1 files changed, 177 insertions, 590 deletions

diff --git a/drivers/net/ethernet/sfc/rx.c b/drivers/net/ethernet/sfc/rx.c
index d5db045535d3..a2042f16babc 100644
--- a/drivers/net/ethernet/sfc/rx.c
+++ b/drivers/net/ethernet/sfc/rx.c
@@ -17,8 +17,11 @@
 #include <linux/iommu.h>
 #include <net/ip.h>
 #include <net/checksum.h>
+#include <net/xdp.h>
+#include <linux/bpf_trace.h>
 #include "net_driver.h"
 #include "efx.h"
+#include "rx_common.h"
 #include "filter.h"
 #include "nic.h"
 #include "selftest.h"
@@ -27,60 +30,16 @@
 /* Preferred number of descriptors to fill at once */
 #define EFX_RX_PREFERRED_BATCH 8U
 
-/* Number of RX buffers to recycle pages for.  When creating the RX page recycle
- * ring, this number is divided by the number of buffers per page to calculate
- * the number of pages to store in the RX page recycle ring.
- */
-#define EFX_RECYCLE_RING_SIZE_IOMMU 4096
-#define EFX_RECYCLE_RING_SIZE_NOIOMMU (2 * EFX_RX_PREFERRED_BATCH)
+/* Maximum rx prefix used by any architecture. */
+#define EFX_MAX_RX_PREFIX_SIZE 16
 
 /* Size of buffer allocated for skb header area. */
 #define EFX_SKB_HEADERS  128u
 
-/* This is the percentage fill level below which new RX descriptors
- * will be added to the RX descriptor ring.
- */
-static unsigned int rx_refill_threshold;
-
 /* Each packet can consume up to ceil(max_frame_len / buffer_size) buffers */
 #define EFX_RX_MAX_FRAGS DIV_ROUND_UP(EFX_MAX_FRAME_LEN(EFX_MAX_MTU), \
 				      EFX_RX_USR_BUF_SIZE)
 
-/*
- * RX maximum head room required.
- *
- * This must be at least 1 to prevent overflow, plus one packet-worth
- * to allow pipelined receives.
- */
-#define EFX_RXD_HEAD_ROOM (1 + EFX_RX_MAX_FRAGS)
-
-static inline u8 *efx_rx_buf_va(struct efx_rx_buffer *buf)
-{
-	return page_address(buf->page) + buf->page_offset;
-}
-
-static inline u32 efx_rx_buf_hash(struct efx_nic *efx, const u8 *eh)
-{
-#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS)
-	return __le32_to_cpup((const __le32 *)(eh + efx->rx_packet_hash_offset));
-#else
-	const u8 *data = eh + efx->rx_packet_hash_offset;
-	return (u32)data[0]	  |
-	       (u32)data[1] << 8  |
-	       (u32)data[2] << 16 |
-	       (u32)data[3] << 24;
-#endif
-}
-
-static inline struct efx_rx_buffer *
-efx_rx_buf_next(struct efx_rx_queue *rx_queue, struct efx_rx_buffer *rx_buf)
-{
-	if (unlikely(rx_buf == efx_rx_buffer(rx_queue, rx_queue->ptr_mask)))
-		return efx_rx_buffer(rx_queue, 0);
-	else
-		return rx_buf + 1;
-}
-
 static inline void efx_sync_rx_buffer(struct efx_nic *efx,
 				      struct efx_rx_buffer *rx_buf,
 				      unsigned int len)
@@ -89,298 +48,6 @@ static inline void efx_sync_rx_buffer(struct efx_nic *efx,
 				DMA_FROM_DEVICE);
 }
 
