ice: Add support for XDP multi-buffer on Rx side

Ice driver needs to be a bit reworked on Rx data path in order to
support multi-buffer XDP. For skb path, it currently works in a way that
Rx ring carries pointer to skb so if driver didn't manage to combine
fragmented frame at current NAPI instance, it can restore the state on
next instance and keep looking for last fragment (so descriptor with EOP
bit set). What needs to be achieved is that xdp_buff needs to be
combined in such way (linear + frags part) in the first place. Then skb
will be ready to go in case of XDP_PASS or BPF program being not present
on interface. If BPF program is there, it would work on multi-buffer
XDP. At this point xdp_buff resides directly on Rx ring, so given the
fact that skb will be built straight from xdp_buff, there will be no
further need to carry skb on Rx ring.

Besides removing skb pointer from Rx ring, lots of members have been
moved around within ice_rx_ring. First and foremost reason was to place
rx_buf with xdp_buff on the same cacheline. This means that once we
touch rx_buf (which is a preceding step before touching xdp_buff),
xdp_buff will already be hot in cache. Second thing was that xdp_rxq is
used rather rarely and it occupies a separate cacheline, so maybe it is
better to have it at the end of ice_rx_ring.

Other change that affects ice_rx_ring is the introduction of
ice_rx_ring::first_desc. Its purpose is twofold - first is to propagate
rx_buf->act to all the parts of current xdp_buff after running XDP
program, so that ice_put_rx_buf() that got moved out of the main Rx
processing loop will be able to tak an appriopriate action on each
buffer. Second is for ice_construct_skb().

ice_construct_skb() has a copybreak mechanism which had an explicit
impact on xdp_buff->skb conversion in the new approach when legacy Rx
flag is toggled. It works in a way that linear part is 256 bytes long,
if frame is bigger than that, remaining bytes are going as a frag to
skb_shared_info.

This means while memcpying frags from xdp_buff to newly allocated skb,
care needs to be taken when picking the destination frag array entry.
Upon the time ice_construct_skb() is called, when dealing with
fragmented frame, current rx_buf points to the *last* fragment, but
copybreak needs to be done against the first one.  That's where
ice_rx_ring::first_desc helps.

When frame building spans across NAPI polls (DD bit is not set on
current descriptor and xdp->data is not NULL) with current Rx buffer
handling state there might be some problems.
Since calls to ice_put_rx_buf() were pulled out of the main Rx
processing loop and were scoped from cached_ntc to current ntc, remember
that now mentioned function relies on rx_buf->act, which is set within
ice_run_xdp(). ice_run_xdp() is called when EOP bit was found, so
currently we could put Rx buffer with rx_buf->act being *uninitialized*.
To address this, change scoping to rely on first_desc on both boundaries
instead.

This also implies that cleaned_count which is used as an input to
ice_alloc_rx_buffers() and tells how many new buffers should be refilled
has to be adjusted. If it stayed as is, what could happen is a case
where ntc would go over ntu.

Therefore, remove cleaned_count altogether and use against allocing
routine newly introduced ICE_RX_DESC_UNUSED() macro which is an
equivalent of ICE_DESC_UNUSED() dedicated for Rx side and based on
struct ice_rx_ring::first_desc instead of next_to_clean.

Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Link: https://lore.kernel.org/bpf/20230131204506.219292-11-maciej.fijalkowski@intel.com

1 files changed, 129 insertions, 73 deletions

diff --git a/drivers/net/ethernet/intel/ice/ice_txrx.c b/drivers/net/ethernet/intel/ice/ice_txrx.c
index 8db712254cb5..05028921fadd 100644
--- a/drivers/net/ethernet/intel/ice/ice_txrx.c
+++ b/drivers/net/ethernet/intel/ice/ice_txrx.c
@@ -382,6 +382,7 @@ err:
  */
 void ice_clean_rx_ring(struct ice_rx_ring *rx_ring)
 {
+	struct xdp_buff *xdp = &rx_ring->xdp;
 	struct device *dev = rx_ring->dev;
 	u32 size;
 	u16 i;
@@ -390,16 +391,16 @@ void ice_clean_rx_ring(struct ice_rx_ring *rx_ring)
 	if (!rx_ring->rx_buf)
 		return;
 
