[RESEND,PATCH net-next v7 0/8] Wangxun netdev features support

[RESEND,PATCH net-next v7 0/8] Wangxun netdev features support

[Mengyuan Lou]

Implement tx_csum and rx_csum to support hardware checksum offload.
Implement ndo_vlan_rx_add_vid and ndo_vlan_rx_kill_vid.
Implement ndo_set_features.
Enable macros in netdev features which wangxun can support.

changes v7:
- Remove fragmented packets parsing.
- Jakub Kicinski:
https://lore.kernel.org/netdev/20230523210454.12963d67@kernel.org/
changes v6:
- Fix some code spelling errors.
changes v5:
- Add ndo_set_features support.
- Move wx_decode_ptype() and wx_ptype_lookup to C file.
- Remove wx_fwd_adapter.
changes v4:
- Yunsheng Lin:
https://lore.kernel.org/netdev/c4b9765d-7213-2718-5de3-5e8231753b95@huawei.com/
changes v3:
- Yunsheng Lin: Tidy up logic for wx_encode_tx_desc_ptype.
changes v2:
- Andrew Lunn:
Add ETH_P_CNM Congestion Notification Message to if_ether.h.
Remove lro support.
- Yunsheng Lin:
https://lore.kernel.org/netdev/eb75ae23-8c19-bbc5-e99a-9b853511affa@huawei.com/

Mengyuan Lou (8):
  net: wangxun: libwx add tx offload functions
  net: wangxun: libwx add rx offload functions
  net: wangxun: Implement vlan add and kill functions
  net: libwx: Implement xx_set_features ops
  net: ngbe: Add netdev features support
  net: ngbe: Implement vlan add and remove ops
  net: txgbe: Add netdev features support
  net: txgbe: Implement vlan add and remove ops

 drivers/net/ethernet/wangxun/libwx/wx_hw.c    | 272 ++++++-
 drivers/net/ethernet/wangxun/libwx/wx_hw.h    |   3 +
 drivers/net/ethernet/wangxun/libwx/wx_lib.c   | 734 +++++++++++++++++-
 drivers/net/ethernet/wangxun/libwx/wx_lib.h   |   1 +
 drivers/net/ethernet/wangxun/libwx/wx_type.h  | 214 ++++-
 drivers/net/ethernet/wangxun/ngbe/ngbe_main.c |  20 +-
 drivers/net/ethernet/wangxun/ngbe/ngbe_type.h |   1 +
 .../net/ethernet/wangxun/txgbe/txgbe_main.c   |  25 +-
 .../net/ethernet/wangxun/txgbe/txgbe_type.h   |   1 +
 9 files changed, 1250 insertions(+), 21 deletions(-)

-- 
2.40.1

[RESEND,PATCH net-next v7 1/8] net: wangxun: libwx add tx offload functions

[Mengyuan Lou]

Add tx offload functions for wx_xmit_frame_ring which
includes wx_encode_tx_desc_ptype, wx_tso and wx_tx_csum.
which supports ngbe and txgbe to implement tx offload
function.

Signed-off-by: Mengyuan Lou <mengyuanlou@net-swift.com>
---
 drivers/net/ethernet/wangxun/libwx/wx_lib.c  | 462 ++++++++++++++++++-
 drivers/net/ethernet/wangxun/libwx/wx_type.h |  87 +++-
 2 files changed, 541 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/wangxun/libwx/wx_lib.c b/drivers/net/ethernet/wangxun/libwx/wx_lib.c
index 1e8d8b7b0c62..34ac30e87b7c 100644
--- a/drivers/net/ethernet/wangxun/libwx/wx_lib.c
+++ b/drivers/net/ethernet/wangxun/libwx/wx_lib.c
@@ -2,9 +2,14 @@
 /* Copyright (c) 2019 - 2022 Beijing WangXun Technology Co., Ltd. */
 
 #include <linux/etherdevice.h>
+#include <net/ip6_checksum.h>
 #include <net/page_pool.h>
+#include <net/inet_ecn.h>
 #include <linux/iopoll.h>
+#include <linux/sctp.h>
 #include <linux/pci.h>
+#include <net/tcp.h>
+#include <net/ip.h>
 
 #include "wx_type.h"
 #include "wx_lib.h"
@@ -707,11 +712,50 @@ static int wx_maybe_stop_tx(struct wx_ring *tx_ring, u16 size)
 	return 0;
 }
 
+static u32 wx_tx_cmd_type(u32 tx_flags)
+{
+	/* set type for advanced descriptor with frame checksum insertion */
+	u32 cmd_type = WX_TXD_DTYP_DATA | WX_TXD_IFCS;
+
+	/* set HW vlan bit if vlan is present */
+	cmd_type |= WX_SET_FLAG(tx_flags, WX_TX_FLAGS_HW_VLAN, WX_TXD_VLE);
+	/* set segmentation enable bits for TSO/FSO */
+	cmd_type |= WX_SET_FLAG(tx_flags, WX_TX_FLAGS_TSO, WX_TXD_TSE);
+	/* set timestamp bit if present */
+	cmd_type |= WX_SET_FLAG(tx_flags, WX_TX_FLAGS_TSTAMP, WX_TXD_MAC_TSTAMP);
+	cmd_type |= WX_SET_FLAG(tx_flags, WX_TX_FLAGS_LINKSEC, WX_TXD_LINKSEC);
+
+	return cmd_type;
+}
+
+static void wx_tx_olinfo_status(union wx_tx_desc *tx_desc,
+				u32 tx_flags, unsigned int paylen)
+{
+	u32 olinfo_status = paylen << WX_TXD_PAYLEN_SHIFT;
+
+	/* enable L4 checksum for TSO and TX checksum offload */
+	olinfo_status |= WX_SET_FLAG(tx_flags, WX_TX_FLAGS_CSUM, WX_TXD_L4CS);
+	/* enable IPv4 checksum for TSO */
+	olinfo_status |= WX_SET_FLAG(tx_flags, WX_TX_FLAGS_IPV4, WX_TXD_IIPCS);
+	/* enable outer IPv4 checksum for TSO */
+	olinfo_status |= WX_SET_FLAG(tx_flags, WX_TX_FLAGS_OUTER_IPV4,
+				     WX_TXD_EIPCS);
+	/* Check Context must be set if Tx switch is enabled, which it
+	 * always is for case where virtual functions are running
+	 */
+	olinfo_status |= WX_SET_FLAG(tx_flags, WX_TX_FLAGS_CC, WX_TXD_CC);
+	olinfo_status |= WX_SET_FLAG(tx_flags, WX_TX_FLAGS_IPSEC,
+				     WX_TXD_IPSEC);
+	tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
+}
+
 static void wx_tx_map(struct wx_ring *tx_ring,
-		      struct wx_tx_buffer *first)
+		      struct wx_tx_buffer *first,
+		      const u8 hdr_len)
 {
 	struct sk_buff *skb = first->skb;
 	struct wx_tx_buffer *tx_buffer;
+	u32 tx_flags = first->tx_flags;
 	u16 i = tx_ring->next_to_use;
 	unsigned int data_len, size;
 	union wx_tx_desc *tx_desc;
@@ -719,10 +763,9 @@ static void wx_tx_map(struct wx_ring *tx_ring,
 	dma_addr_t dma;
 	u32 cmd_type;
 
-	cmd_type = WX_TXD_DTYP_DATA | WX_TXD_IFCS;
+	cmd_type = wx_tx_cmd_type(tx_flags);
 	tx_desc = WX_TX_DESC(tx_ring, i);
-
-	tx_desc->read.olinfo_status = cpu_to_le32(skb->len << WX_TXD_PAYLEN_SHIFT);
+	wx_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
 
 	size = skb_headlen(skb);
 	data_len = skb->data_len;
@@ -838,12 +881,399 @@ static void wx_tx_map(struct wx_ring *tx_ring,
 	tx_ring->next_to_use = i;
 }
 