-void efx_rx_config_page_split(struct efx_nic *efx)
-{
-	efx->rx_page_buf_step = ALIGN(efx->rx_dma_len + efx->rx_ip_align,
-				      EFX_RX_BUF_ALIGNMENT);
-	efx->rx_bufs_per_page = efx->rx_buffer_order ? 1 :
-		((PAGE_SIZE - sizeof(struct efx_rx_page_state)) /
-		 efx->rx_page_buf_step);
-	efx->rx_buffer_truesize = (PAGE_SIZE << efx->rx_buffer_order) /
-		efx->rx_bufs_per_page;
-	efx->rx_pages_per_batch = DIV_ROUND_UP(EFX_RX_PREFERRED_BATCH,
-					       efx->rx_bufs_per_page);
-}
-
-/* Check the RX page recycle ring for a page that can be reused. */
-static struct page *efx_reuse_page(struct efx_rx_queue *rx_queue)
-{
-	struct efx_nic *efx = rx_queue->efx;
-	struct page *page;
-	struct efx_rx_page_state *state;
-	unsigned index;
-
-	index = rx_queue->page_remove & rx_queue->page_ptr_mask;
-	page = rx_queue->page_ring[index];
-	if (page == NULL)
-		return NULL;
-
-	rx_queue->page_ring[index] = NULL;
-	/* page_remove cannot exceed page_add. */
-	if (rx_queue->page_remove != rx_queue->page_add)
-		++rx_queue->page_remove;
-
-	/* If page_count is 1 then we hold the only reference to this page. */
-	if (page_count(page) == 1) {
-		++rx_queue->page_recycle_count;
-		return page;
-	} else {
-		state = page_address(page);
-		dma_unmap_page(&efx->pci_dev->dev, state->dma_addr,
-			       PAGE_SIZE << efx->rx_buffer_order,
-			       DMA_FROM_DEVICE);
-		put_page(page);
-		++rx_queue->page_recycle_failed;
-	}
-
-	return NULL;
-}
-
-/**
- * efx_init_rx_buffers - create EFX_RX_BATCH page-based RX buffers
- *
- * @rx_queue:		Efx RX queue
- *
- * This allocates a batch of pages, maps them for DMA, and populates
- * struct efx_rx_buffers for each one. Return a negative error code or
- * 0 on success. If a single page can be used for multiple buffers,
- * then the page will either be inserted fully, or not at all.
- */
-static int efx_init_rx_buffers(struct efx_rx_queue *rx_queue, bool atomic)
-{
-	struct efx_nic *efx = rx_queue->efx;
-	struct efx_rx_buffer *rx_buf;
-	struct page *page;
-	unsigned int page_offset;
-	struct efx_rx_page_state *state;
-	dma_addr_t dma_addr;
-	unsigned index, count;
-
-	count = 0;
-	do {
-		page = efx_reuse_page(rx_queue);
-		if (page == NULL) {
-			page = alloc_pages(__GFP_COMP |
-					   (atomic ? GFP_ATOMIC : GFP_KERNEL),
-					   efx->rx_buffer_order);
-			if (unlikely(page == NULL))
-				return -ENOMEM;
-			dma_addr =
-				dma_map_page(&efx->pci_dev->dev, page, 0,
-					     PAGE_SIZE << efx->rx_buffer_order,
-					     DMA_FROM_DEVICE);
-			if (unlikely(dma_mapping_error(&efx->pci_dev->dev,
-						       dma_addr))) {
-				__free_pages(page, efx->rx_buffer_order);
-				return -EIO;
-			}
-			state = page_address(page);
-			state->dma_addr = dma_addr;
-		} else {
-			state = page_address(page);
-			dma_addr = state->dma_addr;
-		}
-
-		dma_addr += sizeof(struct efx_rx_page_state);
-		page_offset = sizeof(struct efx_rx_page_state);
-
-		do {
-			index = rx_queue->added_count & rx_queue->ptr_mask;
-			rx_buf = efx_rx_buffer(rx_queue, index);
-			rx_buf->dma_addr = dma_addr + efx->rx_ip_align;
-			rx_buf->page = page;
-			rx_buf->page_offset = page_offset + efx->rx_ip_align;
-			rx_buf->len = efx->rx_dma_len;
-			rx_buf->flags = 0;
-			++rx_queue->added_count;
-			get_page(page);
-			dma_addr += efx->rx_page_buf_step;
-			page_offset += efx->rx_page_buf_step;
-		} while (page_offset + efx->rx_page_buf_step <= PAGE_SIZE);
-
-		rx_buf->flags = EFX_RX_BUF_LAST_IN_PAGE;
-	} while (++count < efx->rx_pages_per_batch);
-
-	return 0;
-}
-
-/* Unmap a DMA-mapped page.  This function is only called for the final RX
- * buffer in a page.
- */
-static void efx_unmap_rx_buffer(struct efx_nic *efx,
-				struct efx_rx_buffer *rx_buf)
-{
-	struct page *page = rx_buf->page;
-
-	if (page) {
-		struct efx_rx_page_state *state = page_address(page);
-		dma_unmap_page(&efx->pci_dev->dev,
-			       state->dma_addr,
-			       PAGE_SIZE << efx->rx_buffer_order,
-			       DMA_FROM_DEVICE);
-	}
-}
-
-static void efx_free_rx_buffers(struct efx_rx_queue *rx_queue,
-				struct efx_rx_buffer *rx_buf,
-				unsigned int num_bufs)
-{
-	do {
-		if (rx_buf->page) {
-			put_page(rx_buf->page);
-			rx_buf->page = NULL;
-		}
-		rx_buf = efx_rx_buf_next(rx_queue, rx_buf);
-	} while (--num_bufs);
-}
-
-/* Attempt to recycle the page if there is an RX recycle ring; the page can
- * only be added if this is the final RX buffer, to prevent pages being used in
- * the descriptor ring and appearing in the recycle ring simultaneously.
- */
-static void efx_recycle_rx_page(struct efx_channel *channel,
-				struct efx_rx_buffer *rx_buf)
-{
