goto construct_skb; - } + if (++ntc == cnt) + ntc = 0; - xdp.data = page_address(rx_buf->page) + rx_buf->page_offset; - xdp.data_hard_start = xdp.data - ice_rx_offset(rx_ring); - xdp.data_meta = xdp.data; - xdp.data_end = xdp.data + size; -#if (PAGE_SIZE > 4096) - /* At larger PAGE_SIZE, frame_sz depend on len size */ - xdp.frame_sz = ice_rx_frame_truesize(rx_ring, size); -#endif - - rcu_read_lock(); - xdp_prog = READ_ONCE(rx_ring->xdp_prog); - if (!xdp_prog) { - rcu_read_unlock(); - goto construct_skb; - } + /* skip if it is NOP desc */ + if (ice_is_non_eop(rx_ring, rx_desc) || unlikely(!xdp->data)) + continue; - xdp_res = ice_run_xdp(rx_ring, &xdp, xdp_prog); - rcu_read_unlock(); - if (!xdp_res) + xdp_verdict = ice_run_xdp(rx_ring, xdp, xdp_prog, xdp_ring, rx_desc); + if (xdp_verdict == ICE_XDP_PASS) goto construct_skb; - if (xdp_res & (ICE_XDP_TX | ICE_XDP_REDIR)) { - xdp_xmit |= xdp_res; - ice_rx_buf_adjust_pg_offset(rx_buf, xdp.frame_sz); - } else { - rx_buf->pagecnt_bias++; - } - total_rx_bytes += size; + + if (xdp_verdict & (ICE_XDP_TX | ICE_XDP_REDIR)) + xdp_xmit |= xdp_verdict; + total_rx_bytes += xdp_get_buff_len(&xdp->base); total_rx_pkts++; - cleaned_count++; - ice_put_rx_buf(rx_ring, rx_buf); + xdp->data = NULL; continue; + construct_skb: - if (skb) { - ice_add_rx_frag(rx_ring, rx_buf, skb, size); - } else if (likely(xdp.data)) { - if (ice_ring_uses_build_skb(rx_ring)) - skb = ice_build_skb(rx_ring, rx_buf, &xdp); - else - skb = ice_construct_skb(rx_ring, rx_buf, &xdp); - } + skb = xdp_build_skb_from_buff(&xdp->base); + xdp->data = NULL; + /* exit if we failed to retrieve a buffer */ if (!skb) { - rx_ring->rx_stats.alloc_buf_failed++; - if (rx_buf) - rx_buf->pagecnt_bias++; - break; - } - - ice_put_rx_buf(rx_ring, rx_buf); - cleaned_count++; - - /* skip if it is NOP desc */ - if (ice_is_non_eop(rx_ring, rx_desc, skb)) - continue; - - stat_err_bits = BIT(ICE_RX_FLEX_DESC_STATUS0_RXE_S); - if (unlikely(ice_test_staterr(rx_desc, stat_err_bits))) { - dev_kfree_skb_any(skb); + libeth_xdp_return_buff_slow(xdp); + rx_ring->ring_stats->rx_stats.alloc_buf_failed++; continue; } - stat_err_bits = BIT(ICE_RX_FLEX_DESC_STATUS0_L2TAG1P_S); - if (ice_test_staterr(rx_desc, stat_err_bits)) - vlan_tag = le16_to_cpu(rx_desc->wb.l2tag1); - - /* pad the skb if needed, to make a valid ethernet frame */ - if (eth_skb_pad(skb)) { - skb = NULL; - continue; - } + vlan_tci = ice_get_vlan_tci(rx_desc); /* probably a little skewed due to removing CRC */ total_rx_bytes += skb->len; /* populate checksum, VLAN, and protocol */ - rx_ptype = le16_to_cpu(rx_desc->wb.ptype_flex_flags0) & - ICE_RX_FLEX_DESC_PTYPE_M; - - ice_process_skb_fields(rx_ring, rx_desc, skb, rx_ptype); + ice_process_skb_fields(rx_ring, rx_desc, skb); + ice_trace(clean_rx_irq_indicate, rx_ring, rx_desc, skb); /* send completed skb up the stack */ - ice_receive_skb(rx_ring, skb, vlan_tag); + ice_receive_skb(rx_ring, skb, vlan_tci); /* update budget accounting */ total_rx_pkts++; } + rx_ring->next_to_clean = ntc; /* return up to cleaned_count buffers to hardware */ - failure = ice_alloc_rx_bufs(rx_ring, cleaned_count); + failure = ice_alloc_rx_bufs(rx_ring, ICE_DESC_UNUSED(rx_ring)); - if (xdp_prog) - ice_finalize_xdp_rx(rx_ring, xdp_xmit); + if (xdp_xmit) + ice_finalize_xdp_rx(xdp_ring, xdp_xmit, cached_ntu); - ice_update_rx_ring_stats(rx_ring, total_rx_pkts, total_rx_bytes); + libeth_xdp_save_buff(&rx_ring->xdp, xdp); + + if (rx_ring->ring_stats) + ice_update_rx_ring_stats(rx_ring, total_rx_pkts, + total_rx_bytes); /* guarantee a trip back through this routine if there was a failure */ return failure ? budget : (int)total_rx_pkts; } -/** - * ice_adjust_itr_by_size_and_speed - Adjust ITR based on current traffic - * @port_info: port_info structure containing the current link speed - * @avg_pkt_size: average size of Tx or Rx packets based on clean routine - * @itr: ITR value to update - * - * Calculate how big of an increment should be applied to the ITR value passed - * in based on wmem_default, SKB overhead, Ethernet overhead, and the current - * link speed. - * - * The following is a calculation derived from: - * wmem_default / (size + overhead) = desired_pkts_per_int - * rate / bits_per_byte / (size + Ethernet overhead) = pkt_rate - * (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value - * - * Assuming wmem_default is 212992 and overhead is 640 bytes per - * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the - * formula down to: - * - * wmem_default * bits_per_byte * usecs_per_sec pkt_size + 24 - * ITR = -------------------------------------------- * -------------- - * rate pkt_size + 640 - */ -static unsigned int -ice_adjust_itr_by_size_and_speed(struct ice_port_info *port_info, - unsigned int avg_pkt_size, - unsigned int itr) -{ - switch (port_info->phy.link_info.link_speed) { - case ICE_AQ_LINK_SPEED_100GB: - itr += DIV_ROUND_UP(17 * (avg_pkt_size + 24), - avg_pkt_size + 640); - break; - case ICE_AQ_LINK_SPEED_50GB: - itr += DIV_ROUND_UP(34 * (avg_pkt_size + 24), - avg_pkt_size + 640); - break; - case ICE_AQ_LINK_SPEED_40GB: - itr += DIV_ROUND_UP(43 * (avg_pkt_size + 24), - avg_pkt_size + 640); - break; - case ICE_AQ_LINK_SPEED_25GB: - itr += DIV_ROUND_UP(68 * (avg_pkt_size + 24), - avg_pkt_size + 640); - break; - case ICE_AQ_LINK_SPEED_20GB: - itr += DIV_ROUND_UP(85 * (avg_pkt_size + 24), - avg_pkt_size + 640); - break; - case ICE_AQ_LINK_SPEED_10GB: - default: - itr += DIV_ROUND_UP(170 * (avg_pkt_size + 24), - avg_pkt_size + 640); - break; - } - - if ((itr & ICE_ITR_MASK) > ICE_ITR_ADAPTIVE_MAX_USECS) { - itr &= ICE_ITR_ADAPTIVE_LATENCY; - itr += ICE_ITR_ADAPTIVE_MAX_USECS; - } - - return itr; -} - -/** - * ice_update_itr - update the adaptive ITR value based on statistics - * @q_vector: structure containing interrupt and ring information - * @rc: structure containing ring performance data - * - * 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 void -ice_update_itr(struct ice_q_vector *q_vector, struct ice_ring_container *rc) +static void __ice_update_sample(struct ice_q_vector *q_vector, + struct ice_ring_container *rc, + struct dim_sample *sample, + bool is_tx) { - unsigned long next_update = jiffies; - unsigned int packets, bytes, itr; - bool container_is_rx; - - if (!rc->ring || !ITR_IS_DYNAMIC(rc->itr_setting)) - return; - - /* If itr_countdown is set it means we programmed an ITR within - * the last 4 interrupt cycles. This has a side effect of us - * potentially firing an early interrupt. In order to work around - * this we need to throw out any data received for a few - * interrupts following the update. - */ - if (q_vector->itr_countdown) { - itr = rc->target_itr; - goto clear_counts; - } - - container_is_rx = (&q_vector->rx == rc); - /* For Rx we want to push the delay up and default to low latency. - * for Tx we want to pull the delay down and default to high latency. - */ - itr = container_is_rx ? - ICE_ITR_ADAPTIVE_MIN_USECS | ICE_ITR_ADAPTIVE_LATENCY : - ICE_ITR_ADAPTIVE_MAX_USECS | ICE_ITR_ADAPTIVE_LATENCY; - - /* If we didn't update within up to 1 - 2 jiffies we can assume - * that either packets are coming in so slow there hasn't been - * any work, or that there is so much work that NAPI is dealing - * with interrupt moderation and we don't need to do anything. - */ - if (time_after(next_update, rc->next_update)) - goto clear_counts; + u64 packets = 0, bytes = 0; - prefetch(q_vector->vsi->port_info); + if (is_tx) { + struct ice_tx_ring *tx_ring; - packets = rc->total_pkts; - bytes = rc->total_bytes; + ice_for_each_tx_ring(tx_ring, *rc) { + struct ice_ring_stats *ring_stats; - if (container_is_rx) { - /* If Rx there are 1 to 4 packets and bytes are less than - * 9000 assume insufficient data to use bulk rate limiting - * approach unless Tx is already in bulk rate limiting. We - * are likely latency driven. - */ - if (packets && packets < 4 && bytes < 9000 && - (q_vector->tx.target_itr & ICE_ITR_ADAPTIVE_LATENCY)) { - itr = ICE_ITR_ADAPTIVE_LATENCY; - goto adjust_by_size_and_speed; + ring_stats = tx_ring->ring_stats; + if (!ring_stats) + continue; + packets += ring_stats->stats.pkts; + bytes += ring_stats->stats.bytes; } - } else if (packets < 4) { - /* If we have Tx and Rx ITR maxed and Tx ITR is running in - * bulk mode and we are receiving 4 or fewer packets just - * reset the ITR_ADAPTIVE_LATENCY bit for latency mode so - * that the Rx can relax. - */ - if (rc->target_itr == ICE_ITR_ADAPTIVE_MAX_USECS && - (q_vector->rx.target_itr & ICE_ITR_MASK) == - ICE_ITR_ADAPTIVE_MAX_USECS) - goto clear_counts; - } else if (packets > 32) { - /* If we have processed over 32 packets in a single interrupt - * for Tx assume we need to switch over to "bulk" mode. - */ - rc->target_itr &= ~ICE_ITR_ADAPTIVE_LATENCY; - } + } else { + struct ice_rx_ring *rx_ring; - /* We have no packets to actually measure against. This means - * either one of the other queues on this vector is active or - * we are a Tx queue doing TSO with too high of an interrupt rate. - * - * Between 4 and 56 we can assume that our current interrupt delay - * is only slightly too low. As such we should increase it by a small - * fixed amount. - */ - if (packets < 56) { - itr = rc->target_itr + ICE_ITR_ADAPTIVE_MIN_INC; - if ((itr & ICE_ITR_MASK) > ICE_ITR_ADAPTIVE_MAX_USECS) { - itr &= ICE_ITR_ADAPTIVE_LATENCY; - itr += ICE_ITR_ADAPTIVE_MAX_USECS; + ice_for_each_rx_ring(rx_ring, *rc) { + struct ice_ring_stats *ring_stats; + + ring_stats = rx_ring->ring_stats; + if (!ring_stats) + continue; + packets += ring_stats->stats.pkts; + bytes += ring_stats->stats.bytes; } - goto clear_counts; } - if (packets <= 256) { - itr = min(q_vector->tx.current_itr, q_vector->rx.current_itr); - itr &= ICE_ITR_MASK; + dim_update_sample(q_vector->total_events, packets, bytes, sample); + sample->comp_ctr = 0; - /* Between 56 and 112 is our "goldilocks" zone where we are - * working out "just right". Just report that our current - * ITR is good for us. - */ - if (packets <= 112) - goto clear_counts; - - /* If packet count is 128 or greater we are likely looking - * at a slight overrun of the delay we want. Try halving - * our delay to see if that will cut the number of packets - * in half per interrupt. - */ - itr >>= 1; - itr &= ICE_ITR_MASK; - if (itr < ICE_ITR_ADAPTIVE_MIN_USECS) - itr = ICE_ITR_ADAPTIVE_MIN_USECS; - - goto clear_counts; - } - - /* The paths below assume we are dealing with a bulk ITR since - * number of packets is greater than 256. We are just going to have - * to compute a value and try to bring the count under control, - * though for smaller packet sizes there isn't much we can do as - * NAPI polling will likely be kicking in sooner rather than later. + /* if dim settings get stale, like when not updated for 1 + * second or longer, force it to start again. This addresses the + * frequent case of an idle queue being switched to by the + * scheduler. The 1,000 here means 1,000 milliseconds. */ - itr = ICE_ITR_ADAPTIVE_BULK; + if (ktime_ms_delta(sample->time, rc->dim.start_sample.time) >= 1000) + rc->dim.state = DIM_START_MEASURE; +} -adjust_by_size_and_speed: +/** + * ice_net_dim - Update net DIM algorithm + * @q_vector: the vector associated with the interrupt + * + * Create a DIM sample and notify net_dim() so that it can possibly decide + * a new ITR value based on incoming packets, bytes, and interrupts. + * + * This function is a no-op if the ring is not configured to dynamic ITR. + */ +static void ice_net_dim(struct ice_q_vector *q_vector) +{ + struct ice_ring_container *tx = &q_vector->tx; + struct ice_ring_container *rx = &q_vector->rx; - /* based on checks above packets cannot be 0 so division is safe */ - itr = ice_adjust_itr_by_size_and_speed(q_vector->vsi->port_info, - bytes / packets, itr); + if (ITR_IS_DYNAMIC(tx)) { + struct dim_sample dim_sample; -clear_counts: - /* write back value */ - rc->target_itr = itr; + __ice_update_sample(q_vector, tx, &dim_sample, true); + net_dim(&tx->dim, &dim_sample); + } - /* next update should occur within next jiffy */ - rc->next_update = next_update + 1; + if (ITR_IS_DYNAMIC(rx)) { + struct dim_sample dim_sample; - rc->total_bytes = 0; - rc->total_pkts = 0; + __ice_update_sample(q_vector, rx, &dim_sample, false); + net_dim(&rx->dim, &dim_sample); + } } /** @@ -1481,85 +1177,45 @@ static u32 ice_buildreg_itr(u16 itr_idx, u16 itr) (itr << (GLINT_DYN_CTL_INTERVAL_S - ICE_ITR_GRAN_S)); } -/* The act of updating the ITR will cause it to immediately trigger. In order - * to prevent this from throwing off adaptive update statistics we defer the - * update so that it can only happen so often. So after either Tx or Rx are - * updated we make the adaptive scheme wait until either the ITR completely - * expires via the next_update expiration or we have been through at least - * 3 interrupts. - */ -#define ITR_COUNTDOWN_START 3 - /** - * ice_update_ena_itr - Update ITR and re-enable MSIX interrupt - * @q_vector: q_vector for which ITR is being updated and interrupt enabled + * ice_enable_interrupt - re-enable MSI-X interrupt + * @q_vector: the vector associated with the interrupt to enable + * + * If the VSI is down, the interrupt will not be re-enabled. Also, + * when enabling the interrupt always reset the wb_on_itr to false + * and trigger a software interrupt to clean out internal state. */ -static void ice_update_ena_itr(struct ice_q_vector *q_vector) +static void ice_enable_interrupt(struct ice_q_vector *q_vector) { - struct ice_ring_container *tx = &q_vector->tx; - struct ice_ring_container *rx = &q_vector->rx; struct ice_vsi *vsi = q_vector->vsi; + bool wb_en = q_vector->wb_on_itr; u32 itr_val; - /* when exiting WB_ON_ITR lets set a low ITR value and trigger - * interrupts to expire right away in case we have more work ready to go - * already - */ - if (q_vector->itr_countdown == ICE_IN_WB_ON_ITR_MODE) { - itr_val = ice_buildreg_itr(rx->itr_idx, ICE_WB_ON_ITR_USECS); - wr32(&vsi->back->hw, GLINT_DYN_CTL(q_vector->reg_idx), itr_val); - /* set target back to last user set value */ - rx->target_itr = rx->itr_setting; - /* set current to what we just wrote and dynamic if needed */ - rx->current_itr = ICE_WB_ON_ITR_USECS | - (rx->itr_setting & ICE_ITR_DYNAMIC); - /* allow normal interrupt flow to start */ - q_vector->itr_countdown = 0; + if (test_bit(ICE_DOWN, vsi->state)) return; - } - /* This will