kernel: pick patches for MediaTek Ethernet from linux-next

Pick patches with several fixes and improvements, preparation for
upcoming WED (TX) [1] as well as basic XDP support [2] with MediaTek's
Filogic SoCs to the mtk_eth_soc driver.

Also pick follow-up patch fixing Ethernet on MT7621 [3].

Tested on Bananapi BPi-R3 (MT7986), Bananapi BPi-R64 (MT7622),
Bananapi BPi-R2 (MT7623), MikroTik RouterBoard M11G (MT7621).

[1]: https://patchwork.kernel.org/project/netdevbpf/list/?series=662108&state=*
[2]: https://patchwork.kernel.org/project/netdevbpf/list/?series=675368&state=*
     (the first part of the series adding wed nodes to mt7986a.dtsi was
      applied to the copy of mt7986a.dtsi in our tree)
[3]: https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net.git/commit/?id=5e69163d3b9931098922b3fc2f8e786af8c1f37e

Signed-off-by: Daniel Golle <daniel@makrotopia.org>

1 files changed, 340 insertions, 0 deletions

diff --git a/target/linux/generic/backport-5.15/706-03-v6.0-net-ethernet-mtk_eth_soc-add-xmit-XDP-support.patch b/target/linux/generic/backport-5.15/706-03-v6.0-net-ethernet-mtk_eth_soc-add-xmit-XDP-support.patch
new file mode 100644
index 0000000000..42b3ea215d
--- /dev/null
+++ b/target/linux/generic/backport-5.15/706-03-v6.0-net-ethernet-mtk_eth_soc-add-xmit-XDP-support.patch
@@ -0,0 +1,340 @@
+From 5886d26fd25bbe26130e3e5f7474b9b3e98a3469 Mon Sep 17 00:00:00 2001
+From: Lorenzo Bianconi <lorenzo@kernel.org>
+Date: Fri, 22 Jul 2022 09:19:39 +0200
+Subject: [PATCH] net: ethernet: mtk_eth_soc: add xmit XDP support
+
+Introduce XDP support for XDP_TX verdict and ndo_xdp_xmit function
+pointer.
+
+Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
+Signed-off-by: David S. Miller <davem@davemloft.net>
+---
+ drivers/net/ethernet/mediatek/mtk_eth_soc.c | 192 +++++++++++++++++---
+ drivers/net/ethernet/mediatek/mtk_eth_soc.h |  10 +-
+ 2 files changed, 180 insertions(+), 22 deletions(-)
+
+--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
++++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+@@ -987,15 +987,26 @@ static void mtk_tx_unmap(struct mtk_eth
+ 		}
+ 	}
+ 
+-	tx_buf->flags = 0;
+-	if (tx_buf->skb &&
+-	    (tx_buf->skb != (struct sk_buff *)MTK_DMA_DUMMY_DESC)) {
+-		if (napi)
+-			napi_consume_skb(tx_buf->skb, napi);
++	if (tx_buf->type == MTK_TYPE_SKB) {
++		if (tx_buf->data &&
++		    tx_buf->data != (void *)MTK_DMA_DUMMY_DESC) {
++			struct sk_buff *skb = tx_buf->data;
++
++			if (napi)
++				napi_consume_skb(skb, napi);
++			else
++				dev_kfree_skb_any(skb);
++		}
++	} else if (tx_buf->data) {
++		struct xdp_frame *xdpf = tx_buf->data;
++
++		if (napi && tx_buf->type == MTK_TYPE_XDP_TX)
++			xdp_return_frame_rx_napi(xdpf);
+ 		else
+-			dev_kfree_skb_any(tx_buf->skb);
++			xdp_return_frame(xdpf);
+ 	}
+-	tx_buf->skb = NULL;