-	struct page *page = rx_buf->page;
-	struct efx_rx_queue *rx_queue = efx_channel_get_rx_queue(channel);
-	struct efx_nic *efx = rx_queue->efx;
-	unsigned index;
-
-	/* Only recycle the page after processing the final buffer. */
-	if (!(rx_buf->flags & EFX_RX_BUF_LAST_IN_PAGE))
-		return;
-
-	index = rx_queue->page_add & rx_queue->page_ptr_mask;
-	if (rx_queue->page_ring[index] == NULL) {
-		unsigned read_index = rx_queue->page_remove &
-			rx_queue->page_ptr_mask;
-
-		/* The next slot in the recycle ring is available, but
-		 * increment page_remove if the read pointer currently
-		 * points here.
-		 */
-		if (read_index == index)
-			++rx_queue->page_remove;
-		rx_queue->page_ring[index] = page;
-		++rx_queue->page_add;
-		return;
-	}
-	++rx_queue->page_recycle_full;
-	efx_unmap_rx_buffer(efx, rx_buf);
-	put_page(rx_buf->page);
-}
-
-static void efx_fini_rx_buffer(struct efx_rx_queue *rx_queue,
-			       struct efx_rx_buffer *rx_buf)
-{
-	/* Release the page reference we hold for the buffer. */
-	if (rx_buf->page)
-		put_page(rx_buf->page);
-
-	/* If this is the last buffer in a page, unmap and free it. */
-	if (rx_buf->flags & EFX_RX_BUF_LAST_IN_PAGE) {
-		efx_unmap_rx_buffer(rx_queue->efx, rx_buf);
-		efx_free_rx_buffers(rx_queue, rx_buf, 1);
-	}
-	rx_buf->page = NULL;
-}
-
-/* Recycle the pages that are used by buffers that have just been received. */
-static void efx_recycle_rx_pages(struct efx_channel *channel,
-				 struct efx_rx_buffer *rx_buf,
-				 unsigned int n_frags)
-{
-	struct efx_rx_queue *rx_queue = efx_channel_get_rx_queue(channel);
-
-	do {
-		efx_recycle_rx_page(channel, rx_buf);
-		rx_buf = efx_rx_buf_next(rx_queue, rx_buf);
-	} while (--n_frags);
-}
-
-static void efx_discard_rx_packet(struct efx_channel *channel,
-				  struct efx_rx_buffer *rx_buf,
-				  unsigned int n_frags)
-{
-	struct efx_rx_queue *rx_queue = efx_channel_get_rx_queue(channel);
-
-	efx_recycle_rx_pages(channel, rx_buf, n_frags);
-
-	efx_free_rx_buffers(rx_queue, rx_buf, n_frags);
-}
-
-/**
- * efx_fast_push_rx_descriptors - push new RX descriptors quickly
- * @rx_queue:		RX descriptor queue
- *
- * This will aim to fill the RX descriptor queue up to
- * @rx_queue->@max_fill. If there is insufficient atomic
- * memory to do so, a slow fill will be scheduled.
- *
- * The caller must provide serialisation (none is used here). In practise,
- * this means this function must run from the NAPI handler, or be called
- * when NAPI is disabled.
- */
-void efx_fast_push_rx_descriptors(struct efx_rx_queue *rx_queue, bool atomic)
-{
-	struct efx_nic *efx = rx_queue->efx;
-	unsigned int fill_level, batch_size;
-	int space, rc = 0;
-
-	if (!rx_queue->refill_enabled)
-		return;
-
-	/* Calculate current fill level, and exit if we don't need to fill */
-	fill_level = (rx_queue->added_count - rx_queue->removed_count);
-	EFX_WARN_ON_ONCE_PARANOID(fill_level > rx_queue->efx->rxq_entries);
-	if (fill_level >= rx_queue->fast_fill_trigger)
-		goto out;
-
-	/* Record minimum fill level */
-	if (unlikely(fill_level < rx_queue->min_fill)) {
-		if (fill_level)
-			rx_queue->min_fill = fill_level;
-	}
-
-	batch_size = efx->rx_pages_per_batch * efx->rx_bufs_per_page;
-	space = rx_queue->max_fill - fill_level;
-	EFX_WARN_ON_ONCE_PARANOID(space < batch_size);
-
-	netif_vdbg(rx_queue->efx, rx_status, rx_queue->efx->net_dev,
-		   "RX queue %d fast-filling descriptor ring from"
-		   " level %d to level %d\n",
-		   efx_rx_queue_index(rx_queue), fill_level,
-		   rx_queue->max_fill);
-
-
-	do {
-		rc = efx_init_rx_buffers(rx_queue, atomic);
-		if (unlikely(rc)) {
-			/* Ensure that we don't leave the rx queue empty */
-			efx_schedule_slow_fill(rx_queue);
-			goto out;
-		}
-	} while ((space -= batch_size) >= batch_size);
-
-	netif_vdbg(rx_queue->efx, rx_status, rx_queue->efx->net_dev,
-		   "RX queue %d fast-filled descriptor ring "
-		   "to level %d\n", efx_rx_queue_index(rx_queue),
-		   rx_queue->added_count - rx_queue->removed_count);
-
- out:
-	if (rx_queue->notified_count != rx_queue->added_count)
-		efx_nic_notify_rx_desc(rx_queue);
-}
-
-void efx_rx_slow_fill(struct timer_list *t)
-{
-	struct efx_rx_queue *rx_queue = from_timer(rx_queue, t, slow_fill);
-
-	/* Post an event to cause NAPI to run and refill the queue */
-	efx_nic_generate_fill_event(rx_queue);
-	++rx_queue->slow_fill_count;
-}
-
 static void efx_rx_packet__check_len(struct efx_rx_queue *rx_queue,
 				     struct efx_rx_buffer *rx_buf,
 				     int len)
@@ -404,56 +71,6 @@ static void efx_rx_packet__check_len(struct efx_rx_queue *rx_queue,
 	efx_rx_queue_channel(rx_queue)->n_rx_overlength++;
 }
 