do nothing if dynamic updates are not enabled */ - ice_update_itr(q_vector, tx); - ice_update_itr(q_vector, rx); - - /* This block of logic allows us to get away with only updating - * one ITR value with each interrupt. The idea is to perform a - * pseudo-lazy update with the following criteria. - * - * 1. Rx is given higher priority than Tx if both are in same state - * 2. If we must reduce an ITR that is given highest priority. - * 3. We then give priority to increasing ITR based on amount. + /* trigger an ITR delayed software interrupt when exiting busy poll, to + * make sure to catch any pending cleanups that might have been missed + * due to interrupt state transition. If busy poll or poll isn't + * enabled, then don't update ITR, and just enable the interrupt. */ - if (rx->target_itr < rx->current_itr) { - /* Rx ITR needs to be reduced, this is highest priority */ - itr_val = ice_buildreg_itr(rx->itr_idx, rx->target_itr); - rx->current_itr = rx->target_itr; - q_vector->itr_countdown = ITR_COUNTDOWN_START; - } else if ((tx->target_itr < tx->current_itr) || - ((rx->target_itr - rx->current_itr) < - (tx->target_itr - tx->current_itr))) { - /* Tx ITR needs to be reduced, this is second priority - * Tx ITR needs to be increased more than Rx, fourth priority - */ - itr_val = ice_buildreg_itr(tx->itr_idx, tx->target_itr); - tx->current_itr = tx->target_itr; - q_vector->itr_countdown = ITR_COUNTDOWN_START; - } else if (rx->current_itr != rx->target_itr) { - /* Rx ITR needs to be increased, third priority */ - itr_val = ice_buildreg_itr(rx->itr_idx, rx->target_itr); - rx->current_itr = rx->target_itr; - q_vector->itr_countdown = ITR_COUNTDOWN_START; - } else { - /* Still have to re-enable the interrupts */ + if (!wb_en) { itr_val = ice_buildreg_itr(ICE_ITR_NONE, 0); - if (q_vector->itr_countdown) - q_vector->itr_countdown--; - } + } else { + q_vector->wb_on_itr = false; - if (!test_bit(__ICE_DOWN, q_vector->vsi->state)) - wr32(&q_vector->vsi->back->hw, - GLINT_DYN_CTL(q_vector->reg_idx), - itr_val); + /* do two things here with a single write. Set up the third ITR + * index to be used for software interrupt moderation, and then + * trigger a software interrupt with a rate limit of 20K on + * software interrupts, this will help avoid high interrupt + * loads due to frequently polling and exiting polling. + */ + itr_val = ice_buildreg_itr(ICE_IDX_ITR2, ICE_ITR_20K); + itr_val |= GLINT_DYN_CTL_SWINT_TRIG_M | + ICE_IDX_ITR2 << GLINT_DYN_CTL_SW_ITR_INDX_S | + GLINT_DYN_CTL_SW_ITR_INDX_ENA_M; + } + wr32(&vsi->back->hw, GLINT_DYN_CTL(q_vector->reg_idx), itr_val); } /** @@ -1569,32 +1225,31 @@ static void ice_update_ena_itr(struct ice_q_vector *q_vector) * We need to tell hardware to write-back completed descriptors even when * interrupts are disabled. Descriptors will be written back on cache line * boundaries without WB_ON_ITR enabled, but if we don't enable WB_ON_ITR - * descriptors may not be written back if they don't fill a cache line until the - * next interrupt. + * descriptors may not be written back if they don't fill a cache line until + * the next interrupt. * - * This sets the write-back frequency to 2 microseconds as that is the minimum - * value that's not 0 due to ITR granularity. Also, set the INTENA_MSK bit to - * make sure hardware knows we aren't meddling with the INTENA_M bit. + * This sets the write-back frequency to whatever was set previously for the + * ITR indices. Also, set the INTENA_MSK bit to make sure hardware knows we + * aren't meddling with the INTENA_M bit. */ static void ice_set_wb_on_itr(struct ice_q_vector *q_vector) { struct ice_vsi *vsi = q_vector->vsi; - /* already in WB_ON_ITR mode no need to change it */ - if (q_vector->itr_countdown == ICE_IN_WB_ON_ITR_MODE) + /* already in wb_on_itr mode no need to change it */ + if (q_vector->wb_on_itr) return; - if (q_vector->num_ring_rx) - wr32(&vsi->back->hw, GLINT_DYN_CTL(q_vector->reg_idx), - ICE_GLINT_DYN_CTL_WB_ON_ITR(ICE_WB_ON_ITR_USECS, - ICE_RX_ITR)); - - if (q_vector->num_ring_tx) - wr32(&vsi->back->hw, GLINT_DYN_CTL(q_vector->reg_idx), - ICE_GLINT_DYN_CTL_WB_ON_ITR(ICE_WB_ON_ITR_USECS, - ICE_TX_ITR)); + /* use previously set ITR values for all of the ITR indices by + * specifying ICE_ITR_NONE, which will vary in adaptive (AIM) mode and + * be static in non-adaptive mode (user configured) + */ + wr32(&vsi->back->hw, GLINT_DYN_CTL(q_vector->reg_idx), + FIELD_PREP(GLINT_DYN_CTL_ITR_INDX_M, ICE_ITR_NONE) | + FIELD_PREP(GLINT_DYN_CTL_INTENA_MSK_M, 1) | + FIELD_PREP(GLINT_DYN_CTL_WB_ON_ITR_M, 1)); - q_vector->itr_countdown = ICE_IN_WB_ON_ITR_MODE; + q_vector->wb_on_itr = true; } /** @@ -1610,18 +1265,25 @@ int ice_napi_poll(struct napi_struct *napi, int budget) { struct ice_q_vector *q_vector = container_of(napi, struct