-	if (rx_ring->skb) {
-		dev_kfree_skb(rx_ring->skb);
-		rx_ring->skb = NULL;
-	}
-
 	if (rx_ring->xsk_pool) {
 		ice_xsk_clean_rx_ring(rx_ring);
 		goto rx_skip_free;
 	}
 
+	if (xdp->data) {
+		xdp_return_buff(xdp);
+		xdp->data = NULL;
+	}
+
 	/* 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];
@@ -437,6 +438,7 @@ rx_skip_free:
 
 	rx_ring->next_to_alloc = 0;
 	rx_ring->next_to_clean = 0;
+	rx_ring->first_desc = 0;
 	rx_ring->next_to_use = 0;
 }
 
@@ -506,6 +508,7 @@ int ice_setup_rx_ring(struct ice_rx_ring *rx_ring)
 
 	rx_ring->next_to_use = 0;
 	rx_ring->next_to_clean = 0;
+	rx_ring->first_desc = 0;
 
 	if (ice_is_xdp_ena_vsi(rx_ring->vsi))
 		WRITE_ONCE(rx_ring->xdp_prog, rx_ring->vsi->xdp_prog);
@@ -598,6 +601,8 @@ out_failure:
 	}
 exit:
 	rx_buf->act = ret;
+	if (unlikely(xdp_buff_has_frags(xdp)))
+		ice_set_rx_bufs_act(xdp, rx_ring, ret);
 }
 
 /**
@@ -721,7 +726,7 @@ ice_alloc_mapped_page(struct ice_rx_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_rx_ring *rx_ring, u16 cleaned_count)
+bool ice_alloc_rx_bufs(struct ice_rx_ring *rx_ring, unsigned int cleaned_count)
 {
 	union ice_32b_rx_flex_desc *rx_desc;
 	u16 ntu = rx_ring->next_to_use;
@@ -838,26 +843,44 @@ ice_can_reuse_rx_page(struct ice_rx_buf *rx_buf)
 }
 
 /**
- * ice_add_rx_frag - Add contents of Rx buffer to sk_buff as a frag
+ * ice_add_xdp_frag - Add contents of Rx buffer to xdp buf as a frag
  * @rx_ring: Rx descriptor ring to transact packets on
- * @xdp: XDP buffer
+ * @xdp: xdp buff to place the data into
  * @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.
+ * This function will add the data contained in rx_buf->page to the xdp buf.
+ * It will just attach the page as a frag.
  */
-static void
-ice_add_rx_frag(struct ice_rx_ring *rx_ring, struct xdp_buff *xdp,
-		struct ice_rx_buf *rx_buf, struct sk_buff *skb,
-		unsigned int size)
+static int
+ice_add_xdp_frag(struct ice_rx_ring *rx_ring, struct xdp_buff *xdp,
+		 struct ice_rx_buf *rx_buf, const unsigned int size)
 {
+	struct skb_shared_info *sinfo = xdp_get_shared_info_from_buff(xdp);
+
 	if (!size)
-		return;
-	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buf->page,
-			rx_buf->page_offset, size, xdp->frame_sz);
+		return 0;
+
+	if (!xdp_buff_has_frags(xdp)) {
+		sinfo->nr_frags = 0;
+		sinfo->xdp_frags_size = 0;
+		xdp_buff_set_frags_flag(xdp);
+	}
+
+	if (unlikely(sinfo->nr_frags == MAX_SKB_FRAGS)) {
+		if (unlikely(xdp_buff_has_frags(xdp)))
+			ice_set_rx_bufs_act(xdp, rx_ring, ICE_XDP_CONSUMED);
+		return -ENOMEM;
+	}
+
+	__skb_fill_page_desc_noacc(sinfo, sinfo->nr_frags++, rx_buf->page,
+				   rx_buf->page_offset, size);
+	sinfo->xdp_frags_size += size;
+
+	if (page_is_pfmemalloc(rx_buf->page))
+		xdp_buff_set_frag_pfmemalloc(xdp);
+
+	return 0;
 }
 