-/* Pass a received packet up through GRO.  GRO can handle pages
- * regardless of checksum state and skbs with a good checksum.
- */
-static void
-efx_rx_packet_gro(struct efx_channel *channel, struct efx_rx_buffer *rx_buf,
-		  unsigned int n_frags, u8 *eh)
-{
-	struct napi_struct *napi = &channel->napi_str;
-	gro_result_t gro_result;
-	struct efx_nic *efx = channel->efx;
-	struct sk_buff *skb;
-
-	skb = napi_get_frags(napi);
-	if (unlikely(!skb)) {
-		struct efx_rx_queue *rx_queue;
-
-		rx_queue = efx_channel_get_rx_queue(channel);
-		efx_free_rx_buffers(rx_queue, rx_buf, n_frags);
-		return;
-	}
-
-	if (efx->net_dev->features & NETIF_F_RXHASH)
-		skb_set_hash(skb, efx_rx_buf_hash(efx, eh),
-			     PKT_HASH_TYPE_L3);
-	skb->ip_summed = ((rx_buf->flags & EFX_RX_PKT_CSUMMED) ?
-			  CHECKSUM_UNNECESSARY : CHECKSUM_NONE);
-	skb->csum_level = !!(rx_buf->flags & EFX_RX_PKT_CSUM_LEVEL);
-
-	for (;;) {
-		skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags,
-				   rx_buf->page, rx_buf->page_offset,
-				   rx_buf->len);
-		rx_buf->page = NULL;
-		skb->len += rx_buf->len;
-		if (skb_shinfo(skb)->nr_frags == n_frags)
-			break;
-
-		rx_buf = efx_rx_buf_next(&channel->rx_queue, rx_buf);
-	}
-
-	skb->data_len = skb->len;
-	skb->truesize += n_frags * efx->rx_buffer_truesize;
-
-	skb_record_rx_queue(skb, channel->rx_queue.core_index);
-
-	gro_result = napi_gro_frags(napi);
-	if (gro_result != GRO_DROP)
-		channel->irq_mod_score += 2;
-}
-
 /* Allocate and construct an SKB around page fragments */
 static struct sk_buff *efx_rx_mk_skb(struct efx_channel *channel,
 				     struct efx_rx_buffer *rx_buf,
@@ -638,6 +255,126 @@ static void efx_rx_deliver(struct efx_channel *channel, u8 *eh,
 		netif_receive_skb(skb);
 }
 
