ice: split ice_ring onto Tx/Rx separate structs

While it was convenient to have a generic ring structure that served
both Tx and Rx sides, next commits are going to introduce several
Tx-specific fields, so in order to avoid hurting the Rx side, let's
pull out the Tx ring onto new ice_tx_ring and ice_rx_ring structs.

Rx ring could be handled by the old ice_ring which would reduce the code
churn within this patch, but this would make things asymmetric.

Make the union out of the ring container within ice_q_vector so that it
is possible to iterate over newly introduced ice_tx_ring.

Remove the @size as it's only accessed from control path and it can be
calculated pretty easily.

Change definitions of ice_update_ring_stats and
ice_fetch_u64_stats_per_ring so that they are ring agnostic and can be
used for both Rx and Tx rings.

Sizes of Rx and Tx ring structs are 256 and 192 bytes, respectively. In
Rx ring xdp_rxq_info occupies its own cacheline, so it's the major
difference now.

Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

1 files changed, 74 insertions, 57 deletions

diff --git a/drivers/net/ethernet/intel/ice/ice_txrx.c b/drivers/net/ethernet/intel/ice/ice_txrx.c
index 4da9420df1b0..b031f754c5cb 100644
--- a/drivers/net/ethernet/intel/ice/ice_txrx.c
+++ b/drivers/net/ethernet/intel/ice/ice_txrx.c
@@ -34,7 +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;
@@ -108,7 +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)
@@ -135,7 +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);
 }
@@ -144,8 +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) {
@@ -164,8 +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;
@@ -183,14 +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;
 	}
@@ -203,7 +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;
@@ -331,9 +338,10 @@ 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;
@@ -347,13 +355,13 @@ int ice_setup_tx_ring(struct ice_ring *tx_ring)
 		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;
 	}
 
@@ -372,9 +380,10 @@ 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 */
@@ -419,7 +428,9 @@ 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;
@@ -432,8 +443,10 @@ 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))
@@ -443,7 +456,9 @@ 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;
 	}
@@ -455,9 +470,10 @@ 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;
@@ -471,13 +487,13 @@ int ice_setup_rx_ring(struct ice_ring *rx_ring)
 		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;
 	}
 
@@ -501,7 +517,7 @@ 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;
 
@@ -525,10 +541,10 @@ ice_rx_frame_truesize(struct ice_ring *rx_ring, unsigned int __maybe_unused size
  * Returns any of ICE_XDP_{PASS, CONSUMED, TX, REDIR}
  */
 static int
-ice_run_xdp(struct ice_ring *rx_ring, struct xdp_buff *xdp,
+ice_run_xdp(struct ice_rx_ring *rx_ring, struct xdp_buff *xdp,
 	    struct bpf_prog *xdp_prog)
 {
-	struct ice_ring *xdp_ring;
+	struct ice_tx_ring *xdp_ring;
 	int err, result;
 	u32 act;
 
@@ -578,7 +594,7 @@ 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))
@@ -616,7 +632,7 @@ 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;
@@ -667,7 +683,7 @@ 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;
@@ -796,7 +812,7 @@ 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)
@@ -822,7 +838,7 @@ 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;
@@ -853,7 +869,7 @@ 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;
@@ -890,7 +906,7 @@ 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;
@@ -942,7 +958,7 @@ 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;
@@ -1002,7 +1018,7 @@ 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;
@@ -1038,7 +1054,7 @@ 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)
@@ -1062,7 +1078,7 @@ 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);
@@ -1249,9 +1265,9 @@ static void ice_net_dim(struct ice_q_vector *q_vector)
 	if (ITR_IS_DYNAMIC(tx)) {
 		struct dim_sample dim_sample = {};
 		u64 packets = 0, bytes = 0;
-		struct ice_ring *ring;
+		struct ice_tx_ring *ring;
 
-		ice_for_each_ring(ring, q_vector->tx) {
+		ice_for_each_tx_ring(ring, q_vector->tx) {
 			packets += ring->stats.pkts;
 			bytes += ring->stats.bytes;
 		}
@@ -1265,9 +1281,9 @@ static void ice_net_dim(struct ice_q_vector *q_vector)
 	if (ITR_IS_DYNAMIC(rx)) {
 		struct dim_sample dim_sample = {};
 		u64 packets = 0, bytes = 0;
-		struct ice_ring *ring;
+		struct ice_rx_ring *ring;
 
-		ice_for_each_ring(ring, q_vector->rx) {
+		ice_for_each_rx_ring(ring, q_vector->rx) {
 			packets += ring->stats.pkts;
 			bytes += ring->stats.bytes;
 		}
@@ -1389,18 +1405,19 @@ 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 = tx_ring->xsk_pool ?
+			  ice_clean_tx_irq_zc(tx_ring, budget) :
+			  ice_clean_tx_irq(tx_ring, budget);
 
 		if (!wd)
 			clean_complete = false;
@@ -1421,16 +1438,16 @@ 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)
@@ -1464,7 +1481,7 @@ 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 */
@@ -1487,7 +1504,7 @@ 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;
@@ -1506,7 +1523,7 @@ 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;
@@ -1842,7 +1859,7 @@ 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;
 
@@ -2148,7 +2165,7 @@ 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;
@@ -2183,7 +2200,7 @@ 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;
@@ -2286,7 +2303,7 @@ 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];
 
@@ -2336,7 +2353,7 @@ ice_select_queue(struct net_device *netdev, struct sk_buff *skb,
  * 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;