ice_q_vector, napi); + struct ice_tx_ring *tx_ring; + struct ice_rx_ring *rx_ring; bool clean_complete = true; - struct ice_ring *ring; int budget_per_ring; int work_done = 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. */ - ice_for_each_ring(ring, q_vector->tx) { - bool wd = ring->xsk_umem ? - ice_clean_tx_irq_zc(ring, budget) : - ice_clean_tx_irq(ring, budget); + ice_for_each_tx_ring(tx_ring, q_vector->tx) { + struct xsk_buff_pool *xsk_pool = READ_ONCE(tx_ring->xsk_pool); + bool wd; + + if (xsk_pool) + wd = ice_xmit_zc(tx_ring, xsk_pool); + else if (ice_ring_is_xdp(tx_ring)) + wd = true; + else + wd = ice_clean_tx_irq(tx_ring, budget); if (!wd) clean_complete = false; @@ -1642,16 +1304,17 @@ int ice_napi_poll(struct napi_struct *napi, int budget) /* Max of 1 Rx ring in this q_vector so give it the budget */ budget_per_ring = budget; - ice_for_each_ring(ring, q_vector->rx) { + ice_for_each_rx_ring(rx_ring, q_vector->rx) { + struct xsk_buff_pool *xsk_pool = READ_ONCE(rx_ring->xsk_pool); int cleaned; /* A dedicated path for zero-copy allows making a single * comparison in the irq context instead of many inside the * ice_clean_rx_irq function and makes the codebase cleaner. */ - cleaned = ring->xsk_umem ? - ice_clean_rx_irq_zc(ring, budget_per_ring) : - ice_clean_rx_irq(ring, budget_per_ring); + cleaned = rx_ring->xsk_pool ? + ice_clean_rx_irq_zc(rx_ring, xsk_pool, budget_per_ring) : + ice_clean_rx_irq(rx_ring, budget_per_ring); work_done += cleaned; /* if we clean as many as budgeted, we must not be done */ if (cleaned >= budget_per_ring) @@ -1659,16 +1322,23 @@ int ice_napi_poll(struct napi_struct *napi, int budget) } /* If work not completed, return budget and polling will return */ - if (!clean_complete) + if (!clean_complete) { + /* Set the writeback on ITR so partial completions of + * cache-lines will still continue even if we're polling. + */ + ice_set_wb_on_itr(q_vector); return budget; + } /* Exit the polling mode, but don't re-enable interrupts if stack might * poll us due to busy-polling */ - if (likely(napi_complete_done(napi, work_done))) - ice_update_ena_itr(q_vector); - else + if (napi_complete_done(napi, work_done)) { + ice_net_dim(q_vector); + ice_enable_interrupt(q_vector); + } else { ice_set_wb_on_itr(q_vector); + } return min_t(int, work_done, budget - 1); } @@ -1680,9 +1350,9 @@ int ice_napi_poll(struct napi_struct *napi, int budget) * * Returns -EBUSY if a stop is needed, else 0 */ -static int __ice_maybe_stop_tx(struct ice_ring *tx_ring, unsigned int size) +static int __ice_maybe_stop_tx(struct ice_tx_ring *tx_ring, unsigned int size) { - netif_stop_subqueue(tx_ring->netdev, tx_ring->q_index); + netif_tx_stop_queue(txring_txq(tx_ring)); /* Memory barrier before checking head and tail */ smp_mb(); @@ -1690,9 +1360,9 @@ static int __ice_maybe_stop_tx(struct ice_ring *tx_ring, unsigned int size) if (likely(ICE_DESC_UNUSED(tx_ring) < size)) return -EBUSY; - /* A reprieve! - use start_subqueue because it doesn't call schedule */ - netif_start_subqueue(tx_ring->netdev, tx_ring->q_index); - ++tx_ring->tx_stats.restart_q; + /* A reprieve! - use start_queue because it doesn't call schedule */ + netif_tx_start_queue(txring_txq(tx_ring)); + ++tx_ring->ring_stats->tx_stats.restart_q; return 0; } @@ -1703,7 +1373,7 @@ static int __ice_maybe_stop_tx(struct ice_ring *tx_ring, unsigned int size) * * Returns 0 if stop is not needed */ -static int ice_maybe_stop_tx(struct ice_ring *tx_ring, unsigned int size) +static int ice_maybe_stop_tx(struct ice_tx_ring *tx_ring, unsigned int size) { if (likely(ICE_DESC_UNUSED(tx_ring) >= size)) return 0; @@ -1722,7 +1392,7 @@ static int ice_maybe_stop_tx(struct ice_ring *tx_ring, unsigned int size) * it and the length into the transmit descriptor. */ static void -ice_tx_map(struct ice_ring *tx_ring, struct ice_tx_buf *first, +ice_tx_map(struct ice_tx_ring *tx_ring, struct ice_tx_buf *first, struct ice_tx_offload_params *off) { u64 td_offset, td_tag, td_cmd; @@ -1733,6 +1403,7 @@ ice_tx_map(struct ice_ring *tx_ring, struct ice_tx_buf *first, struct sk_buff *skb; skb_frag_t *frag; dma_addr_t dma; + bool kick; td_tag = off->td_l2tag1; td_cmd = off->td_cmd; @@ -1746,8 +1417,7 @@ ice_tx_map(struct ice_ring *tx_ring, struct ice_tx_buf *first, if (first->tx_flags & ICE_TX_FLAGS_HW_VLAN) { td_cmd |= (u64)ICE_TX_DESC_CMD_IL2TAG1; - td_tag = (first->tx_flags & ICE_TX_FLAGS_VLAN_M) >> - ICE_TX_FLAGS_VLAN_S; + td_tag = first->vid; } dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE); @@ -1812,11 +1482,9 @@ ice_tx_map(struct ice_ring *tx_ring, struct ice_tx_buf *first, DMA_TO_DEVICE); tx_buf = &tx_ring->tx_buf[i]; + tx_buf->type = ICE_TX_BUF_FRAG; } - /* record bytecount for BQL */ - netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount); - /* record