+/** efx_do_xdp: perform XDP processing on a received packet
+ *
+ * Returns true if packet should still be delivered.
+ */
+static bool efx_do_xdp(struct efx_nic *efx, struct efx_channel *channel,
+		       struct efx_rx_buffer *rx_buf, u8 **ehp)
+{
+	u8 rx_prefix[EFX_MAX_RX_PREFIX_SIZE];
+	struct efx_rx_queue *rx_queue;
+	struct bpf_prog *xdp_prog;
+	struct xdp_frame *xdpf;
+	struct xdp_buff xdp;
+	u32 xdp_act;
+	s16 offset;
+	int err;
+
+	rcu_read_lock();
+	xdp_prog = rcu_dereference(efx->xdp_prog);
+	if (!xdp_prog) {
+		rcu_read_unlock();
+		return true;
+	}
+
+	rx_queue = efx_channel_get_rx_queue(channel);
+
+	if (unlikely(channel->rx_pkt_n_frags > 1)) {
+		/* We can't do XDP on fragmented packets - drop. */
+		rcu_read_unlock();
+		efx_free_rx_buffers(rx_queue, rx_buf,
+				    channel->rx_pkt_n_frags);
+		if (net_ratelimit())
+			netif_err(efx, rx_err, efx->net_dev,
+				  "XDP is not possible with multiple receive fragments (%d)\n",
+				  channel->rx_pkt_n_frags);
+		channel->n_rx_xdp_bad_drops++;
+		return false;
+	}
+
+	dma_sync_single_for_cpu(&efx->pci_dev->dev, rx_buf->dma_addr,
+				rx_buf->len, DMA_FROM_DEVICE);
+
+	/* Save the rx prefix. */
+	EFX_WARN_ON_PARANOID(efx->rx_prefix_size > EFX_MAX_RX_PREFIX_SIZE);
+	memcpy(rx_prefix, *ehp - efx->rx_prefix_size,
+	       efx->rx_prefix_size);
+
+	xdp.data = *ehp;
+	xdp.data_hard_start = xdp.data - XDP_PACKET_HEADROOM;
+
+	/* No support yet for XDP metadata */
+	xdp_set_data_meta_invalid(&xdp);
+	xdp.data_end = xdp.data + rx_buf->len;
+	xdp.rxq = &rx_queue->xdp_rxq_info;
+
+	xdp_act = bpf_prog_run_xdp(xdp_prog, &xdp);
+	rcu_read_unlock();
+
+	offset = (u8 *)xdp.data - *ehp;
+
+	switch (xdp_act) {
+	case XDP_PASS:
+		/* Fix up rx prefix. */
+		if (offset) {
+			*ehp += offset;
+			rx_buf->page_offset += offset;
+			rx_buf->len -= offset;
+			memcpy(*ehp - efx->rx_prefix_size, rx_prefix,
+			       efx->rx_prefix_size);
+		}
+		break;
+
+	case XDP_TX:
+		/* Buffer ownership passes to tx on success. */
+		xdpf = convert_to_xdp_frame(&xdp);
+		err = efx_xdp_tx_buffers(efx, 1, &xdpf, true);
+		if (unlikely(err != 1)) {
+			efx_free_rx_buffers(rx_queue, rx_buf, 1);
+			if (net_ratelimit())
+				netif_err(efx, rx_err, efx->net_dev,
+					  "XDP TX failed (%d)\n", err);
+			channel->n_rx_xdp_bad_drops++;
+			trace_xdp_exception(efx->net_dev, xdp_prog, xdp_act);
+		} else {
+			channel->n_rx_xdp_tx++;
+		}
+		break;
+
+	case XDP_REDIRECT:
+		err = xdp_do_redirect(efx->net_dev, &xdp, xdp_prog);
+		if (unlikely(err)) {
+			efx_free_rx_buffers(rx_queue, rx_buf, 1);
+			if (net_ratelimit())
+				netif_err(efx, rx_err, efx->net_dev,
+					  "XDP redirect failed (%d)\n", err);
+			channel->n_rx_xdp_bad_drops++;
+			trace_xdp_exception(efx->net_dev, xdp_prog, xdp_act);
+		} else {
+			channel->n_rx_xdp_redirect++;
+		}
+		break;
+
+	default:
+		bpf_warn_invalid_xdp_action(xdp_act);
+		efx_free_rx_buffers(rx_queue, rx_buf, 1);
+		channel->n_rx_xdp_bad_drops++;
+		trace_xdp_exception(efx->net_dev, xdp_prog, xdp_act);
+		break;
+
+	case XDP_ABORTED:
+		trace_xdp_exception(efx->net_dev, xdp_prog, xdp_act);
+		/* Fall through */
+	case XDP_DROP:
+		efx_free_rx_buffers(rx_queue, rx_buf, 1);
+		channel->n_rx_xdp_drops++;
+		break;
+	}
+
+	return xdp_act == XDP_PASS;
+}
+
 /* Handle a received packet.  Second half: Touches packet payload. */
 void __efx_rx_packet(struct efx_channel *channel)
 {
@@ -666,6 +403,9 @@ void __efx_rx_packet(struct efx_channel *channel)
 		goto out;
 	}
 