+static void wx_tx_ctxtdesc(struct wx_ring *tx_ring, u32 vlan_macip_lens,
+			   u32 fcoe_sof_eof, u32 type_tucmd, u32 mss_l4len_idx)
+{
+	struct wx_tx_context_desc *context_desc;
+	u16 i = tx_ring->next_to_use;
+
+	context_desc = WX_TX_CTXTDESC(tx_ring, i);
+	i++;
+	tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
+
+	/* set bits to identify this as an advanced context descriptor */
+	type_tucmd |= WX_TXD_DTYP_CTXT;
+	context_desc->vlan_macip_lens   = cpu_to_le32(vlan_macip_lens);
+	context_desc->seqnum_seed       = cpu_to_le32(fcoe_sof_eof);
+	context_desc->type_tucmd_mlhl   = cpu_to_le32(type_tucmd);
+	context_desc->mss_l4len_idx     = cpu_to_le32(mss_l4len_idx);
+}
+
+static void wx_get_ipv6_proto(struct sk_buff *skb, int offset, u8 *nexthdr)
+{
+	struct ipv6hdr *hdr = (struct ipv6hdr *)(skb->data + offset);
+
+	*nexthdr = hdr->nexthdr;
+	offset += sizeof(struct ipv6hdr);
+	while (ipv6_ext_hdr(*nexthdr)) {
+		struct ipv6_opt_hdr _hdr, *hp;
+
+		if (*nexthdr == NEXTHDR_NONE)
+			return;
+		hp = skb_header_pointer(skb, offset, sizeof(_hdr), &_hdr);
+		if (!hp)
+			return;
+		if (*nexthdr == NEXTHDR_FRAGMENT)
+			break;
+		*nexthdr = hp->nexthdr;
+	}
+}
+
+union network_header {
+	struct iphdr *ipv4;
+	struct ipv6hdr *ipv6;
+	void *raw;
+};
+
+static u8 wx_encode_tx_desc_ptype(const struct wx_tx_buffer *first)
+{
+	u8 tun_prot = 0, l4_prot = 0, ptype = 0;
+	struct sk_buff *skb = first->skb;
+
+	if (skb->encapsulation) {
+		union network_header hdr;
+
+		switch (first->protocol) {
+		case htons(ETH_P_IP):
+			tun_prot = ip_hdr(skb)->protocol;
+			ptype = WX_PTYPE_TUN_IPV4;
+			break;
+		case htons(ETH_P_IPV6):
+			wx_get_ipv6_proto(skb, skb_network_offset(skb), &tun_prot);
+			ptype = WX_PTYPE_TUN_IPV6;
+			break;
+		default:
+			return ptype;
+		}
+
+		if (tun_prot == IPPROTO_IPIP) {
+			hdr.raw = (void *)inner_ip_hdr(skb);
+			ptype |= WX_PTYPE_PKT_IPIP;
+		} else if (tun_prot == IPPROTO_UDP) {
+			hdr.raw = (void *)inner_ip_hdr(skb);
+			if (skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
+			    skb->inner_protocol != htons(ETH_P_TEB)) {
+				ptype |= WX_PTYPE_PKT_IG;
+			} else {
+				if (((struct ethhdr *)skb_inner_mac_header(skb))->h_proto
+				     == htons(ETH_P_8021Q))
+					ptype |= WX_PTYPE_PKT_IGMV;
+				else
+					ptype |= WX_PTYPE_PKT_IGM;
+			}
+
+		} else if (tun_prot == IPPROTO_GRE) {
+			hdr.raw = (void *)inner_ip_hdr(skb);
+			if (skb->inner_protocol ==  htons(ETH_P_IP) ||
+			    skb->inner_protocol ==  htons(ETH_P_IPV6)) {
+				ptype |= WX_PTYPE_PKT_IG;
+			} else {
+				if (((struct ethhdr *)skb_inner_mac_header(skb))->h_proto
+				    == htons(ETH_P_8021Q))
+					ptype |= WX_PTYPE_PKT_IGMV;
+				else
+					ptype |= WX_PTYPE_PKT_IGM;
+			}
+		} else {
+			return ptype;
+		}
+
+		switch (hdr.ipv4->version) {
+		case IPVERSION:
+			l4_prot = hdr.ipv4->protocol;
+			break;
+		case 6:
+			wx_get_ipv6_proto(skb, skb_inner_network_offset(skb), &l4_prot);
+			ptype |= WX_PTYPE_PKT_IPV6;
+			break;
+		default:
+			return ptype;
+		}
+	} else {
+		switch (first->protocol) {
+		case htons(ETH_P_IP):
+			l4_prot = ip_hdr(skb)->protocol;
+			ptype = WX_PTYPE_PKT_IP;
+			break;
+		case htons(ETH_P_IPV6):
+			wx_get_ipv6_proto(skb, skb_network_offset(skb), &l4_prot);
+			ptype = WX_PTYPE_PKT_IP | WX_PTYPE_PKT_IPV6;
+			break;
+		default:
+			return WX_PTYPE_PKT_MAC | WX_PTYPE_TYP_MAC;
+		}
+	}
+	switch (l4_prot) {
+	case IPPROTO_TCP:
+		ptype |= WX_PTYPE_TYP_TCP;
+		break;
+	case IPPROTO_UDP:
+		ptype |= WX_PTYPE_TYP_UDP;
+		break;
+	case IPPROTO_SCTP:
+		ptype |= WX_PTYPE_TYP_SCTP;
+		break;
+	default:
+		ptype |= WX_PTYPE_TYP_IP;
+		break;
+	}
+
+	return ptype;
+}
+
+static int wx_tso(struct wx_ring *tx_ring, struct wx_tx_buffer *first,
+		  u8 *hdr_len, u8 ptype)
+{
+	u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
+	struct net_device *netdev = tx_ring->netdev;
+	u32 l4len, tunhdr_eiplen_tunlen = 0;
+	struct sk_buff *skb = first->skb;
+	bool enc = skb->encapsulation;
+	struct ipv6hdr *ipv6h;
+	struct tcphdr *tcph;
+	struct iphdr *iph;
+	u8 tun_prot = 0;
+	int err;
+
+	if (skb->ip_summed != CHECKSUM_PARTIAL)
+		return 0;
+
+	if (!skb_is_gso(skb))
+		return 0;
+
+	err = skb_cow_head(skb, 0);
+	if (err < 0)
+		return err;
+
+	/* indicates the inner headers in the skbuff are valid. */
+	iph = enc ? inner_ip_hdr(skb) : ip_hdr(skb);
+	if (iph->version == 4) {
+		tcph = enc ? inner_tcp_hdr(skb) : tcp_hdr(skb);
+		iph->tot_len = 0;
+		iph->check = 0;
+		tcph->check = ~csum_tcpudp_magic(iph->saddr,
+						 iph->daddr, 0,
+						 IPPROTO_TCP, 0);
+		first->tx_flags |= WX_TX_FLAGS_TSO |
+				   WX_TX_FLAGS_CSUM |
+				   WX_TX_FLAGS_IPV4 |
+				   WX_TX_FLAGS_CC;
+	} else if (iph->version == 6 && skb_is_gso_v6(skb)) {
+		ipv6h = enc ? inner_ipv6_hdr(skb) : ipv6_hdr(skb);
+		tcph = enc ? inner_tcp_hdr(skb) : tcp_hdr(skb);
+		ipv6h->payload_len = 0;
+		tcph->check = ~csum_ipv6_magic(&ipv6h->saddr,
+					       &ipv6h->daddr, 0,
+					       IPPROTO_TCP, 0);
+		first->tx_flags |= WX_TX_FLAGS_TSO |
+				   WX_TX_FLAGS_CSUM |
+				   WX_TX_FLAGS_CC;
+	}
+
+	/* compute header lengths */
+	l4len = enc ? inner_tcp_hdrlen(skb) : tcp_hdrlen(skb);
+	*hdr_len = enc ? (skb_inner_transport_header(skb) - skb->data) :
+			 skb_transport_offset(skb);
+	*hdr_len += l4len;
+
+	/* update gso size and bytecount with header size */
+	first->gso_segs = skb_shinfo(skb)->gso_segs;
+	first->bytecount += (first->gso_segs - 1) * *hdr_len;
+
+	/* mss_l4len_id: use 0 as index for TSO */
+	mss_l4len_idx = l4len << WX_TXD_L4LEN_SHIFT;