SW timestamp if HW timestamp is not available */ skb_tx_timestamp(first->skb); @@ -1845,9 +1513,48 @@ ice_tx_map(struct ice_ring *tx_ring, struct ice_tx_buf *first, ice_maybe_stop_tx(tx_ring, DESC_NEEDED); /* notify HW of packet */ - if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more()) - writel(i, tx_ring->tail); + kick = __netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount, + netdev_xmit_more()); + if (!kick) + return; + if (ice_is_txtime_cfg(tx_ring)) { + struct ice_tstamp_ring *tstamp_ring = tx_ring->tstamp_ring; + u32 tstamp_count = tstamp_ring->count; + u32 j = tstamp_ring->next_to_use; + struct ice_ts_desc *ts_desc; + struct timespec64 ts; + u32 tstamp; + + ts = ktime_to_timespec64(first->skb->tstamp); + tstamp = ts.tv_nsec >> ICE_TXTIME_CTX_RESOLUTION_128NS; + + ts_desc = ICE_TS_DESC(tstamp_ring, j); + ts_desc->tx_desc_idx_tstamp = ice_build_tstamp_desc(i, tstamp); + + j++; + if (j == tstamp_count) { + u32 fetch = tstamp_count - tx_ring->count; + + j = 0; + + /* To prevent an MDD, when wrapping the tstamp ring + * create additional TS descriptors equal to the number + * of the fetch TS descriptors value. HW will merge the + * TS descriptors with the same timestamp value into a + * single descriptor. + */ + for (; j < fetch; j++) { + ts_desc = ICE_TS_DESC(tstamp_ring, j); + ts_desc->tx_desc_idx_tstamp = + ice_build_tstamp_desc(i, tstamp); + } + } + tstamp_ring->next_to_use = j; + writel_relaxed(j, tstamp_ring->tail); + } else { + writel_relaxed(i, tx_ring->tail); + } return; dma_error: @@ -1875,6 +1582,7 @@ dma_error: static int ice_tx_csum(struct ice_tx_buf *first, struct ice_tx_offload_params *off) { + const struct ice_tx_ring *tx_ring = off->tx_ring; u32 l4_len = 0, l3_len = 0, l2_len = 0; struct sk_buff *skb = first->skb; union { @@ -1894,18 +1602,26 @@ int ice_tx_csum(struct ice_tx_buf *first, struct ice_tx_offload_params *off) if (skb->ip_summed != CHECKSUM_PARTIAL) return 0; - ip.hdr = skb_network_header(skb); - l4.hdr = skb_transport_header(skb); + protocol = vlan_get_protocol(skb); + + if (eth_p_mpls(protocol)) { + ip.hdr = skb_inner_network_header(skb); + l4.hdr = skb_checksum_start(skb); + } else { + ip.hdr = skb_network_header(skb); + l4.hdr = skb_transport_header(skb); + } /* compute outer L2 header size */ l2_len = ip.hdr - skb->data; offset = (l2_len / 2) << ICE_TX_DESC_LEN_MACLEN_S; - protocol = vlan_get_protocol(skb); - - if (protocol == htons(ETH_P_IP)) + /* set the tx_flags to indicate the IP protocol type. this is + * required so that checksum header computation below is accurate. + */ + if (ip.v4->version == 4) first->tx_flags |= ICE_TX_FLAGS_IPV4; - else if (protocol == htons(ETH_P_IPV6)) + else if (ip.v6->version == 6) first->tx_flags |= ICE_TX_FLAGS_IPV6; if (skb->encapsulation) { @@ -1919,12 +1635,15 @@ int ice_tx_csum(struct ice_tx_buf *first, struct ice_tx_offload_params *off) ICE_TX_CTX_EIPT_IPV4_NO_CSUM; l4_proto = ip.v4->protocol; } else if (first->tx_flags & ICE_TX_FLAGS_IPV6) { + int ret; + tunnel |= ICE_TX_CTX_EIPT_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); + ret = ipv6_skip_exthdr(skb, exthdr - skb->data, + &l4_proto, &frag_off); + if (ret < 0) + return -1; } /* define outer transport */ @@ -2013,6 +1732,30 @@ int ice_tx_csum(struct ice_tx_buf *first, struct ice_tx_offload_params *off) l3_len = l4.hdr - ip.hdr; offset |= (l3_len / 4) << ICE_TX_DESC_LEN_IPLEN_S; + if ((tx_ring->netdev->features & NETIF_F_HW_CSUM) && + !(first->tx_flags & ICE_TX_FLAGS_TSO) && + !skb_csum_is_sctp(skb)) { + /* Set GCS */ + u16 csum_start = (skb->csum_start - skb->mac_header) / 2; + u16 csum_offset = skb->csum_offset / 2; + u16 gcs_params; + + gcs_params = FIELD_PREP(ICE_TX_GCS_DESC_START_M, csum_start) | + FIELD_PREP(ICE_TX_GCS_DESC_OFFSET_M, csum_offset) | + FIELD_PREP(ICE_TX_GCS_DESC_TYPE_M, + ICE_TX_GCS_DESC_CSUM_PSH); + + /* Unlike legacy HW checksums, GCS requires a context + * descriptor. + */ + off->cd_qw1 |= ICE_TX_DESC_DTYPE_CTX; + off->cd_gcs_params = gcs_params; + /* Fill out CSO info in data descriptors */ + off->td_offset |= offset; + off->td_cmd |= cmd; + return 1; + } + /* Enable L4 checksum offloads */ switch (l4_proto) { case IPPROTO_TCP: @@ -2055,7 +1798,7 @@ int ice_tx_csum(struct ice_tx_buf *first, struct ice_tx_offload_params *off) * related to VLAN tagging for the HW, such as VLAN, DCB, etc. */ static void -ice_tx_prepare_vlan_flags(struct ice_ring *tx_ring, struct ice_tx_buf *first) +ice_tx_prepare_vlan_flags(struct ice_tx_ring *tx_ring, struct ice_tx_buf *first) { struct sk_buff *skb = first->skb; @@ -2063,12 +1806,16 @@ ice_tx_prepare_vlan_flags(struct ice_ring *tx_ring, struct ice_tx_buf *first) if (!skb_vlan_tag_present(skb) && eth_type_vlan(skb->protocol)) return; - /* currently, we always assume 802.1Q for VLAN