++	tx_buf->flags = 0;
++	tx_buf->data = NULL;
+ }
+ 
+ static void setup_tx_buf(struct mtk_eth *eth, struct mtk_tx_buf *tx_buf,
+@@ -1012,7 +1023,7 @@ static void setup_tx_buf(struct mtk_eth
+ 			dma_unmap_addr_set(tx_buf, dma_addr1, mapped_addr);
+ 			dma_unmap_len_set(tx_buf, dma_len1, size);
+ 		} else {
+-			tx_buf->skb = (struct sk_buff *)MTK_DMA_DUMMY_DESC;
++			tx_buf->data = (void *)MTK_DMA_DUMMY_DESC;
+ 			txd->txd1 = mapped_addr;
+ 			txd->txd2 = TX_DMA_PLEN0(size);
+ 			dma_unmap_addr_set(tx_buf, dma_addr0, mapped_addr);
+@@ -1188,7 +1199,7 @@ static int mtk_tx_map(struct sk_buff *sk
+ 						    soc->txrx.txd_size);
+ 			if (new_desc)
+ 				memset(tx_buf, 0, sizeof(*tx_buf));
+-			tx_buf->skb = (struct sk_buff *)MTK_DMA_DUMMY_DESC;
++			tx_buf->data = (void *)MTK_DMA_DUMMY_DESC;
+ 			tx_buf->flags |= MTK_TX_FLAGS_PAGE0;
+ 			tx_buf->flags |= (!mac->id) ? MTK_TX_FLAGS_FPORT0 :
+ 					 MTK_TX_FLAGS_FPORT1;
+@@ -1202,7 +1213,8 @@ static int mtk_tx_map(struct sk_buff *sk
+ 	}
+ 
+ 	/* store skb to cleanup */
+-	itx_buf->skb = skb;
++	itx_buf->type = MTK_TYPE_SKB;
++	itx_buf->data = skb;
+ 
+ 	if (!MTK_HAS_CAPS(soc->caps, MTK_QDMA)) {
+ 		if (k & 0x1)
+@@ -1414,13 +1426,14 @@ static struct page_pool *mtk_create_page
+ 		.pool_size = size,
+ 		.nid = NUMA_NO_NODE,
+ 		.dev = eth->dma_dev,
+-		.dma_dir = DMA_FROM_DEVICE,
+ 		.offset = MTK_PP_HEADROOM,
+ 		.max_len = MTK_PP_MAX_BUF_SIZE,
+ 	};
+ 	struct page_pool *pp;
+ 	int err;
+ 
++	pp_params.dma_dir = rcu_access_pointer(eth->prog) ? DMA_BIDIRECTIONAL
++							  : DMA_FROM_DEVICE;
+ 	pp = page_pool_create(&pp_params);
+ 	if (IS_ERR(pp))
+ 		return pp;
+@@ -1466,6 +1479,122 @@ static void mtk_rx_put_buff(struct mtk_r
+ 		skb_free_frag(data);
+ }
+ 
++static int mtk_xdp_submit_frame(struct mtk_eth *eth, struct xdp_frame *xdpf,
++				struct net_device *dev, bool dma_map)
++{
++	const struct mtk_soc_data *soc = eth->soc;
++	struct mtk_tx_ring *ring = &eth->tx_ring;
++	struct mtk_tx_dma_desc_info txd_info = {
++		.size	= xdpf->len,
++		.first	= true,
++		.last	= true,
++	};
++	struct mtk_mac *mac = netdev_priv(dev);
++	struct mtk_tx_dma *txd, *txd_pdma;
++	int err = 0, index = 0, n_desc = 1;
++	struct mtk_tx_buf *tx_buf;
++
++	if (unlikely(test_bit(MTK_RESETTING, &eth->state)))
++		return -EBUSY;
++
++	if (unlikely(atomic_read(&ring->free_count) <= 1))
++		return -EBUSY;
++
++	spin_lock(&eth->page_lock);
++
++	txd = ring->next_free;
++	if (txd == ring->last_free) {
++		err = -ENOMEM;
++		goto out;
++	}
++
++	tx_buf = mtk_desc_to_tx_buf(ring, txd, soc->txrx.txd_size);
++	memset(tx_buf, 0, sizeof(*tx_buf));
++
++	if (dma_map) {  /* ndo_xdp_xmit */
++		txd_info.addr = dma_map_single(eth->dma_dev, xdpf->data,
++					       txd_info.size, DMA_TO_DEVICE);
++		if (unlikely(dma_mapping_error(eth->dma_dev, txd_info.addr))) {