+	mss_l4len_idx |= skb_shinfo(skb)->gso_size << WX_TXD_MSS_SHIFT;
+
+	/* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
+	if (enc) {
+		switch (first->protocol) {
+		case htons(ETH_P_IP):
+			tun_prot = ip_hdr(skb)->protocol;
+			first->tx_flags |= WX_TX_FLAGS_OUTER_IPV4;
+			break;
+		case htons(ETH_P_IPV6):
+			tun_prot = ipv6_hdr(skb)->nexthdr;
+			break;
+		default:
+			break;
+		}
+		switch (tun_prot) {
+		case IPPROTO_UDP:
+			tunhdr_eiplen_tunlen = WX_TXD_TUNNEL_UDP;
+			tunhdr_eiplen_tunlen |= ((skb_network_header_len(skb) >> 2) <<
+						 WX_TXD_OUTER_IPLEN_SHIFT) |
+						(((skb_inner_mac_header(skb) -
+						skb_transport_header(skb)) >> 1) <<
+						WX_TXD_TUNNEL_LEN_SHIFT);
+			break;
+		case IPPROTO_GRE:
+			tunhdr_eiplen_tunlen = WX_TXD_TUNNEL_GRE;
+			tunhdr_eiplen_tunlen |= ((skb_network_header_len(skb) >> 2) <<
+						 WX_TXD_OUTER_IPLEN_SHIFT) |
+						(((skb_inner_mac_header(skb) -
+						skb_transport_header(skb)) >> 1) <<
+						WX_TXD_TUNNEL_LEN_SHIFT);
+			break;
+		case IPPROTO_IPIP:
+			tunhdr_eiplen_tunlen = (((char *)inner_ip_hdr(skb) -
+						(char *)ip_hdr(skb)) >> 2) <<
+						WX_TXD_OUTER_IPLEN_SHIFT;
+			break;
+		default:
+			break;
+		}
+		vlan_macip_lens = skb_inner_network_header_len(skb) >> 1;
+	} else {
+		vlan_macip_lens = skb_network_header_len(skb) >> 1;
+	}
+
+	vlan_macip_lens |= skb_network_offset(skb) << WX_TXD_MACLEN_SHIFT;
+	vlan_macip_lens |= first->tx_flags & WX_TX_FLAGS_VLAN_MASK;
+
+	type_tucmd = ptype << 24;
+	if (skb->vlan_proto == htons(ETH_P_8021AD) &&
+	    netdev->features & NETIF_F_HW_VLAN_STAG_TX)
+		type_tucmd |= WX_SET_FLAG(first->tx_flags,
+					  WX_TX_FLAGS_HW_VLAN,
+					  0x1 << WX_TXD_TAG_TPID_SEL_SHIFT);
+	wx_tx_ctxtdesc(tx_ring, vlan_macip_lens, tunhdr_eiplen_tunlen,
+		       type_tucmd, mss_l4len_idx);
+
+	return 1;
+}
+
+static void wx_tx_csum(struct wx_ring *tx_ring, struct wx_tx_buffer *first,
+		       u8 ptype)
+{
+	u32 tunhdr_eiplen_tunlen = 0, vlan_macip_lens = 0;
+	struct net_device *netdev = tx_ring->netdev;
+	u32 mss_l4len_idx = 0, type_tucmd;
+	struct sk_buff *skb = first->skb;
+	u8 tun_prot = 0;
+
+	if (skb->ip_summed != CHECKSUM_PARTIAL) {
+		if (!(first->tx_flags & WX_TX_FLAGS_HW_VLAN) &&
+		    !(first->tx_flags & WX_TX_FLAGS_CC))
+			return;
+		vlan_macip_lens = skb_network_offset(skb) <<
+				  WX_TXD_MACLEN_SHIFT;
+	} else {
+		u8 l4_prot = 0;
+		union {
+			struct iphdr *ipv4;
+			struct ipv6hdr *ipv6;
+			u8 *raw;
+		} network_hdr;
+		union {
+			struct tcphdr *tcphdr;
+			u8 *raw;
+		} transport_hdr;
+
+		if (skb->encapsulation) {
+			network_hdr.raw = skb_inner_network_header(skb);
+			transport_hdr.raw = skb_inner_transport_header(skb);
+			vlan_macip_lens = skb_network_offset(skb) <<
+					  WX_TXD_MACLEN_SHIFT;
+			switch (first->protocol) {
+			case htons(ETH_P_IP):
+				tun_prot = ip_hdr(skb)->protocol;
+				break;
+			case htons(ETH_P_IPV6):
+				tun_prot = ipv6_hdr(skb)->nexthdr;
+				break;
+			default:
+				return;
+			}
+			switch (tun_prot) {
+			case IPPROTO_UDP:
+				tunhdr_eiplen_tunlen = WX_TXD_TUNNEL_UDP;
+				tunhdr_eiplen_tunlen |=
+					((skb_network_header_len(skb) >> 2) <<
+					WX_TXD_OUTER_IPLEN_SHIFT) |
+					(((skb_inner_mac_header(skb) -
+					skb_transport_header(skb)) >> 1) <<
+					WX_TXD_TUNNEL_LEN_SHIFT);
+				break;
+			case IPPROTO_GRE:
+				tunhdr_eiplen_tunlen = WX_TXD_TUNNEL_GRE;
+				tunhdr_eiplen_tunlen |= ((skb_network_header_len(skb) >> 2) <<
+							 WX_TXD_OUTER_IPLEN_SHIFT) |
+							 (((skb_inner_mac_header(skb) -
+							    skb_transport_header(skb)) >> 1) <<
+							  WX_TXD_TUNNEL_LEN_SHIFT);
+				break;
+			case IPPROTO_IPIP:
+				tunhdr_eiplen_tunlen = (((char *)inner_ip_hdr(skb) -
+							(char *)ip_hdr(skb)) >> 2) <<
+							WX_TXD_OUTER_IPLEN_SHIFT;
+				break;
+			default:
+				break;
+			}
+
+		} else {
+			network_hdr.raw = skb_network_header(skb);
+			transport_hdr.raw = skb_transport_header(skb);
+			vlan_macip_lens = skb_network_offset(skb) <<
+					  WX_TXD_MACLEN_SHIFT;
+		}
+
+		switch (network_hdr.ipv4->version) {
+		case IPVERSION:
+			vlan_macip_lens |= (transport_hdr.raw - network_hdr.raw) >> 1;
+			l4_prot = network_hdr.ipv4->protocol;
+			break;
+		case 6:
+			vlan_macip_lens |= (transport_hdr.raw - network_hdr.raw) >> 1;
+			l4_prot = network_hdr.ipv6->nexthdr;
+			break;
+		default:
+			break;
+		}
+
+		switch (l4_prot) {
+		case IPPROTO_TCP:
+		mss_l4len_idx = (transport_hdr.tcphdr->doff * 4) <<
+				WX_TXD_L4LEN_SHIFT;
+			break;
+		case IPPROTO_SCTP:
+			mss_l4len_idx = sizeof(struct sctphdr) <<
+					WX_TXD_L4LEN_SHIFT;
+			break;
+		case IPPROTO_UDP:
+			mss_l4len_idx = sizeof(struct udphdr) <<
+					WX_TXD_L4LEN_SHIFT;
+			break;
+		default:
+			break;
+		}
+
+		/* update TX checksum flag */
+		first->tx_flags |= WX_TX_FLAGS_CSUM;
+	}
+	first->tx_flags |= WX_TX_FLAGS_CC;
+	/* vlan_macip_lens: MACLEN, VLAN tag */
+	vlan_macip_lens |= first->tx_flags & WX_TX_FLAGS_VLAN_MASK;
+
+	type_tucmd = ptype << 24;
+	if (skb->vlan_proto == htons(ETH_P_8021AD) &&
+	    netdev->features & NETIF_F_HW_VLAN_STAG_TX)
+		type_tucmd |= WX_SET_FLAG(first->tx_flags,
+					  WX_TX_FLAGS_HW_VLAN,
+					  0x1 << WX_TXD_TAG_TPID_SEL_SHIFT);
+	wx_tx_ctxtdesc(tx_ring, vlan_macip_lens, tunhdr_eiplen_tunlen,
+		       type_tucmd, mss_l4len_idx);
+}
+
 static netdev_tx_t wx_xmit_frame_ring(struct sk_buff *skb,
 				      struct wx_ring *tx_ring)
 {
 	u16 count = TXD_USE_COUNT(skb_headlen(skb));
 	struct wx_tx_buffer *first;
+	u8 hdr_len = 0, ptype;
 	unsigned short f;
+	u32 tx_flags = 0;
+	int tso;
 