+	if (!efx_do_xdp(efx, channel, rx_buf, &eh))
+		goto out;
+
 	if (unlikely(!(efx->net_dev->features & NETIF_F_RXCSUM)))
 		rx_buf->flags &= ~EFX_RX_PKT_CSUMMED;
 
@@ -677,155 +417,6 @@ out:
 	channel->rx_pkt_n_frags = 0;
 }
 
-int efx_probe_rx_queue(struct efx_rx_queue *rx_queue)
-{
-	struct efx_nic *efx = rx_queue->efx;
-	unsigned int entries;
-	int rc;
-
-	/* Create the smallest power-of-two aligned ring */
-	entries = max(roundup_pow_of_two(efx->rxq_entries), EFX_MIN_DMAQ_SIZE);
-	EFX_WARN_ON_PARANOID(entries > EFX_MAX_DMAQ_SIZE);
-	rx_queue->ptr_mask = entries - 1;
-
-	netif_dbg(efx, probe, efx->net_dev,
-		  "creating RX queue %d size %#x mask %#x\n",
-		  efx_rx_queue_index(rx_queue), efx->rxq_entries,
-		  rx_queue->ptr_mask);
-
-	/* Allocate RX buffers */
-	rx_queue->buffer = kcalloc(entries, sizeof(*rx_queue->buffer),
-				   GFP_KERNEL);
-	if (!rx_queue->buffer)
-		return -ENOMEM;
-
-	rc = efx_nic_probe_rx(rx_queue);
-	if (rc) {
-		kfree(rx_queue->buffer);
-		rx_queue->buffer = NULL;
-	}
-
-	return rc;
-}
-
-static void efx_init_rx_recycle_ring(struct efx_nic *efx,
-				     struct efx_rx_queue *rx_queue)
-{
-	unsigned int bufs_in_recycle_ring, page_ring_size;
-
-	/* Set the RX recycle ring size */
-#ifdef CONFIG_PPC64
-	bufs_in_recycle_ring = EFX_RECYCLE_RING_SIZE_IOMMU;
-#else
-	if (iommu_present(&pci_bus_type))
-		bufs_in_recycle_ring = EFX_RECYCLE_RING_SIZE_IOMMU;
-	else
-		bufs_in_recycle_ring = EFX_RECYCLE_RING_SIZE_NOIOMMU;
-#endif /* CONFIG_PPC64 */
-
-	page_ring_size = roundup_pow_of_two(bufs_in_recycle_ring /
-					    efx->rx_bufs_per_page);
-	rx_queue->page_ring = kcalloc(page_ring_size,
-				      sizeof(*rx_queue->page_ring), GFP_KERNEL);
-	rx_queue->page_ptr_mask = page_ring_size - 1;
-}
-
-void efx_init_rx_queue(struct efx_rx_queue *rx_queue)
-{
-	struct efx_nic *efx = rx_queue->efx;
-	unsigned int max_fill, trigger, max_trigger;
-
-	netif_dbg(rx_queue->efx, drv, rx_queue->efx->net_dev,
-		  "initialising RX queue %d\n", efx_rx_queue_index(rx_queue));
-
-	/* Initialise ptr fields */
-	rx_queue->added_count = 0;
-	rx_queue->notified_count = 0;
-	rx_queue->removed_count = 0;
-	rx_queue->min_fill = -1U;
-	efx_init_rx_recycle_ring(efx, rx_queue);
-
-	rx_queue->page_remove = 0;
-	rx_queue->page_add = rx_queue->page_ptr_mask + 1;
-	rx_queue->page_recycle_count = 0;
-	rx_queue->page_recycle_failed = 0;
-	rx_queue->page_recycle_full = 0;
-
-	/* Initialise limit fields */
-	max_fill = efx->rxq_entries - EFX_RXD_HEAD_ROOM;
-	max_trigger =
-		max_fill - efx->rx_pages_per_batch * efx->rx_bufs_per_page;
-	if (rx_refill_threshold != 0) {
-		trigger = max_fill * min(rx_refill_threshold, 100U) / 100U;
-		if (trigger > max_trigger)
-			trigger = max_trigger;
-	} else {
-		trigger = max_trigger;
-	}
-
-	rx_queue->max_fill = max_fill;
-	rx_queue->fast_fill_trigger = trigger;
-	rx_queue->refill_enabled = true;
-
-	/* Set up RX descriptor ring */
-	efx_nic_init_rx(rx_queue);
-}
-
-void efx_fini_rx_queue(struct efx_rx_queue *rx_queue)
-{
-	int i;
-	struct efx_nic *efx = rx_queue->efx;
-	struct efx_rx_buffer *rx_buf;
-
-	netif_dbg(rx_queue->efx, drv, rx_queue->efx->net_dev,
-		  "shutting down RX queue %d\n", efx_rx_queue_index(rx_queue));
-
-	del_timer_sync(&rx_queue->slow_fill);
-
-	/* Release RX buffers from the current read ptr to the write ptr */
-	if (rx_queue->buffer) {
-		for (i = rx_queue->removed_count; i < rx_queue->added_count;
-		     i++) {
-			unsigned index = i & rx_queue->ptr_mask;
-			rx_buf = efx_rx_buffer(rx_queue, index);
-			efx_fini_rx_buffer(rx_queue, rx_buf);
-		}
-	}
-
-	/* Unmap and release the pages in the recycle ring. Remove the ring. */
-	for (i = 0; i <= rx_queue->page_ptr_mask; i++) {
-		struct page *page = rx_queue->page_ring[i];
-		struct efx_rx_page_state *state;
-
-		if (page == NULL)
-			continue;
-
-		state = page_address(page);
-		dma_unmap_page(&efx->pci_dev->dev, state->dma_addr,
-			       PAGE_SIZE << efx->rx_buffer_order,
-			       DMA_FROM_DEVICE);
-		put_page(page);
-	}
-	kfree(rx_queue->page_ring);
-	rx_queue->page_ring = NULL;
-}
-
-void efx_remove_rx_queue(struct efx_rx_queue *rx_queue)
-{
-	netif_dbg(rx_queue->efx, drv, rx_queue->efx->net_dev,
-		  "destroying RX queue %d\n", efx_rx_queue_index(rx_queue));
-
-	efx_nic_remove_rx(rx_queue);
-
-	kfree(rx_queue->buffer);
-	rx_queue->buffer = NULL;
-}
-
-
-module_param(rx_refill_threshold, uint, 0444);
-MODULE_PARM_DESC(rx_refill_threshold,
-		 "RX descriptor ring refill threshold (%)");
-
 #ifdef CONFIG_RFS_ACCEL
 