insertion as VLAN - * insertion for 802.1AD is not supported + /* the VLAN ethertype/tpid is determined by VSI configuration and netdev + * feature flags, which the driver only allows either 802.1Q or 802.1ad + * VLAN offloads exclusively so we only care about the VLAN ID here */ if (skb_vlan_tag_present(skb)) { - first->tx_flags |= skb_vlan_tag_get(skb) << ICE_TX_FLAGS_VLAN_S; - first->tx_flags |= ICE_TX_FLAGS_HW_VLAN; + first->vid = skb_vlan_tag_get(skb); + if (tx_ring->flags & ICE_TX_FLAGS_RING_VLAN_L2TAG2) + first->tx_flags |= ICE_TX_FLAGS_HW_OUTER_SINGLE_VLAN; + else + first->tx_flags |= ICE_TX_FLAGS_HW_VLAN; } ice_tx_prepare_vlan_flags_dcb(tx_ring, first); @@ -2096,6 +1843,7 @@ int ice_tso(struct ice_tx_buf *first, struct ice_tx_offload_params *off) unsigned char *hdr; } l4; u64 cd_mss, cd_tso_len; + __be16 protocol; u32 paylen; u8 l4_start; int err; @@ -2110,9 +1858,13 @@ int ice_tso(struct ice_tx_buf *first, struct ice_tx_offload_params *off) if (err < 0) return err; - /* cppcheck-suppress unreadVariable */ - ip.hdr = skb_network_header(skb); - l4.hdr = skb_transport_header(skb); + protocol = vlan_get_protocol(skb); + + if (eth_p_mpls(protocol)) + ip.hdr = skb_inner_network_header(skb); + else + ip.hdr = skb_network_header(skb); + l4.hdr = skb_checksum_start(skb); /* initialize outer IP header fields */ if (ip.v4->version == 4) { @@ -2142,8 +1894,6 @@ int ice_tso(struct ice_tx_buf *first, struct ice_tx_offload_params *off) } /* reset pointers to inner headers */ - - /* cppcheck-suppress unreadVariable */ ip.hdr = skb_inner_network_header(skb); l4.hdr = skb_inner_transport_header(skb); @@ -2294,10 +2044,30 @@ static bool __ice_chk_linearize(struct sk_buff *skb) /* Walk through fragments adding latest fragment, testing it, and * then removing stale fragments from the sum. */ - stale = &skb_shinfo(skb)->frags[0]; - for (;;) { + for (stale = &skb_shinfo(skb)->frags[0];; stale++) { + int stale_size = skb_frag_size(stale); + sum += skb_frag_size(frag++); + /* The stale fragment may present us with a smaller + * descriptor than the actual fragment size. To account + * for that we need to remove all the data on the front and + * figure out what the remainder would be in the last + * descriptor associated with the fragment. + */ + if (stale_size > ICE_MAX_DATA_PER_TXD) { + int align_pad = -(skb_frag_off(stale)) & + (ICE_MAX_READ_REQ_SIZE - 1); + + sum -= align_pad; + stale_size -= align_pad; + + do { + sum -= ICE_MAX_DATA_PER_TXD_ALIGNED; + stale_size -= ICE_MAX_DATA_PER_TXD_ALIGNED; + } while (stale_size > ICE_MAX_DATA_PER_TXD); + } + /* if sum is negative we failed to make sufficient progress */ if (sum < 0) return true; @@ -2305,7 +2075,7 @@ static bool __ice_chk_linearize(struct sk_buff *skb) if (!nr_frags--) break; - sum -= skb_frag_size(stale++); + sum -= stale_size; } return false; @@ -2334,6 +2104,40 @@ static bool ice_chk_linearize(struct sk_buff *skb, unsigned int count) } /** + * ice_tstamp - set up context descriptor for hardware timestamp + * @tx_ring: pointer to the Tx ring to send buffer on + * @skb: pointer to the SKB we're sending + * @first: Tx buffer + * @off: Tx offload parameters + */ +static void +ice_tstamp(struct ice_tx_ring *tx_ring, struct sk_buff *skb, + struct ice_tx_buf *first, struct ice_tx_offload_params *off) +{ + s8 idx; + + /* only timestamp the outbound packet if the user has requested it */ + if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP))) + return; + + /* Tx timestamps cannot be sampled when doing TSO */ + if (first->tx_flags & ICE_TX_FLAGS_TSO) + return; + + /* Grab an open timestamp slot */ + idx = ice_ptp_request_ts(tx_ring->tx_tstamps, skb); + if (idx < 0) { + tx_ring->vsi->back->ptp.tx_hwtstamp_skipped++; + return; + } + + off->cd_qw1 |= (u64)(ICE_TX_DESC_DTYPE_CTX | + (ICE_TX_CTX_DESC_TSYN << ICE_TXD_CTX_QW1_CMD_S) | + ((u64)idx << ICE_TXD_CTX_QW1_TSO_LEN_S)); + first->tx_flags |= ICE_TX_FLAGS_TSYN; +} + +/** * ice_xmit_frame_ring - Sends buffer on Tx ring * @skb: send buffer * @tx_ring: ring to send buffer on @@ -2341,20 +2145,26 @@ static bool ice_chk_linearize(struct sk_buff *skb, unsigned int count) * Returns NETDEV_TX_OK if sent, else an error code */ static netdev_tx_t -ice_xmit_frame_ring(struct sk_buff *skb, struct ice_ring *tx_ring) +ice_xmit_frame_ring(struct sk_buff *skb, struct ice_tx_ring *tx_ring) { struct ice_tx_offload_params offload = { 0 }; struct ice_vsi *vsi = tx_ring->vsi; struct ice_tx_buf *first; + struct ethhdr *eth; unsigned int count; int tso, csum; + ice_trace(xmit_frame_ring, tx_ring, skb); + + if (unlikely(ipv6_hopopt_jumbo_remove(skb))) + goto out_drop; + count = ice_xmit_desc_count(skb); if (ice_chk_linearize(skb, count)) { if (__skb_linearize(skb)) goto out_drop; count = ice_txd_use_count(skb->len); - tx_ring->tx_stats.tx_linearize++; + tx_ring->ring_stats->tx_stats.tx_linearize++; } /* need: 1 descriptor per page * PAGE_SIZE/ICE_MAX_DATA_PER_TXD, @@ -2365,21 +2175,31 @@ ice_xmit_frame_ring(struct sk_buff *skb, struct ice_ring *tx_ring) */ if (ice_maybe_stop_tx(tx_ring, count + ICE_DESCS_PER_CACHE_LINE + ICE_DESCS_FOR_CTX_DESC)) { - tx_ring->tx_stats.tx_busy++; + tx_ring->ring_stats->tx_stats.tx_busy++; return NETDEV_TX_BUSY; } + /* prefetch for bql data which is infrequently used */ + netdev_txq_bql_enqueue_prefetchw(txring_txq(tx_ring)); + offload.tx_ring = tx_ring; /* record the location of the first descriptor for this packet */ first = &tx_ring->tx_buf[tx_ring->next_to_use]; first->skb = skb; + first->type = ICE_TX_BUF_SKB; first->bytecount = max_t(unsigned int, skb->len, ETH_ZLEN); first->gso_segs = 1; first->tx_flags = 0; /* prepare the VLAN tagging flags for Tx */ ice_tx_prepare_vlan_flags(tx_ring, first); + if (first->tx_flags & ICE_TX_FLAGS_HW_OUTER_SINGLE_VLAN) { + offload.cd_qw1 |= (u64)(ICE_TX_DESC_DTYPE_CTX | + (ICE_TX_CTX_DESC_IL2TAG2 << + ICE_TXD_CTX_QW1_CMD_S)); + offload.cd_l2tag2 = first->vid; + } /* set up TSO offload */ tso = ice_tso(first, &offload); @@ -2392,13 +2212,22 @@ ice_xmit_frame_ring(struct sk_buff *skb, struct ice_ring *tx_ring) goto out_drop; /* allow CONTROL frames egress from main VSI if FW LLDP disabled */ - if (unlikely(skb->priority == TC_PRIO_CONTROL && - vsi->type == ICE_VSI_PF && - vsi->port_info->is_sw_lldp)) + eth = (struct ethhdr *)skb_mac_header(skb); + + if ((ice_is_switchdev_running(vsi->back) || + ice_lag_is_switchdev_running(vsi->back)) && + vsi->type != ICE_VSI_SF) + ice_eswitch_set_target_vsi(skb, &offload); + else if (unlikely((skb->priority == TC_PRIO_CONTROL || + eth->h_proto == htons(ETH_P_LLDP)) && + vsi->type == ICE_VSI_PF && + vsi->port_info->qos_cfg.is_sw_lldp)) offload.cd_qw1 |= (u64)(ICE_TX_DESC_DTYPE_CTX | ICE_TX_CTX_DESC_SWTCH_UPLINK << ICE_TXD_CTX_QW1_CMD_S); + ice_tstamp(tx_ring, skb, first, &offload); + if (offload.cd_qw1 & ICE_TX_DESC_DTYPE_CTX) { struct ice_tx_ctx_desc *cdesc; u16 i = tx_ring->next_to_use; @@ -2411,7 +2240,7 @@ ice_xmit_frame_ring(struct sk_buff *skb, struct ice_ring *tx_ring) /* setup context descriptor */ cdesc->tunneling_params = cpu_to_le32(offload.cd_tunnel_params); cdesc->l2tag2 = cpu_to_le16(offload.cd_l2tag2); - cdesc->rsvd = cpu_to_le16(0); + cdesc->gcs = cpu_to_le16(offload.cd_gcs_params); cdesc->qw1 = cpu_to_le64(offload.cd_qw1); } @@ -2419,6 +2248,7 @@ ice_xmit_frame_ring(struct sk_buff *skb, struct ice_ring *tx_ring) return NETDEV_TX_OK; out_drop: + ice_trace(xmit_frame_ring_drop, tx_ring, skb); dev_kfree_skb_any(skb); return NETDEV_TX_OK; } @@ -2434,7 +2264,7 @@ netdev_tx_t ice_start_xmit(struct sk_buff *skb, struct net_device *netdev) { struct ice_netdev_priv *np = netdev_priv(netdev); struct ice_vsi *vsi = np->vsi; - struct ice_ring *tx_ring; + struct ice_tx_ring *tx_ring; tx_ring = vsi->tx_rings[skb->queue_mapping]; @@ -2448,10 +2278,43 @@ netdev_tx_t ice_start_xmit(struct sk_buff *skb, struct net_device *netdev) } /** + * ice_get_dscp_up - return the UP/TC value for a SKB + * @dcbcfg: DCB config that contains DSCP to UP/TC mapping + * @skb: SKB to query for info to determine UP/TC + * + * This function is to only be called when the PF is in L3 DSCP PFC mode + */ +static u8 ice_get_dscp_up(struct ice_dcbx_cfg *dcbcfg, struct sk_buff *skb) +{ + u8 dscp = 0; + + if (skb->protocol == htons(ETH_P_IP)) + dscp = ipv4_get_dsfield(ip_hdr(skb)) >> 2; + else if (skb->protocol == htons(ETH_P_IPV6)) + dscp = ipv6_get_dsfield(ipv6_hdr(skb)) >> 2; + + return dcbcfg->dscp_map[dscp]; +} + +u16 +ice_select_queue(struct net_device *netdev, struct sk_buff *skb, + struct net_device *sb_dev) +{ + struct ice_pf *pf = ice_netdev_to_pf(netdev); + struct ice_dcbx_cfg *dcbcfg; + + dcbcfg = &pf->hw.port_info->qos_cfg.local_dcbx_cfg; + if (dcbcfg->pfc_mode == ICE_QOS_MODE_DSCP) + skb->priority = ice_get_dscp_up(dcbcfg, skb); + + return netdev_pick_tx(netdev, skb, sb_dev); +} + +/** * ice_clean_ctrl_tx_irq - interrupt handler for flow director Tx queue * @tx_ring: tx_ring to clean */ -void ice_clean_ctrl_tx_irq(struct ice_ring *tx_ring) +void ice_clean_ctrl_tx_irq(struct ice_tx_ring *tx_ring) { struct ice_vsi *vsi = tx_ring->vsi; s16 i = tx_ring->next_to_clean; @@ -2499,11 +2362,11 @@ void ice_clean_ctrl_tx_irq(struct ice_ring *tx_ring) dma_unmap_addr(tx_buf, dma), dma_unmap_len(tx_buf, len), DMA_TO_DEVICE); - if (tx_buf->tx_flags & ICE_TX_FLAGS_DUMMY_PKT) + if (tx_buf->type == ICE_TX_BUF_DUMMY) devm_kfree(tx_ring->dev, tx_buf->raw_buf); /* clear next_to_watch to prevent false hangs */ - tx_buf->raw_buf = NULL; + tx_buf->type = ICE_TX_BUF_EMPTY; tx_buf->tx_flags = 0; tx_buf->next_to_watch = NULL; dma_unmap_len_set(tx_buf, len, 0); |