 	/* need: 1 descriptor per page * PAGE_SIZE/WX_MAX_DATA_PER_TXD,
 	 *       + 1 desc for skb_headlen/WX_MAX_DATA_PER_TXD,
@@ -864,7 +1294,29 @@ static netdev_tx_t wx_xmit_frame_ring(struct sk_buff *skb,
 	first->bytecount = skb->len;
 	first->gso_segs = 1;
 
-	wx_tx_map(tx_ring, first);
+	/* if we have a HW VLAN tag being added default to the HW one */
+	if (skb_vlan_tag_present(skb)) {
+		tx_flags |= skb_vlan_tag_get(skb) << WX_TX_FLAGS_VLAN_SHIFT;
+		tx_flags |= WX_TX_FLAGS_HW_VLAN;
+	}
+
+	/* record initial flags and protocol */
+	first->tx_flags = tx_flags;
+	first->protocol = vlan_get_protocol(skb);
+
+	ptype = wx_encode_tx_desc_ptype(first);
+
+	tso = wx_tso(tx_ring, first, &hdr_len, ptype);
+	if (tso < 0)
+		goto out_drop;
+	else if (!tso)
+		wx_tx_csum(tx_ring, first, ptype);
+	wx_tx_map(tx_ring, first, hdr_len);
+
+	return NETDEV_TX_OK;
+out_drop:
+	dev_kfree_skb_any(first->skb);
+	first->skb = NULL;
 
 	return NETDEV_TX_OK;
 }
diff --git a/drivers/net/ethernet/wangxun/libwx/wx_type.h b/drivers/net/ethernet/wangxun/libwx/wx_type.h
index 32f952d93009..d2e0584e3b44 100644
--- a/drivers/net/ethernet/wangxun/libwx/wx_type.h
+++ b/drivers/net/ethernet/wangxun/libwx/wx_type.h
@@ -6,6 +6,7 @@
 
 #include <linux/bitfield.h>
 #include <linux/netdevice.h>
+#include <net/ip.h>
 
 #define WX_NCSI_SUP                             0x8000
 #define WX_NCSI_MASK                            0x8000
@@ -315,9 +316,6 @@
 #define TXD_USE_COUNT(S)     DIV_ROUND_UP((S), WX_MAX_DATA_PER_TXD)
 #define DESC_NEEDED          (MAX_SKB_FRAGS + 4)
 
-/* Ether Types */
-#define WX_ETH_P_CNM                 0x22E7
-
 #define WX_CFG_PORT_ST               0x14404
 
 /******************* Receive Descriptor bit definitions **********************/
@@ -326,6 +324,29 @@
 
 #define WX_RXD_ERR_RXE               BIT(29) /* Any MAC Error */
 
+/* TUN */
+#define WX_PTYPE_TUN_IPV4            0x80
+#define WX_PTYPE_TUN_IPV6            0xC0
+
+/* PKT for TUN */
+#define WX_PTYPE_PKT_IPIP            0x00 /* IP+IP */
+#define WX_PTYPE_PKT_IG              0x10 /* IP+GRE */
+#define WX_PTYPE_PKT_IGM             0x20 /* IP+GRE+MAC */
+#define WX_PTYPE_PKT_IGMV            0x30 /* IP+GRE+MAC+VLAN */
+/* PKT for !TUN */
+#define WX_PTYPE_PKT_MAC             0x10
+#define WX_PTYPE_PKT_IP              0x20
+
+/* TYP for PKT=mac */
+#define WX_PTYPE_TYP_MAC             0x01
+/* TYP for PKT=ip */
+#define WX_PTYPE_PKT_IPV6            0x08
+#define WX_PTYPE_TYP_IPFRAG          0x01
+#define WX_PTYPE_TYP_IP              0x02
+#define WX_PTYPE_TYP_UDP             0x03
+#define WX_PTYPE_TYP_TCP             0x04
+#define WX_PTYPE_TYP_SCTP            0x05
+
 /*********************** Transmit Descriptor Config Masks ****************/
 #define WX_TXD_STAT_DD               BIT(0)  /* Descriptor Done */
 #define WX_TXD_DTYP_DATA             0       /* Adv Data Descriptor */
@@ -334,6 +355,49 @@
 #define WX_TXD_IFCS                  BIT(25) /* Insert FCS */
 #define WX_TXD_RS                    BIT(27) /* Report Status */
 
+/*********************** Adv Transmit Descriptor Config Masks ****************/
+#define WX_TXD_MAC_TSTAMP            BIT(19) /* IEEE1588 time stamp */
+#define WX_TXD_DTYP_CTXT             BIT(20) /* Adv Context Desc */
+#define WX_TXD_LINKSEC               BIT(26) /* enable linksec */
+#define WX_TXD_VLE                   BIT(30) /* VLAN pkt enable */
+#define WX_TXD_TSE                   BIT(31) /* TCP Seg enable */
+#define WX_TXD_CC                    BIT(7) /* Check Context */
+#define WX_TXD_IPSEC                 BIT(8) /* enable ipsec esp */
+#define WX_TXD_L4CS                  BIT(9)
+#define WX_TXD_IIPCS                 BIT(10)
+#define WX_TXD_EIPCS                 BIT(11)
+#define WX_TXD_PAYLEN_SHIFT          13 /* Adv desc PAYLEN shift */
+#define WX_TXD_MACLEN_SHIFT          9  /* Adv ctxt desc mac len shift */
+#define WX_TXD_TAG_TPID_SEL_SHIFT    11
+
+#define WX_TXD_L4LEN_SHIFT           8  /* Adv ctxt L4LEN shift */
+#define WX_TXD_MSS_SHIFT             16  /* Adv ctxt MSS shift */
+
+#define WX_TXD_OUTER_IPLEN_SHIFT     12 /* Adv ctxt OUTERIPLEN shift */
+#define WX_TXD_TUNNEL_LEN_SHIFT      21 /* Adv ctxt TUNNELLEN shift */
+#define WX_TXD_TUNNEL_TYPE_SHIFT     11 /* Adv Tx Desc Tunnel Type shift */
+#define WX_TXD_TUNNEL_UDP            FIELD_PREP(BIT(WX_TXD_TUNNEL_TYPE_SHIFT), 0)
+#define WX_TXD_TUNNEL_GRE            FIELD_PREP(BIT(WX_TXD_TUNNEL_TYPE_SHIFT), 1)
+
+enum wx_tx_flags {
+	/* cmd_type flags */
+	WX_TX_FLAGS_HW_VLAN	= 0x01,
+	WX_TX_FLAGS_TSO		= 0x02,
+	WX_TX_FLAGS_TSTAMP	= 0x04,
+
+	/* olinfo flags */
+	WX_TX_FLAGS_CC		= 0x08,
+	WX_TX_FLAGS_IPV4	= 0x10,
+	WX_TX_FLAGS_CSUM	= 0x20,
+	WX_TX_FLAGS_OUTER_IPV4	= 0x100,
+	WX_TX_FLAGS_LINKSEC	= 0x200,
+	WX_TX_FLAGS_IPSEC	= 0x400,
+};
+
+/* VLAN info */
+#define WX_TX_FLAGS_VLAN_MASK			GENMASK(31, 16)
+#define WX_TX_FLAGS_VLAN_SHIFT			16
+
 /* Host Interface Command Structures */
 struct wx_hic_hdr {
 	u8 cmd;
@@ -508,10 +572,25 @@ union wx_rx_desc {
 	} wb;  /* writeback */
 };
 
