diff options
Diffstat (limited to 'drivers/net/ethernet/intel/ice/ice_txrx.c')
| -rw-r--r-- | drivers/net/ethernet/intel/ice/ice_txrx.c | 326 |
1 files changed, 207 insertions, 119 deletions
diff --git a/drivers/net/ethernet/intel/ice/ice_txrx.c b/drivers/net/ethernet/intel/ice/ice_txrx.c index 6ee8e0032d52..bc3ba19dc88f 100644 --- a/drivers/net/ethernet/intel/ice/ice_txrx.c +++ b/drivers/net/ethernet/intel/ice/ice_txrx.c @@ -6,6 +6,7 @@ #include <linux/prefetch.h> #include <linux/mm.h> #include <linux/bpf_trace.h> +#include <net/dsfield.h> #include <net/xdp.h> #include "ice_txrx_lib.h" #include "ice_lib.h" @@ -13,6 +14,7 @@ #include "ice_trace.h" #include "ice_dcb_lib.h" #include "ice_xsk.h" +#include "ice_eswitch.h" #define ICE_RX_HDR_SIZE 256 @@ -32,7 +34,7 @@ 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; @@ -106,7 +108,7 @@ 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) @@ -133,7 +135,7 @@ ice_unmap_and_free_tx_buf(struct ice_ring *ring, struct ice_tx_buf *tx_buf) /* 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); } @@ -142,8 +144,9 @@ static struct netdev_queue *txring_txq(const struct ice_ring *ring) * 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_pool) { @@ -162,8 +165,10 @@ 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; @@ -181,14 +186,18 @@ 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; } @@ -201,7 +210,7 @@ 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; @@ -238,11 +247,8 @@ 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, @@ -298,9 +304,6 @@ static bool ice_clean_tx_irq(struct ice_ring *tx_ring, int napi_budget) 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); @@ -329,9 +332,10 @@ static bool ice_clean_tx_irq(struct ice_ring *tx_ring, int napi_budget) * * 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; @@ -339,19 +343,19 @@ 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; } @@ -370,9 +374,10 @@ err: * 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; + u32 size; u16 i; /* ring already cleared, nothing to do */ @@ -417,7 +422,9 @@ rx_skip_free: memset(rx_ring->rx_buf, 0, sizeof(*rx_ring->rx_buf) * rx_ring->count); /* 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; @@ -430,8 +437,10 @@ 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) { + 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)) @@ -441,7 +450,9 @@ void ice_free_rx_ring(struct ice_ring *rx_ring) rx_ring->rx_buf = NULL; if (rx_ring->desc) { - dmam_free_coherent(rx_ring->dev, rx_ring->size, + size = ALIGN(rx_ring->count * sizeof(union ice_32byte_rx_desc), + PAGE_SIZE); + dmam_free_coherent(rx_ring->dev, size, rx_ring->desc, rx_ring->dma); rx_ring->desc = NULL; } @@ -453,9 +464,10 @@ 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; + u32 size; if (!dev) return -ENOMEM; @@ -463,19 +475,19 @@ int ice_setup_rx_ring(struct ice_ring *rx_ring) /* 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, + devm_kcalloc(dev, sizeof(*rx_ring->rx_buf), rx_ring->count, GFP_KERNEL); if (!rx_ring->rx_buf) return -ENOMEM; /* 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); + size); goto err; } @@ -499,7 +511,7 @@ err: } static unsigned int -ice_rx_frame_truesize(struct ice_ring *rx_ring, unsigned int __maybe_unused size) +ice_rx_frame_truesize(struct ice_rx_ring *rx_ring, unsigned int __maybe_unused size) { unsigned int truesize; @@ -519,15 +531,15 @@ ice_rx_frame_truesize(struct ice_ring *rx_ring, unsigned int __maybe_unused size * @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 * * 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) +ice_run_xdp(struct ice_rx_ring *rx_ring, struct xdp_buff *xdp, + struct bpf_prog *xdp_prog, struct ice_tx_ring *xdp_ring) { - struct ice_ring *xdp_ring; - int err, result; + int err; u32 act; act = bpf_prog_run_xdp(xdp_prog, xdp); @@ -535,11 +547,14 @@ ice_run_xdp(struct ice_ring *rx_ring, struct xdp_buff *xdp, case XDP_PASS: return ICE_XDP_PASS; case XDP_TX: - xdp_ring = rx_ring->vsi->xdp_rings[smp_processor_id()]; - result = ice_xmit_xdp_buff(xdp, xdp_ring); - if (result == ICE_XDP_CONSUMED) + if (static_branch_unlikely(&ice_xdp_locking_key)) + spin_lock(&xdp_ring->tx_lock); + err = ice_xmit_xdp_ring(xdp->data, xdp->data_end - xdp->data, xdp_ring); + if (static_branch_unlikely(&ice_xdp_locking_key)) + spin_unlock(&xdp_ring->tx_lock); + if (err == ICE_XDP_CONSUMED) goto out_failure; - return result; + return err; case XDP_REDIRECT: err = xdp_do_redirect(rx_ring->netdev, xdp, xdp_prog); if (err) @@ -576,7 +591,7 @@ 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; + struct ice_tx_ring *xdp_ring; int nxmit = 0, i; if (test_bit(ICE_VSI_DOWN, vsi->state)) @@ -588,7 +603,14 @@ ice_xdp_xmit(struct net_device *dev, int n, struct xdp_frame **frames, 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 { + xdp_ring = vsi->xdp_rings[queue_index]; + } + for (i = 0; i < n; i++) { struct xdp_frame *xdpf = frames[i]; int err; @@ -602,6 +624,9 @@ ice_xdp_xmit(struct net_device *dev, int n, struct xdp_frame **frames, if (unlikely(flags & XDP_XMIT_FLUSH)) ice_xdp_ring_update_tail(xdp_ring); + if (static_branch_unlikely(&ice_xdp_locking_key)) + spin_unlock(&xdp_ring->tx_lock); + return nxmit; } @@ -614,7 +639,7 @@ ice_xdp_xmit(struct net_device *dev, int n, struct xdp_frame **frames, * reused. */ static bool -ice_alloc_mapped_page(struct ice_ring *rx_ring, struct ice_rx_buf *bi) +ice_alloc_mapped_page(struct ice_rx_ring *rx_ring, struct ice_rx_buf *bi) { struct page *page = bi->page; dma_addr_t dma; @@ -665,7 +690,7 @@ 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, u16 cleaned_count) { union ice_32b_rx_flex_desc *rx_desc; u16 ntu = rx_ring->next_to_use; @@ -794,7 +819,7 @@ ice_can_reuse_rx_page(struct ice_rx_buf *rx_buf, int rx_buf_pgcnt) * 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, +ice_add_rx_frag(struct ice_rx_ring *rx_ring, struct ice_rx_buf *rx_buf, struct sk_buff *skb, unsigned int size) { #if (PAGE_SIZE >= 8192) @@ -820,7 +845,7 @@ ice_add_rx_frag(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf, * Synchronizes page for reuse by the adapter */ static void -ice_reuse_rx_page(struct ice_ring *rx_ring, struct ice_rx_buf *old_buf) +ice_reuse_rx_page(struct ice_rx_ring *rx_ring, struct ice_rx_buf *old_buf) { u16 nta = rx_ring->next_to_alloc; struct ice_rx_buf *new_buf; @@ -851,7 +876,7 @@ ice_reuse_rx_page(struct ice_ring *rx_ring, struct ice_rx_buf *old_buf) * for use by the CPU. */ static struct ice_rx_buf * -ice_get_rx_buf(struct ice_ring *rx_ring, const unsigned int size, +ice_get_rx_buf(struct ice_rx_ring *rx_ring, const unsigned int size, int *rx_buf_pgcnt) { struct ice_rx_buf *rx_buf; @@ -888,7 +913,7 @@ ice_get_rx_buf(struct ice_ring *rx_ring, const unsigned int size, * 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, +ice_build_skb(struct ice_rx_ring *rx_ring, struct ice_rx_buf *rx_buf, struct xdp_buff *xdp) { u8 metasize = xdp->data - xdp->data_meta; @@ -940,7 +965,7 @@ ice_build_skb(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf, * skb correctly. */ static struct sk_buff * -ice_construct_skb(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf, +ice_construct_skb(struct ice_rx_ring *rx_ring, struct ice_rx_buf *rx_buf, struct xdp_buff *xdp) { unsigned int size = xdp->data_end - xdp->data; @@ -1000,7 +1025,7 @@ ice_construct_skb(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf, * the