 /**
@@ -928,19 +951,25 @@ ice_get_rx_buf(struct ice_rx_ring *rx_ring, const unsigned int size,
 /**
  * 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.
+ * This function builds an skb around an existing XDP buffer, taking care
+ * to set up the skb correctly and avoid any memcpy overhead. Driver has
+ * already combined frags (if any) to skb_shared_info.
  */
 static struct sk_buff *
-ice_build_skb(struct ice_rx_ring *rx_ring, struct ice_rx_buf *rx_buf,
-	      struct xdp_buff *xdp)
+ice_build_skb(struct ice_rx_ring *rx_ring, struct xdp_buff *xdp)
 {
 	u8 metasize = xdp->data - xdp->data_meta;
+	struct skb_shared_info *sinfo = NULL;
+	unsigned int nr_frags;
 	struct sk_buff *skb;
 
+	if (unlikely(xdp_buff_has_frags(xdp))) {
+		sinfo = xdp_get_shared_info_from_buff(xdp);
+		nr_frags = sinfo->nr_frags;
+	}
+
 	/* 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
@@ -963,6 +992,12 @@ ice_build_skb(struct ice_rx_ring *rx_ring, struct ice_rx_buf *rx_buf,
 	if (metasize)
 		skb_metadata_set(skb, metasize);
 
+	if (unlikely(xdp_buff_has_frags(xdp)))
+		xdp_update_skb_shared_info(skb, nr_frags,
+					   sinfo->xdp_frags_size,
+					   nr_frags * xdp->frame_sz,
+					   xdp_buff_is_frag_pfmemalloc(xdp));
+
 	return skb;
 }
 
@@ -977,22 +1012,30 @@ ice_build_skb(struct ice_rx_ring *rx_ring, struct ice_rx_buf *rx_buf,
  * skb correctly.
  */
 static struct sk_buff *
-ice_construct_skb(struct ice_rx_ring *rx_ring, struct ice_rx_buf *rx_buf,
-		  struct xdp_buff *xdp)
+ice_construct_skb(struct ice_rx_ring *rx_ring, struct xdp_buff *xdp)
 {
 	unsigned int size = xdp->data_end - xdp->data;
+	struct skb_shared_info *sinfo = NULL;
+	struct ice_rx_buf *rx_buf;
+	unsigned int nr_frags = 0;
 	unsigned int headlen;
 	struct sk_buff *skb;
 
 	/* prefetch first cache line of first page */
 	net_prefetch(xdp->data);
 
+	if (unlikely(xdp_buff_has_frags(xdp))) {
+		sinfo = xdp_get_shared_info_from_buff(xdp);
+		nr_frags = sinfo->nr_frags;
+	}
+
 	/* 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;
 