+struct wx_tx_context_desc {
+	__le32 vlan_macip_lens;
+	__le32 seqnum_seed;
+	__le32 type_tucmd_mlhl;
+	__le32 mss_l4len_idx;
+};
+
+/* if _flag is in _input, return _result */
+#define WX_SET_FLAG(_input, _flag, _result) \
+	(((_flag) <= (_result)) ? \
+	 ((u32)((_input) & (_flag)) * ((_result) / (_flag))) : \
+	 ((u32)((_input) & (_flag)) / ((_flag) / (_result))))
+
 #define WX_RX_DESC(R, i)     \
 	(&(((union wx_rx_desc *)((R)->desc))[i]))
 #define WX_TX_DESC(R, i)     \
 	(&(((union wx_tx_desc *)((R)->desc))[i]))
+#define WX_TX_CTXTDESC(R, i) \
+	(&(((struct wx_tx_context_desc *)((R)->desc))[i]))
 
 /* wrapper around a pointer to a socket buffer,
  * so a DMA handle can be stored along with the buffer
@@ -523,6 +602,8 @@ struct wx_tx_buffer {
 	unsigned short gso_segs;
 	DEFINE_DMA_UNMAP_ADDR(dma);
 	DEFINE_DMA_UNMAP_LEN(len);
+	__be16 protocol;
+	u32 tx_flags;
 };
 
 struct wx_rx_buffer {
-- 
2.40.1

[RESEND,PATCH net-next v7 2/8] net: wangxun: libwx add rx offload functions

[Mengyuan Lou]

Add rx offload functions for wx_clean_rx_irq
which supports ngbe and txgbe to implement
rx offload function.

Signed-off-by: Mengyuan Lou <mengyuanlou@net-swift.com>
---
 drivers/net/ethernet/wangxun/libwx/wx_lib.c  | 234 ++++++++++++++++++-
 drivers/net/ethernet/wangxun/libwx/wx_type.h | 100 +++++++-
 2 files changed, 331 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/wangxun/libwx/wx_lib.c b/drivers/net/ethernet/wangxun/libwx/wx_lib.c
index 34ac30e87b7c..84107208401e 100644
--- a/drivers/net/ethernet/wangxun/libwx/wx_lib.c
+++ b/drivers/net/ethernet/wangxun/libwx/wx_lib.c
@@ -15,6 +15,144 @@
 #include "wx_lib.h"
 #include "wx_hw.h"
 