++			err = -ENOMEM;
++			goto out;
++		}
++		tx_buf->flags |= MTK_TX_FLAGS_SINGLE0;
++	} else {
++		struct page *page = virt_to_head_page(xdpf->data);
++
++		txd_info.addr = page_pool_get_dma_addr(page) +
++				sizeof(*xdpf) + xdpf->headroom;
++		dma_sync_single_for_device(eth->dma_dev, txd_info.addr,
++					   txd_info.size,
++					   DMA_BIDIRECTIONAL);
++	}
++	mtk_tx_set_dma_desc(dev, txd, &txd_info);
++
++	tx_buf->flags |= !mac->id ? MTK_TX_FLAGS_FPORT0 : MTK_TX_FLAGS_FPORT1;
++
++	txd_pdma = qdma_to_pdma(ring, txd);
++	setup_tx_buf(eth, tx_buf, txd_pdma, txd_info.addr, txd_info.size,
++		     index++);
++
++	/* store xdpf for cleanup */
++	tx_buf->type = dma_map ? MTK_TYPE_XDP_NDO : MTK_TYPE_XDP_TX;
++	tx_buf->data = xdpf;
++
++	if (!MTK_HAS_CAPS(soc->caps, MTK_QDMA)) {
++		if (index & 1)
++			txd_pdma->txd2 |= TX_DMA_LS0;
++		else
++			txd_pdma->txd2 |= TX_DMA_LS1;
++	}
++
++	ring->next_free = mtk_qdma_phys_to_virt(ring, txd->txd2);
++	atomic_sub(n_desc, &ring->free_count);
++
++	/* make sure that all changes to the dma ring are flushed before we
++	 * continue
++	 */
++	wmb();
++
++	if (MTK_HAS_CAPS(soc->caps, MTK_QDMA)) {
++		mtk_w32(eth, txd->txd2, soc->reg_map->qdma.ctx_ptr);
++	} else {
++		int idx;
++
++		idx = txd_to_idx(ring, txd, soc->txrx.txd_size);
++		mtk_w32(eth, NEXT_DESP_IDX(idx, ring->dma_size),
++			MT7628_TX_CTX_IDX0);
++	}
++out:
++	spin_unlock(&eth->page_lock);
++
++	return err;
++}
++
++static int mtk_xdp_xmit(struct net_device *dev, int num_frame,
++			struct xdp_frame **frames, u32 flags)
++{
++	struct mtk_mac *mac = netdev_priv(dev);
++	struct mtk_hw_stats *hw_stats = mac->hw_stats;
++	struct mtk_eth *eth = mac->hw;
++	int i, nxmit = 0;
++
++	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
++		return -EINVAL;
++
++	for (i = 0; i < num_frame; i++) {
++		if (mtk_xdp_submit_frame(eth, frames[i], dev, true))
++			break;
++		nxmit++;
++	}
++
++	u64_stats_update_begin(&hw_stats->syncp);
++	hw_stats->xdp_stats.tx_xdp_xmit += nxmit;
++	hw_stats->xdp_stats.tx_xdp_xmit_errors += num_frame - nxmit;
++	u64_stats_update_end(&hw_stats->syncp);
++
++	return nxmit;
++}
++
+ static u32 mtk_xdp_run(struct mtk_eth *eth, struct mtk_rx_ring *ring,
+ 		       struct xdp_buff *xdp, struct net_device *dev)
+ {
+@@ -1494,6 +1623,18 @@ static u32 mtk_xdp_run(struct mtk_eth *e
+ 
+ 		count = &hw_stats->xdp_stats.rx_xdp_redirect;
+ 		goto update_stats;
++	case XDP_TX: {
++		struct xdp_frame *xdpf = xdp_convert_buff_to_frame(xdp);
++
++		if (mtk_xdp_submit_frame(eth, xdpf, dev, false)) {
++			count = &hw_stats->xdp_stats.rx_xdp_tx_errors;
++			act = XDP_DROP;
++			break;
++		}
++
++		count = &hw_stats->xdp_stats.rx_xdp_tx;
++		goto update_stats;
++	}
+ 	default:
+ 		bpf_warn_invalid_xdp_action(act);
+ 		fallthrough;
+@@ -1727,9 +1868,8 @@ static int mtk_poll_tx_qdma(struct mtk_e