 static void efx_filter_rfs_work(struct work_struct *data)
@@ -841,6 +432,7 @@ static void efx_filter_rfs_work(struct work_struct *data)
 
 	rc = efx->type->filter_insert(efx, &req->spec, true);
 	if (rc >= 0)
+		/* Discard 'priority' part of EF10+ filter ID (mcdi_filters) */
 		rc %= efx->type->max_rx_ip_filters;
 	if (efx->rps_hash_table) {
 		spin_lock_bh(&efx->rps_hash_lock);
@@ -865,8 +457,9 @@ static void efx_filter_rfs_work(struct work_struct *data)
 		 * later.
 		 */
 		mutex_lock(&efx->rps_mutex);
+		if (channel->rps_flow_id[rc] == RPS_FLOW_ID_INVALID)
+			channel->rfs_filter_count++;
 		channel->rps_flow_id[rc] = req->flow_id;
-		++channel->rfs_filters_added;
 		mutex_unlock(&efx->rps_mutex);
 
 		if (req->spec.ether_type == htons(ETH_P_IP))
@@ -883,6 +476,28 @@ static void efx_filter_rfs_work(struct work_struct *data)
 				   req->spec.rem_host, ntohs(req->spec.rem_port),
 				   req->spec.loc_host, ntohs(req->spec.loc_port),
 				   req->rxq_index, req->flow_id, rc, arfs_id);
+		channel->n_rfs_succeeded++;
+	} else {
+		if (req->spec.ether_type == htons(ETH_P_IP))
+			netif_dbg(efx, rx_status, efx->net_dev,
+				  "failed to steer %s %pI4:%u:%pI4:%u to queue %u [flow %u rc %d id %u]\n",
+				  (req->spec.ip_proto == IPPROTO_TCP) ? "TCP" : "UDP",
+				  req->spec.rem_host, ntohs(req->spec.rem_port),
+				  req->spec.loc_host, ntohs(req->spec.loc_port),
+				  req->rxq_index, req->flow_id, rc, arfs_id);
+		else
+			netif_dbg(efx, rx_status, efx->net_dev,
+				  "failed to steer %s [%pI6]:%u:[%pI6]:%u to queue %u [flow %u rc %d id %u]\n",
+				  (req->spec.ip_proto == IPPROTO_TCP) ? "TCP" : "UDP",
+				  req->spec.rem_host, ntohs(req->spec.rem_port),
+				  req->spec.loc_host, ntohs(req->spec.loc_port),
+				  req->rxq_index, req->flow_id, rc, arfs_id);
+		channel->n_rfs_failed++;
+		/* We're overloading the NIC's filter tables, so let's do a
+		 * chunk of extra expiry work.
+		 */
+		__efx_filter_rfs_expire(channel, min(channel->rfs_filter_count,
+						     100u));
 	}
 