+/* Lookup table mapping the HW PTYPE to the bit field for decoding */
+static struct wx_dec_ptype wx_ptype_lookup[256] = {
+	/* L2: mac */
+	[0x11] = WX_PTT(L2, NONE, NONE, NONE, NONE, PAY2),
+	[0x12] = WX_PTT(L2, NONE, NONE, NONE, TS,   PAY2),
+	[0x13] = WX_PTT(L2, NONE, NONE, NONE, NONE, PAY2),
+	[0x14] = WX_PTT(L2, NONE, NONE, NONE, NONE, PAY2),
+	[0x15] = WX_PTT(L2, NONE, NONE, NONE, NONE, NONE),
+	[0x16] = WX_PTT(L2, NONE, NONE, NONE, NONE, PAY2),
+	[0x17] = WX_PTT(L2, NONE, NONE, NONE, NONE, NONE),
+
+	/* L2: ethertype filter */
+	[0x18 ... 0x1F] = WX_PTT(L2, NONE, NONE, NONE, NONE, NONE),
+
+	/* L3: ip non-tunnel */
+	[0x21] = WX_PTT(IP, FGV4, NONE, NONE, NONE, PAY3),
+	[0x22] = WX_PTT(IP, IPV4, NONE, NONE, NONE, PAY3),
+	[0x23] = WX_PTT(IP, IPV4, NONE, NONE, UDP,  PAY4),
+	[0x24] = WX_PTT(IP, IPV4, NONE, NONE, TCP,  PAY4),
+	[0x25] = WX_PTT(IP, IPV4, NONE, NONE, SCTP, PAY4),
+	[0x29] = WX_PTT(IP, FGV6, NONE, NONE, NONE, PAY3),
+	[0x2A] = WX_PTT(IP, IPV6, NONE, NONE, NONE, PAY3),
+	[0x2B] = WX_PTT(IP, IPV6, NONE, NONE, UDP,  PAY3),
+	[0x2C] = WX_PTT(IP, IPV6, NONE, NONE, TCP,  PAY4),
+	[0x2D] = WX_PTT(IP, IPV6, NONE, NONE, SCTP, PAY4),
+
+	/* L2: fcoe */
+	[0x30 ... 0x34] = WX_PTT(FCOE, NONE, NONE, NONE, NONE, PAY3),
+	[0x38 ... 0x3C] = WX_PTT(FCOE, NONE, NONE, NONE, NONE, PAY3),
+
+	/* IPv4 --> IPv4/IPv6 */
+	[0x81] = WX_PTT(IP, IPV4, IPIP, FGV4, NONE, PAY3),
+	[0x82] = WX_PTT(IP, IPV4, IPIP, IPV4, NONE, PAY3),
+	[0x83] = WX_PTT(IP, IPV4, IPIP, IPV4, UDP,  PAY4),
+	[0x84] = WX_PTT(IP, IPV4, IPIP, IPV4, TCP,  PAY4),
+	[0x85] = WX_PTT(IP, IPV4, IPIP, IPV4, SCTP, PAY4),
+	[0x89] = WX_PTT(IP, IPV4, IPIP, FGV6, NONE, PAY3),
+	[0x8A] = WX_PTT(IP, IPV4, IPIP, IPV6, NONE, PAY3),
+	[0x8B] = WX_PTT(IP, IPV4, IPIP, IPV6, UDP,  PAY4),
+	[0x8C] = WX_PTT(IP, IPV4, IPIP, IPV6, TCP,  PAY4),
+	[0x8D] = WX_PTT(IP, IPV4, IPIP, IPV6, SCTP, PAY4),
+
+	/* IPv4 --> GRE/NAT --> NONE/IPv4/IPv6 */
+	[0x90] = WX_PTT(IP, IPV4, IG, NONE, NONE, PAY3),
+	[0x91] = WX_PTT(IP, IPV4, IG, FGV4, NONE, PAY3),
+	[0x92] = WX_PTT(IP, IPV4, IG, IPV4, NONE, PAY3),
+	[0x93] = WX_PTT(IP, IPV4, IG, IPV4, UDP,  PAY4),
+	[0x94] = WX_PTT(IP, IPV4, IG, IPV4, TCP,  PAY4),
+	[0x95] = WX_PTT(IP, IPV4, IG, IPV4, SCTP, PAY4),
+	[0x99] = WX_PTT(IP, IPV4, IG, FGV6, NONE, PAY3),
+	[0x9A] = WX_PTT(IP, IPV4, IG, IPV6, NONE, PAY3),
+	[0x9B] = WX_PTT(IP, IPV4, IG, IPV6, UDP,  PAY4),
+	[0x9C] = WX_PTT(IP, IPV4, IG, IPV6, TCP,  PAY4),
+	[0x9D] = WX_PTT(IP, IPV4, IG, IPV6, SCTP, PAY4),
+
+	/* IPv4 --> GRE/NAT --> MAC --> NONE/IPv4/IPv6 */
+	[0xA0] = WX_PTT(IP, IPV4, IGM, NONE, NONE, PAY3),
+	[0xA1] = WX_PTT(IP, IPV4, IGM, FGV4, NONE, PAY3),
+	[0xA2] = WX_PTT(IP, IPV4, IGM, IPV4, NONE, PAY3),
+	[0xA3] = WX_PTT(IP, IPV4, IGM, IPV4, UDP,  PAY4),
+	[0xA4] = WX_PTT(IP, IPV4, IGM, IPV4, TCP,  PAY4),
+	[0xA5] = WX_PTT(IP, IPV4, IGM, IPV4, SCTP, PAY4),
+	[0xA9] = WX_PTT(IP, IPV4, IGM, FGV6, NONE, PAY3),
+	[0xAA] = WX_PTT(IP, IPV4, IGM, IPV6, NONE, PAY3),
+	[0xAB] = WX_PTT(IP, IPV4, IGM, IPV6, UDP,  PAY4),
+	[0xAC] = WX_PTT(IP, IPV4, IGM, IPV6, TCP,  PAY4),
+	[0xAD] = WX_PTT(IP, IPV4, IGM, IPV6, SCTP, PAY4),
+
+	/* IPv4 --> GRE/NAT --> MAC+VLAN --> NONE/IPv4/IPv6 */
+	[0xB0] = WX_PTT(IP, IPV4, IGMV, NONE, NONE, PAY3),
+	[0xB1] = WX_PTT(IP, IPV4, IGMV, FGV4, NONE, PAY3),
+	[0xB2] = WX_PTT(IP, IPV4, IGMV, IPV4, NONE, PAY3),
+	[0xB3] = WX_PTT(IP, IPV4, IGMV, IPV4, UDP,  PAY4),
+	[0xB4] = WX_PTT(IP, IPV4, IGMV, IPV4, TCP,  PAY4),
+	[0xB5] = WX_PTT(IP, IPV4, IGMV, IPV4, SCTP, PAY4),
+	[0xB9] = WX_PTT(IP, IPV4, IGMV, FGV6, NONE, PAY3),
+	[0xBA] = WX_PTT(IP, IPV4, IGMV, IPV6, NONE, PAY3),
+	[0xBB] = WX_PTT(IP, IPV4, IGMV, IPV6, UDP,  PAY4),
+	[0xBC] = WX_PTT(IP, IPV4, IGMV, IPV6, TCP,  PAY4),
+	[0xBD] = WX_PTT(IP, IPV4, IGMV, IPV6, SCTP, PAY4),
+
+	/* IPv6 --> IPv4/IPv6 */
+	[0xC1] = WX_PTT(IP, IPV6, IPIP, FGV4, NONE, PAY3),
+	[0xC2] = WX_PTT(IP, IPV6, IPIP, IPV4, NONE, PAY3),
+	[0xC3] = WX_PTT(IP, IPV6, IPIP, IPV4, UDP,  PAY4),
+	[0xC4] = WX_PTT(IP, IPV6, IPIP, IPV4, TCP,  PAY4),
+	[0xC5] = WX_PTT(IP, IPV6, IPIP, IPV4, SCTP, PAY4),
+	[0xC9] = WX_PTT(IP, IPV6, IPIP, FGV6, NONE, PAY3),
+	[0xCA] = WX_PTT(IP, IPV6, IPIP, IPV6, NONE, PAY3),
+	[0xCB] = WX_PTT(IP, IPV6, IPIP, IPV6, UDP,  PAY4),
+	[0xCC] = WX_PTT(IP, IPV6, IPIP, IPV6, TCP,  PAY4),
+	[0xCD] = WX_PTT(IP, IPV6, IPIP, IPV6, SCTP, PAY4),
+
+	/* IPv6 --> GRE/NAT -> NONE/IPv4/IPv6 */
+	[0xD0] = WX_PTT(IP, IPV6, IG, NONE, NONE, PAY3),
+	[0xD1] = WX_PTT(IP, IPV6, IG, FGV4, NONE, PAY3),
+	[0xD2] = WX_PTT(IP, IPV6, IG, IPV4, NONE, PAY3),
+	[0xD3] = WX_PTT(IP, IPV6, IG, IPV4, UDP,  PAY4),
+	[0xD4] = WX_PTT(IP, IPV6, IG, IPV4, TCP,  PAY4),
+	[0xD5] = WX_PTT(IP, IPV6, IG, IPV4, SCTP, PAY4),
+	[0xD9] = WX_PTT(IP, IPV6, IG, FGV6, NONE, PAY3),
+	[0xDA] = WX_PTT(IP, IPV6, IG, IPV6, NONE, PAY3),
+	[0xDB] = WX_PTT(IP, IPV6, IG, IPV6, UDP,  PAY4),
+	[0xDC] = WX_PTT(IP, IPV6, IG, IPV6, TCP,  PAY4),
+	[0xDD] = WX_PTT(IP, IPV6, IG, IPV6, SCTP, PAY4),
+
+	/* IPv6 --> GRE/NAT -> MAC -> NONE/IPv4/IPv6 */
+	[0xE0] = WX_PTT(IP, IPV6, IGM, NONE, NONE, PAY3),
+	[0xE1] = WX_PTT(IP, IPV6, IGM, FGV4, NONE, PAY3),
+	[0xE2] = WX_PTT(IP, IPV6, IGM, IPV4, NONE, PAY3),
+	[0xE3] = WX_PTT(IP, IPV6, IGM, IPV4, UDP,  PAY4),
+	[0xE4] = WX_PTT(IP, IPV6, IGM, IPV4, TCP,  PAY4),
+	[0xE5] = WX_PTT(IP, IPV6, IGM, IPV4, SCTP, PAY4),
+	[0xE9] = WX_PTT(IP, IPV6, IGM, FGV6, NONE, PAY3),
+	[0xEA] = WX_PTT(IP, IPV6, IGM, IPV6, NONE, PAY3),
+	[0xEB] = WX_PTT(IP, IPV6, IGM, IPV6, UDP,  PAY4),
+	[0xEC] = WX_PTT(IP, IPV6, IGM, IPV6, TCP,  PAY4),
+	[0xED] = WX_PTT(IP, IPV6, IGM, IPV6, SCTP, PAY4),
+
+	/* IPv6 --> GRE/NAT -> MAC--> NONE/IPv */
+	[0xF0] = WX_PTT(IP, IPV6, IGMV, NONE, NONE, PAY3),
+	[0xF1] = WX_PTT(IP, IPV6, IGMV, FGV4, NONE, PAY3),
+	[0xF2] = WX_PTT(IP, IPV6, IGMV, IPV4, NONE, PAY3),
+	[0xF3] = WX_PTT(IP, IPV6, IGMV, IPV4, UDP,  PAY4),
+	[0xF4] = WX_PTT(IP, IPV6, IGMV, IPV4, TCP,  PAY4),
+	[0xF5] = WX_PTT(IP, IPV6, IGMV, IPV4, SCTP, PAY4),
+	[0xF9] = WX_PTT(IP, IPV6, IGMV, FGV6, NONE, PAY3),
+	[0xFA] = WX_PTT(IP, IPV6, IGMV, IPV6, NONE, PAY3),
+	[0xFB] = WX_PTT(IP, IPV6, IGMV, IPV6, UDP,  PAY4),
+	[0xFC] = WX_PTT(IP, IPV6, IGMV, IPV6, TCP,  PAY4),
+	[0xFD] = WX_PTT(IP, IPV6, IGMV, IPV6, SCTP, PAY4),
+};
+
+static struct wx_dec_ptype wx_decode_ptype(const u8 ptype)
+{
+	return wx_ptype_lookup[ptype];
+}
+
 /* wx_test_staterr - tests bits in Rx descriptor status and error fields */
 static __le32 wx_test_staterr(union wx_rx_desc *rx_desc,
 			      const u32 stat_err_bits)
@@ -424,6 +562,98 @@ static bool wx_cleanup_headers(struct wx_ring *rx_ring,
 	return false;
 }
 