+ {
+ 	const struct mtk_reg_map *reg_map = eth->soc->reg_map;
+ 	struct mtk_tx_ring *ring = &eth->tx_ring;
+-	struct mtk_tx_dma *desc;
+-	struct sk_buff *skb;
+ 	struct mtk_tx_buf *tx_buf;
++	struct mtk_tx_dma *desc;
+ 	u32 cpu, dma;
+ 
+ 	cpu = ring->last_free_ptr;
+@@ -1750,15 +1890,21 @@ static int mtk_poll_tx_qdma(struct mtk_e
+ 		if (tx_buf->flags & MTK_TX_FLAGS_FPORT1)
+ 			mac = 1;
+ 
+-		skb = tx_buf->skb;
+-		if (!skb)
++		if (!tx_buf->data)
+ 			break;
+ 
+-		if (skb != (struct sk_buff *)MTK_DMA_DUMMY_DESC) {
++		if (tx_buf->type == MTK_TYPE_SKB &&
++		    tx_buf->data != (void *)MTK_DMA_DUMMY_DESC) {
++			struct sk_buff *skb = tx_buf->data;
++
+ 			bytes[mac] += skb->len;
+ 			done[mac]++;
+ 			budget--;
++		} else if (tx_buf->type == MTK_TYPE_XDP_TX ||
++			   tx_buf->type == MTK_TYPE_XDP_NDO) {
++			budget--;
+ 		}
++
+ 		mtk_tx_unmap(eth, tx_buf, true);
+ 
+ 		ring->last_free = desc;
+@@ -1777,9 +1923,8 @@ static int mtk_poll_tx_pdma(struct mtk_e
+ 			    unsigned int *done, unsigned int *bytes)
+ {
+ 	struct mtk_tx_ring *ring = &eth->tx_ring;
+-	struct mtk_tx_dma *desc;
+-	struct sk_buff *skb;
+ 	struct mtk_tx_buf *tx_buf;
++	struct mtk_tx_dma *desc;
+ 	u32 cpu, dma;
+ 
+ 	cpu = ring->cpu_idx;
+@@ -1787,14 +1932,18 @@ static int mtk_poll_tx_pdma(struct mtk_e
+ 
+ 	while ((cpu != dma) && budget) {
+ 		tx_buf = &ring->buf[cpu];
+-		skb = tx_buf->skb;
+-		if (!skb)
++		if (!tx_buf->data)
+ 			break;
+ 
+-		if (skb != (struct sk_buff *)MTK_DMA_DUMMY_DESC) {
++		if (tx_buf->type == MTK_TYPE_SKB &&
++		    tx_buf->data != (void *)MTK_DMA_DUMMY_DESC) {
++			struct sk_buff *skb = tx_buf->data;
+ 			bytes[0] += skb->len;
+ 			done[0]++;
+ 			budget--;
++		} else if (tx_buf->type == MTK_TYPE_XDP_TX ||
++			   tx_buf->type == MTK_TYPE_XDP_NDO) {
++			budget--;
+ 		}
+ 
+ 		mtk_tx_unmap(eth, tx_buf, true);
+@@ -3475,6 +3624,7 @@ static const struct net_device_ops mtk_n
+ #endif
+ 	.ndo_setup_tc		= mtk_eth_setup_tc,
+ 	.ndo_bpf		= mtk_xdp,
++	.ndo_xdp_xmit		= mtk_xdp_xmit,
+ };
+ 
+ static int mtk_add_mac(struct mtk_eth *eth, struct device_node *np)
+--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.h
++++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
+@@ -693,6 +693,12 @@ enum mtk_dev_state {
+ 	MTK_RESETTING
+ };
+ 
++enum mtk_tx_buf_type {
++	MTK_TYPE_SKB,
++	MTK_TYPE_XDP_TX,
++	MTK_TYPE_XDP_NDO,
++};
++
+ /* struct mtk_tx_buf -	This struct holds the pointers to the memory pointed at
+  *			by the TX descriptor	s
+  * @skb:		The SKB pointer of the packet being sent
+@@ -702,7 +708,9 @@ enum mtk_dev_state {
+  * @dma_len1:		The length of the second segment
+  */
+ struct mtk_tx_buf {
+-	struct sk_buff *skb;
++	enum mtk_tx_buf_type type;
++	void *data;
++
+ 	u32 flags;
+ 	DEFINE_DMA_UNMAP_ADDR(dma_addr0);
+ 	DEFINE_DMA_UNMAP_LEN(dma_len0);