 	/* Release references */
@@ -992,74 +607,46 @@ out_clear:
 	return rc;
 }
 
-bool __efx_filter_rfs_expire(struct efx_nic *efx, unsigned int quota)
+bool __efx_filter_rfs_expire(struct efx_channel *channel, unsigned int quota)
 {
 	bool (*expire_one)(struct efx_nic *efx, u32 flow_id, unsigned int index);
-	unsigned int channel_idx, index, size;
+	struct efx_nic *efx = channel->efx;
+	unsigned int index, size, start;
 	u32 flow_id;
 
 	if (!mutex_trylock(&efx->rps_mutex))
 		return false;
 	expire_one = efx->type->filter_rfs_expire_one;
-	channel_idx = efx->rps_expire_channel;
-	index = efx->rps_expire_index;
+	index = channel->rfs_expire_index;
+	start = index;
 	size = efx->type->max_rx_ip_filters;
-	while (quota--) {
-		struct efx_channel *channel = efx_get_channel(efx, channel_idx);
+	while (quota) {
 		flow_id = channel->rps_flow_id[index];
 
-		if (flow_id != RPS_FLOW_ID_INVALID &&
-		    expire_one(efx, flow_id, index)) {
-			netif_info(efx, rx_status, efx->net_dev,
-				   "expired filter %d [queue %u flow %u]\n",
-				   index, channel_idx, flow_id);
-			channel->rps_flow_id[index] = RPS_FLOW_ID_INVALID;
+		if (flow_id != RPS_FLOW_ID_INVALID) {
+			quota--;
+			if (expire_one(efx, flow_id, index)) {
+				netif_info(efx, rx_status, efx->net_dev,
+					   "expired filter %d [channel %u flow %u]\n",
+					   index, channel->channel, flow_id);
+				channel->rps_flow_id[index] = RPS_FLOW_ID_INVALID;
+				channel->rfs_filter_count--;
+			}
 		}
-		if (++index == size) {
-			if (++channel_idx == efx->n_channels)
-				channel_idx = 0;
+		if (++index == size)
 			index = 0;
-		}
+		/* If we were called with a quota that exceeds the total number
+		 * of filters in the table (which shouldn't happen, but could
+		 * if two callers race), ensure that we don't loop forever -
+		 * stop when we've examined every row of the table.
+		 */
+		if (index == start)
+			break;
 	}
-	efx->rps_expire_channel = channel_idx;
-	efx->rps_expire_index = index;
 
+	channel->rfs_expire_index = index;
 	mutex_unlock(&efx->rps_mutex);
 	return true;
 }
 
 #endif /* CONFIG_RFS_ACCEL */
-
-/**
- * efx_filter_is_mc_recipient - test whether spec is a multicast recipient
- * @spec: Specification to test
- *
- * Return: %true if the specification is a non-drop RX filter that
- * matches a local MAC address I/G bit value of 1 or matches a local
- * IPv4 or IPv6 address value in the respective multicast address
- * range.  Otherwise %false.
- */
-bool efx_filter_is_mc_recipient(const struct efx_filter_spec *spec)
-{
-	if (!(spec->flags & EFX_FILTER_FLAG_RX) ||
-	    spec->dmaq_id == EFX_FILTER_RX_DMAQ_ID_DROP)
-		return false;
-
-	if (spec->match_flags &
-	    (EFX_FILTER_MATCH_LOC_MAC | EFX_FILTER_MATCH_LOC_MAC_IG) &&
-	    is_multicast_ether_addr(spec->loc_mac))
-		return true;
-
-	if ((spec->match_flags &
-	     (EFX_FILTER_MATCH_ETHER_TYPE | EFX_FILTER_MATCH_LOC_HOST)) ==
-	    (EFX_FILTER_MATCH_ETHER_TYPE | EFX_FILTER_MATCH_LOC_HOST)) {
-		if (spec->ether_type == htons(ETH_P_IP) &&
-		    ipv4_is_multicast(spec->loc_host[0]))
-			return true;
-		if (spec->ether_type == htons(ETH_P_IPV6) &&
-		    ((const u8 *)spec->loc_host)[0] == 0xff)
-			return true;
-	}
-
-	return false;
-}