+static void wx_rx_hash(struct wx_ring *ring,
+		       union wx_rx_desc *rx_desc,
+		       struct sk_buff *skb)
+{
+	u16 rss_type;
+
+	if (!(ring->netdev->features & NETIF_F_RXHASH))
+		return;
+
+	rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
+			       WX_RXD_RSSTYPE_MASK;
+
+	if (!rss_type)
+		return;
+
+	skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
+		     (WX_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
+		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
+}
+
+/**
+ * wx_rx_checksum - indicate in skb if hw indicated a good cksum
+ * @ring: structure containing ring specific data
+ * @rx_desc: current Rx descriptor being processed
+ * @skb: skb currently being received and modified
+ **/
+static void wx_rx_checksum(struct wx_ring *ring,
+			   union wx_rx_desc *rx_desc,
+			   struct sk_buff *skb)
+{
+	struct wx_dec_ptype dptype = wx_decode_ptype(WX_RXD_PKTTYPE(rx_desc));
+
+	skb_checksum_none_assert(skb);
+	/* Rx csum disabled */
+	if (!(ring->netdev->features & NETIF_F_RXCSUM))
+		return;
+
+	/* if IPv4 header checksum error */
+	if ((wx_test_staterr(rx_desc, WX_RXD_STAT_IPCS) &&
+	     wx_test_staterr(rx_desc, WX_RXD_ERR_IPE)) ||
+	    (wx_test_staterr(rx_desc, WX_RXD_STAT_OUTERIPCS) &&
+	     wx_test_staterr(rx_desc, WX_RXD_ERR_OUTERIPER))) {
+		ring->rx_stats.csum_err++;
+		return;
+	}
+
+	/* L4 checksum offload flag must set for the below code to work */
+	if (!wx_test_staterr(rx_desc, WX_RXD_STAT_L4CS))
+		return;
+
+	/* Hardware can't guarantee csum if IPv6 Dest Header found */
+	if (dptype.prot != WX_DEC_PTYPE_PROT_SCTP && WX_RXD_IPV6EX(rx_desc))
+		return;
+
+	/* if L4 checksum error */
+	if (wx_test_staterr(rx_desc, WX_RXD_ERR_TCPE)) {
+		ring->rx_stats.csum_err++;
+		return;
+	}
+
+	/* It must be a TCP or UDP or SCTP packet with a valid checksum */
+	skb->ip_summed = CHECKSUM_UNNECESSARY;
+
+	/* If there is an outer header present that might contain a checksum
+	 * we need to bump the checksum level by 1 to reflect the fact that
+	 * we are indicating we validated the inner checksum.
+	 */
+	if (dptype.etype >= WX_DEC_PTYPE_ETYPE_IG)
+		__skb_incr_checksum_unnecessary(skb);
+	ring->rx_stats.csum_good_cnt++;
+}
+
+/**
+ * wx_process_skb_fields - Populate skb header fields from Rx descriptor
+ * @rx_ring: rx descriptor ring packet is being transacted on
+ * @rx_desc: pointer to the EOP Rx descriptor
+ * @skb: pointer to current skb being populated
+ *
+ * This function checks the ring, descriptor, and packet information in
+ * order to populate the hash, checksum, protocol, and
+ * other fields within the skb.
+ **/
+static void wx_process_skb_fields(struct wx_ring *rx_ring,
+				  union wx_rx_desc *rx_desc,
+				  struct sk_buff *skb)
+{
+	wx_rx_hash(rx_ring, rx_desc, skb);
+	wx_rx_checksum(rx_ring, rx_desc, skb);
+	skb_record_rx_queue(skb, rx_ring->queue_index);
+	skb->protocol = eth_type_trans(skb, rx_ring->netdev);
+}
+
 /**
  * wx_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
  * @q_vector: structure containing interrupt and ring information
@@ -491,8 +721,8 @@ static int wx_clean_rx_irq(struct wx_q_vector *q_vector,
 		/* probably a little skewed due to removing CRC */
 		total_rx_bytes += skb->len;
 
-		skb_record_rx_queue(skb, rx_ring->queue_index);
-		skb->protocol = eth_type_trans(skb, rx_ring->netdev);
+		/* populate checksum, timestamp, VLAN, and protocol */
+		wx_process_skb_fields(rx_ring, rx_desc, skb);
 		napi_gro_receive(&q_vector->napi, skb);
 
 		/* update budget accounting */
diff --git a/drivers/net/ethernet/wangxun/libwx/wx_type.h b/drivers/net/ethernet/wangxun/libwx/wx_type.h
index d2e0584e3b44..47abefb5a4fc 100644
--- a/drivers/net/ethernet/wangxun/libwx/wx_type.h
+++ b/drivers/net/ethernet/wangxun/libwx/wx_type.h
@@ -321,9 +321,31 @@
 /******************* Receive Descriptor bit definitions **********************/
 #define WX_RXD_STAT_DD               BIT(0) /* Done */
 #define WX_RXD_STAT_EOP              BIT(1) /* End of Packet */
+#define WX_RXD_STAT_L4CS             BIT(7) /* L4 xsum calculated */
+#define WX_RXD_STAT_IPCS             BIT(8) /* IP xsum calculated */
+#define WX_RXD_STAT_OUTERIPCS        BIT(10) /* Cloud IP xsum calculated*/
 
+#define WX_RXD_ERR_OUTERIPER         BIT(26) /* CRC IP Header error */
 #define WX_RXD_ERR_RXE               BIT(29) /* Any MAC Error */
-
+#define WX_RXD_ERR_TCPE              BIT(30) /* TCP/UDP Checksum Error */
+#define WX_RXD_ERR_IPE               BIT(31) /* IP Checksum Error */
+
+/* RSS Hash results */
+#define WX_RXD_RSSTYPE_MASK          GENMASK(3, 0)
+#define WX_RXD_RSSTYPE_IPV4_TCP      0x00000001U
+#define WX_RXD_RSSTYPE_IPV6_TCP      0x00000003U
+#define WX_RXD_RSSTYPE_IPV4_SCTP     0x00000004U
+#define WX_RXD_RSSTYPE_IPV6_SCTP     0x00000006U
+#define WX_RXD_RSSTYPE_IPV4_UDP      0x00000007U
+#define WX_RXD_RSSTYPE_IPV6_UDP      0x00000008U
+
+#define WX_RSS_L4_TYPES_MASK \
+	((1ul << WX_RXD_RSSTYPE_IPV4_TCP) | \
+	 (1ul << WX_RXD_RSSTYPE_IPV4_UDP) | \
+	 (1ul << WX_RXD_RSSTYPE_IPV4_SCTP) | \
+	 (1ul << WX_RXD_RSSTYPE_IPV6_TCP) | \
+	 (1ul << WX_RXD_RSSTYPE_IPV6_UDP) | \
+	 (1ul << WX_RXD_RSSTYPE_IPV6_SCTP))
 /* TUN */
 #define WX_PTYPE_TUN_IPV4            0x80
 #define WX_PTYPE_TUN_IPV6            0xC0
@@ -347,6 +369,10 @@
 #define WX_PTYPE_TYP_TCP             0x04
 #define WX_PTYPE_TYP_SCTP            0x05
 
+#define WX_RXD_PKTTYPE(_rxd) \
+	((le32_to_cpu((_rxd)->wb.lower.lo_dword.data) >> 9) & 0xFF)
+#define WX_RXD_IPV6EX(_rxd) \
+	((le32_to_cpu((_rxd)->wb.lower.lo_dword.data) >> 6) & 0x1)
 /*********************** Transmit Descriptor Config Masks ****************/
 #define WX_TXD_STAT_DD               BIT(0)  /* Descriptor Done */
 #define WX_TXD_DTYP_DATA             0       /* Adv Data Descriptor */
@@ -398,6 +424,70 @@ enum wx_tx_flags {
 #define WX_TX_FLAGS_VLAN_MASK			GENMASK(31, 16)
 #define WX_TX_FLAGS_VLAN_SHIFT			16
 