associated resources. */ static void -ice_put_rx_buf(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf, +ice_put_rx_buf(struct ice_rx_ring *rx_ring, struct ice_rx_buf *rx_buf, int rx_buf_pgcnt) { u16 ntc = rx_ring->next_to_clean + 1; @@ -1036,7 +1061,7 @@ ice_put_rx_buf(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf, * 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) +ice_is_non_eop(struct ice_rx_ring *rx_ring, union ice_32b_rx_flex_desc *rx_desc) { /* 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) @@ -1060,11 +1085,12 @@ 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) +int ice_clean_rx_irq(struct ice_rx_ring *rx_ring, int budget) { unsigned int total_rx_bytes = 0, total_rx_pkts = 0, frame_sz = 0; u16 cleaned_count = ICE_DESC_UNUSED(rx_ring); unsigned int offset = rx_ring->rx_offset; + struct ice_tx_ring *xdp_ring = NULL; unsigned int xdp_res, xdp_xmit = 0; struct sk_buff *skb = rx_ring->skb; struct bpf_prog *xdp_prog = NULL; @@ -1077,6 +1103,10 @@ int ice_clean_rx_irq(struct ice_ring *rx_ring, int budget) #endif xdp_init_buff(&xdp, frame_sz, &rx_ring->xdp_rxq); + xdp_prog = READ_ONCE(rx_ring->xdp_prog); + if (xdp_prog) + xdp_ring = rx_ring->xdp_ring; + /* 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; @@ -1140,11 +1170,10 @@ int ice_clean_rx_irq(struct ice_ring *rx_ring, int budget) xdp.frame_sz = ice_rx_frame_truesize(rx_ring, size); #endif - xdp_prog = READ_ONCE(rx_ring->xdp_prog); if (!xdp_prog) goto construct_skb; - xdp_res = ice_run_xdp(rx_ring, &xdp, xdp_prog); + xdp_res = ice_run_xdp(rx_ring, &xdp, xdp_prog, xdp_ring); if (!xdp_res) goto construct_skb; if (xdp_res & (ICE_XDP_TX | ICE_XDP_REDIR)) { @@ -1221,7 +1250,7 @@ construct_skb: failure = ice_alloc_rx_bufs(rx_ring, cleaned_count); if (xdp_prog) - ice_finalize_xdp_rx(rx_ring, xdp_xmit); + ice_finalize_xdp_rx(xdp_ring, xdp_xmit); rx_ring->skb = skb; ice_update_rx_ring_stats(rx_ring, total_rx_pkts, total_rx_bytes); @@ -1230,6 +1259,41 @@ construct_skb: return failure ? budget : (int)total_rx_pkts; } +static void __ice_update_sample(struct ice_q_vector *q_vector, + struct ice_ring_container *rc, + struct dim_sample *sample, + bool is_tx) +{ + u64 packets = 0, bytes = 0; + + if (is_tx) { + struct ice_tx_ring *tx_ring; + + ice_for_each_tx_ring(tx_ring, *rc) { + packets += tx_ring->stats.pkts; + bytes += tx_ring->stats.bytes; + } + } else { + struct ice_rx_ring *rx_ring; + + ice_for_each_rx_ring(rx_ring, *rc) { + packets += rx_ring->stats.pkts; + bytes += rx_ring->stats.bytes; + } + } + + dim_update_sample(q_vector->total_events, packets, bytes, sample); + sample->comp_ctr = 0; + + /* 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. + */ + if (ktime_ms_delta(sample->time, rc->dim.start_sample.time) >= 1000) + rc->dim.state = DIM_START_MEASURE; +} + /** * ice_net_dim - Update net DIM algorithm * @q_vector: the vector associated with the interrupt @@ -1245,34 +1309,16 @@ static void ice_net_dim(struct ice_q_vector *q_vector) struct ice_ring_container *rx = &q_vector->rx; if (ITR_IS_DYNAMIC(tx)) { - struct dim_sample dim_sample = {}; - u64 packets = 0, bytes = 0; - struct ice_ring *ring; - - ice_for_each_ring(ring, q_vector->tx) { - packets += ring->stats.pkts; - bytes += ring->stats.bytes; - } - - dim_update_sample(q_vector->total_events, packets, bytes, - &dim_sample); + struct dim_sample dim_sample; + __ice_update_sample(q_vector, tx, &dim_sample, true); net_dim(&tx->dim, dim_sample); } if (ITR_IS_DYNAMIC(rx)) { - struct dim_sample dim_sample = {}; - u64 packets = 0, bytes = 0; - struct ice_ring *ring; - - ice_for_each_ring(ring, q_vector->rx) { - packets += ring->stats.pkts; - bytes += ring->stats.bytes; - } - - dim_update_sample(q_vector->total_events, packets, bytes, - &dim_sample); + struct dim_sample dim_sample; + __ice_update_sample(q_vector, rx, &dim_sample, false); net_dim(&rx->dim, dim_sample); } } @@ -1299,15 +1345,14 @@ static u32 ice_buildreg_itr(u16 itr_idx, u16 itr) } /** - * ice_update_ena_itr - Update ITR moderation and re-enable MSI-X interrupt + * ice_enable_interrupt - re-enable MSI-X interrupt * @q_vector: the vector associated with the interrupt to enable * - * Update the net_dim() algorithm and re-enable the interrupt associated with - * this vector. - * - * If the VSI is down, the interrupt will not be re-enabled. + * 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_vsi *vsi = q_vector->vsi; bool wb_en = q_vector->wb_on_itr; @@ -1316,25 +1361,25 @@ static void ice_update_ena_itr(struct ice_q_vector *q_vector) if (test_bit(ICE_DOWN, vsi->state)) return; - /* When exiting WB_ON_ITR, let ITR resume its normal - * interrupts-enabled path. + /* 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 (wb_en) + if (!wb_en) { + itr_val = ice_buildreg_itr(ICE_ITR_NONE, 0); + } else { q_vector->wb_on_itr = false; - /* This will do nothing if dynamic updates are not enabled. */ - ice_net_dim(q_vector); - - /* net_dim() updates ITR out-of-band using a work item */ - itr_val = ice_buildreg_itr(ICE_ITR_NONE, 0); - /* trigger an immediate 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 (wb_en) { + /* 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 | - GLINT_DYN_CTL_SW_ITR_INDX_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); @@ -1387,18 +1432,24 @@ 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_pool ? - ice_clean_tx_irq_zc(ring, budget) : - ice_clean_tx_irq(ring, budget); + ice_for_each_tx_ring(tx_ring, q_vector->tx) { + bool wd; + + if (tx_ring->xsk_pool) + wd = ice_clean_tx_irq_zc(tx_ring, budget); + else if (ice_ring_is_xdp(tx_ring)) + wd = true; + else + wd = ice_clean_tx_irq(tx_ring, budget); if (!wd) clean_complete = false; @@ -1419,16 +1470,16 @@ 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) { 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_pool ? - 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, 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) @@ -1447,10 +1498,12 @@ int ice_napi_poll(struct napi_struct *napi, int 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 (likely(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); } @@ -1462,7 +1515,7 @@ 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); /* Memory barrier before checking head and tail */ @@ -1485,7 +1538,7 @@ 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; @@ -1504,7 +1557,7 @@ 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; @@ -1840,7 +1893,7 @@ 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; @@ -2146,7 +2199,7 @@ static bool ice_chk_linearize(struct sk_buff *skb, unsigned int count) * @off: Tx offload parameters */ static void -ice_tstamp(struct ice_ring *tx_ring, struct sk_buff *skb, +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; @@ -2181,7 +2234,7 @@ ice_tstamp(struct ice_ring *tx_ring, struct sk_buff *skb, * 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; @@ -2245,6 +2298,8 @@ ice_xmit_frame_ring(struct sk_buff *skb, struct ice_ring *tx_ring) ICE_TXD_CTX_QW1_CMD_S); ice_tstamp(tx_ring, skb, first, &offload); + if (ice_is_switchdev_running(vsi->back)) + ice_eswitch_set_target_vsi(skb, &offload); if (offload.cd_qw1 & ICE_TX_DESC_DTYPE_CTX) { struct ice_tx_ctx_desc *cdesc; @@ -2282,7 +2337,7 @@ 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]; @@ -2296,10 +2351,43 @@ 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; |