+	rx_buf = &rx_ring->rx_buf[rx_ring->first_desc];
 	skb_record_rx_queue(skb, rx_ring->q_index);
 	/* Determine available headroom for copy */
 	headlen = size;
@@ -1006,6 +1049,14 @@ ice_construct_skb(struct ice_rx_ring *rx_ring, struct ice_rx_buf *rx_buf,
 	/* if we exhaust the linear part then add what is left as a frag */
 	size -= headlen;
 	if (size) {
+		/* besides adding here a partial frag, we are going to add
+		 * frags from xdp_buff, make sure there is enough space for
+		 * them
+		 */
+		if (unlikely(nr_frags >= MAX_SKB_FRAGS - 1)) {
+			dev_kfree_skb(skb);
+			return NULL;
+		}
 		skb_add_rx_frag(skb, 0, rx_buf->page,
 				rx_buf->page_offset + headlen, size,
 				xdp->frame_sz);
@@ -1015,7 +1066,19 @@ ice_construct_skb(struct ice_rx_ring *rx_ring, struct ice_rx_buf *rx_buf,
 		 * need for adjusting page offset and we can reuse this buffer
 		 * as-is
 		 */
-		rx_buf->act = ICE_XDP_CONSUMED;
+		rx_buf->act = ICE_SKB_CONSUMED;
+	}
+
+	if (unlikely(xdp_buff_has_frags(xdp))) {
+		struct skb_shared_info *skinfo = skb_shinfo(skb);
+
+		memcpy(&skinfo->frags[skinfo->nr_frags], &sinfo->frags[0],
+		       sizeof(skb_frag_t) * nr_frags);
+
+		xdp_update_skb_shared_info(skb, skinfo->nr_frags + nr_frags,
+					   sinfo->xdp_frags_size,
+					   nr_frags * xdp->frame_sz,
+					   xdp_buff_is_frag_pfmemalloc(xdp));
 	}
 
 	return skb;
@@ -1065,17 +1128,16 @@ ice_put_rx_buf(struct ice_rx_ring *rx_ring, struct ice_rx_buf *rx_buf)
 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 offset = rx_ring->rx_offset;
 	struct xdp_buff *xdp = &rx_ring->xdp;
 	struct ice_tx_ring *xdp_ring = NULL;
-	struct sk_buff *skb = rx_ring->skb;
 	struct bpf_prog *xdp_prog = NULL;
 	u32 ntc = rx_ring->next_to_clean;
 	u32 cnt = rx_ring->count;
 	u32 cached_ntc = ntc;
 	u32 xdp_xmit = 0;
 	bool failure;
+	u32 first;
 
 	/* Frame size depend on rx_ring setup when PAGE_SIZE=4K */
 #if (PAGE_SIZE < 8192)
@@ -1090,7 +1152,7 @@ int ice_clean_rx_irq(struct ice_rx_ring *rx_ring, int budget)
 	while (likely(total_rx_pkts < (unsigned int)budget)) {
 		union ice_32b_rx_flex_desc *rx_desc;
 		struct ice_rx_buf *rx_buf;
-		unsigned char *hard_start;
+		struct sk_buff *skb;
 		unsigned int size;
 		u16 stat_err_bits;
 		u16 vlan_tag = 0;
@@ -1123,7 +1185,6 @@ int ice_clean_rx_irq(struct ice_rx_ring *rx_ring, int budget)
 				ice_vc_fdir_irq_handler(ctrl_vsi, rx_desc);
 			if (++ntc == cnt)
 				ntc = 0;
-			cleaned_count++;
 			continue;
 		}
 
@@ -1133,56 +1194,54 @@ int ice_clean_rx_irq(struct ice_rx_ring *rx_ring, int budget)
 		/* retrieve a buffer from the ring */
 		rx_buf = ice_get_rx_buf(rx_ring, size, ntc);
 
-		if (!size) {
-			xdp->data = NULL;
-			xdp->data_end = NULL;
-			xdp->data_hard_start = NULL;
-			xdp->data_meta = NULL;
-			goto construct_skb;
-		}
+		if (!xdp->data) {
+			void *hard_start;
 
-		hard_start = page_address(rx_buf->page) + rx_buf->page_offset -
-			     offset;
-		xdp_prepare_buff(xdp, hard_start, offset, size, !!offset);
+			hard_start = page_address(rx_buf->page) + rx_buf->page_offset -
+				     offset;
+			xdp_prepare_buff(xdp, hard_start, offset, size, !!offset);
 #if (PAGE_SIZE > 4096)
-		/* At larger PAGE_SIZE, frame_sz depend on len size */
-		xdp->frame_sz = ice_rx_frame_truesize(rx_ring, size);
+			/* At larger PAGE_SIZE, frame_sz depend on len size */
+			xdp->frame_sz = ice_rx_frame_truesize(rx_ring, size);
 #endif
+			xdp_buff_clear_frags_flag(xdp);
+		} else if (ice_add_xdp_frag(rx_ring, xdp, rx_buf, size)) {
+			break;
+		}
+		if (++ntc == cnt)
+			ntc = 0;
+
+		/* skip if it is NOP desc */
+		if (ice_is_non_eop(rx_ring, rx_desc))
+			continue;
 
 		ice_run_xdp(rx_ring, xdp, xdp_prog, xdp_ring, rx_buf);
 		if (rx_buf->act == ICE_XDP_PASS)
 			goto construct_skb;
-		total_rx_bytes += size;
+		total_rx_bytes += xdp_get_buff_len(xdp);
 		total_rx_pkts++;
 
-		cleaned_count++;
-		if (++ntc == cnt)
-			ntc = 0;
+		xdp->data = NULL;
+		rx_ring->first_desc = ntc;
 		continue;
 construct_skb:
-		if (skb) {
-			ice_add_rx_frag(rx_ring, xdp, 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);
-		}
+		if (likely(ice_ring_uses_build_skb(rx_ring)))
+			skb = ice_build_skb(rx_ring, xdp);
+		else
+			skb = ice_construct_skb(rx_ring, xdp);
 		/* exit if we failed to retrieve a buffer */
 		if (!skb) {
-			rx_ring->ring_stats->rx_stats.alloc_buf_failed++;
-			if (rx_buf)
-				rx_buf->pagecnt_bias++;
+			rx_ring->ring_stats->rx_stats.alloc_page_failed++;
+			rx_buf->act = ICE_XDP_CONSUMED;
+			if (unlikely(xdp_buff_has_frags(xdp)))
+				ice_set_rx_bufs_act(xdp, rx_ring,
+						    ICE_XDP_CONSUMED);
+			xdp->data = NULL;
+			rx_ring->first_desc = ntc;
 			break;
 		}
-
-		if (++ntc == cnt)
-			ntc = 0;
-		cleaned_count++;
-
-		/* skip if it is NOP desc */
-		if (ice_is_non_eop(rx_ring, rx_desc))
-			continue;
+		xdp->data = NULL;
+		rx_ring->first_desc = ntc;
 
 		stat_err_bits = BIT(ICE_RX_FLEX_DESC_STATUS0_RXE_S);
 		if (unlikely(ice_test_staterr(rx_desc->wb.status_error0,
@@ -1194,10 +1253,8 @@ construct_skb:
 		vlan_tag = ice_get_vlan_tag_from_rx_desc(rx_desc);
 
 		/* pad the skb if needed, to make a valid ethernet frame */
-		if (eth_skb_pad(skb)) {
-			skb = NULL;
+		if (eth_skb_pad(skb))
 			continue;
-		}
 
 		/* probably a little skewed due to removing CRC */
 		total_rx_bytes += skb->len;
@@ -1211,13 +1268,13 @@ construct_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);
-		skb = NULL;
 
 		/* update budget accounting */
 		total_rx_pkts++;
 	}
 
-	while (cached_ntc != ntc) {
+	first = rx_ring->first_desc;
+	while (cached_ntc != first) {
 		struct ice_rx_buf *buf = &rx_ring->rx_buf[cached_ntc];
 
 		if (buf->act & (ICE_XDP_TX | ICE_XDP_REDIR)) {
@@ -1235,11 +1292,10 @@ construct_skb:
 	}
 	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_RX_DESC_UNUSED(rx_ring));
 
 	if (xdp_xmit)
 		ice_finalize_xdp_rx(xdp_ring, xdp_xmit);
-	rx_ring->skb = skb;
 
 	if (rx_ring->ring_stats)
 		ice_update_rx_ring_stats(rx_ring, total_rx_pkts,