+/* wx_dec_ptype.mac: outer mac */
+enum wx_dec_ptype_mac {
+	WX_DEC_PTYPE_MAC_IP	= 0,
+	WX_DEC_PTYPE_MAC_L2	= 2,
+	WX_DEC_PTYPE_MAC_FCOE	= 3,
+};
+
+/* wx_dec_ptype.[e]ip: outer&encaped ip */
+#define WX_DEC_PTYPE_IP_FRAG	0x4
+enum wx_dec_ptype_ip {
+	WX_DEC_PTYPE_IP_NONE = 0,
+	WX_DEC_PTYPE_IP_IPV4 = 1,
+	WX_DEC_PTYPE_IP_IPV6 = 2,
+	WX_DEC_PTYPE_IP_FGV4 = WX_DEC_PTYPE_IP_FRAG | WX_DEC_PTYPE_IP_IPV4,
+	WX_DEC_PTYPE_IP_FGV6 = WX_DEC_PTYPE_IP_FRAG | WX_DEC_PTYPE_IP_IPV6,
+};
+
+/* wx_dec_ptype.etype: encaped type */
+enum wx_dec_ptype_etype {
+	WX_DEC_PTYPE_ETYPE_NONE	= 0,
+	WX_DEC_PTYPE_ETYPE_IPIP	= 1,	/* IP+IP */
+	WX_DEC_PTYPE_ETYPE_IG	= 2,	/* IP+GRE */
+	WX_DEC_PTYPE_ETYPE_IGM	= 3,	/* IP+GRE+MAC */
+	WX_DEC_PTYPE_ETYPE_IGMV	= 4,	/* IP+GRE+MAC+VLAN */
+};
+
+/* wx_dec_ptype.proto: payload proto */
+enum wx_dec_ptype_prot {
+	WX_DEC_PTYPE_PROT_NONE	= 0,
+	WX_DEC_PTYPE_PROT_UDP	= 1,
+	WX_DEC_PTYPE_PROT_TCP	= 2,
+	WX_DEC_PTYPE_PROT_SCTP	= 3,
+	WX_DEC_PTYPE_PROT_ICMP	= 4,
+	WX_DEC_PTYPE_PROT_TS	= 5,	/* time sync */
+};
+
+/* wx_dec_ptype.layer: payload layer */
+enum wx_dec_ptype_layer {
+	WX_DEC_PTYPE_LAYER_NONE = 0,
+	WX_DEC_PTYPE_LAYER_PAY2 = 1,
+	WX_DEC_PTYPE_LAYER_PAY3 = 2,
+	WX_DEC_PTYPE_LAYER_PAY4 = 3,
+};
+
+struct wx_dec_ptype {
+	u32 known:1;
+	u32 mac:2;	/* outer mac */
+	u32 ip:3;	/* outer ip*/
+	u32 etype:3;	/* encaped type */
+	u32 eip:3;	/* encaped ip */
+	u32 prot:4;	/* payload proto */
+	u32 layer:3;	/* payload layer */
+};
+
+/* macro to make the table lines short */
+#define WX_PTT(mac, ip, etype, eip, proto, layer)\
+	      {1, \
+	       WX_DEC_PTYPE_MAC_##mac,		/* mac */\
+	       WX_DEC_PTYPE_IP_##ip,		/* ip */ \
+	       WX_DEC_PTYPE_ETYPE_##etype,	/* etype */\
+	       WX_DEC_PTYPE_IP_##eip,		/* eip */\
+	       WX_DEC_PTYPE_PROT_##proto,	/* proto */\
+	       WX_DEC_PTYPE_LAYER_##layer	/* layer */}
+
 /* Host Interface Command Structures */
 struct wx_hic_hdr {
 	u8 cmd;
@@ -620,6 +710,11 @@ struct wx_queue_stats {
 	u64 bytes;
 };
 
+struct wx_rx_queue_stats {
+	u64 csum_good_cnt;
+	u64 csum_err;
+};
+
 /* iterator for handling rings in ring container */
 #define wx_for_each_ring(posm, headm) \
 	for (posm = (headm).ring; posm; posm = posm->next)
@@ -661,6 +756,9 @@ struct wx_ring {
 
 	struct wx_queue_stats stats;
 	struct u64_stats_sync syncp;
+	union {
+		struct wx_rx_queue_stats rx_stats;
+	};
 } ____cacheline_internodealigned_in_smp;
 
 struct wx_q_vector {
-- 
2.40.1

[RESEND,PATCH net-next v7 3/8] net: wangxun: Implement vlan add and kill functions

[Mengyuan Lou]

Implement vlan add/kill functions which add and remove
vlan id in hardware.

Signed-off-by: Mengyuan Lou <mengyuanlou@net-swift.com>
---
 drivers/net/ethernet/wangxun/libwx/wx_hw.c   | 272 ++++++++++++++++++-
 drivers/net/ethernet/wangxun/libwx/wx_hw.h   |   3 +
 drivers/net/ethernet/wangxun/libwx/wx_lib.c  |  18 ++
 drivers/net/ethernet/wangxun/libwx/wx_type.h |  25 +-
 4 files changed, 316 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/wangxun/libwx/wx_hw.c b/drivers/net/ethernet/wangxun/libwx/wx_hw.c
index ca409b4054d0..39a9aeee7aab 100644
--- a/drivers/net/ethernet/wangxun/libwx/wx_hw.c
+++ b/drivers/net/ethernet/wangxun/libwx/wx_hw.c
@@ -1182,12 +1182,28 @@ static void wx_enable_sec_rx_path(struct wx *wx)
 	WX_WRITE_FLUSH(wx);
 }
 
+static void wx_vlan_strip_control(struct wx *wx, bool enable)
+{
+	int i, j;
+
+	for (i = 0; i < wx->num_rx_queues; i++) {
+		struct wx_ring *ring = wx->rx_ring[i];
+
+		j = ring->reg_idx;
+		wr32m(wx, WX_PX_RR_CFG(j), WX_PX_RR_CFG_VLAN,
+		      enable ? WX_PX_RR_CFG_VLAN : 0);
+	}
+}
+
 void wx_set_rx_mode(struct net_device *netdev)
 {
 	struct wx *wx = netdev_priv(netdev);
+	netdev_features_t features;
 	u32 fctrl, vmolr, vlnctrl;
 	int count;
 
+	features = netdev->features;
+
 	/* Check for Promiscuous and All Multicast modes */
 	fctrl = rd32(wx, WX_PSR_CTL);
 	fctrl &= ~(WX_PSR_CTL_UPE | WX_PSR_CTL_MPE);
@@ -1254,6 +1270,13 @@ void wx_set_rx_mode(struct net_device *netdev)
 	wr32(wx, WX_PSR_VLAN_CTL, vlnctrl);
 	wr32(wx, WX_PSR_CTL, fctrl);
 	wr32(wx, WX_PSR_VM_L2CTL(0), vmolr);
+
+	if ((features & NETIF_F_HW_VLAN_CTAG_RX) &&
+	    (features & NETIF_F_HW_VLAN_STAG_RX))
+		wx_vlan_strip_control(wx, true);
+	else
+		wx_vlan_strip_control(wx, false);
+
 }
 EXPORT_SYMBOL(wx_set_rx_mode);
 
@@ -1462,6 +1485,16 @@ static void wx_configure_tx(struct wx *wx)
 	      WX_MAC_TX_CFG_TE, WX_MAC_TX_CFG_TE);
 }
 
+static void wx_restore_vlan(struct wx *wx)
+{
+	u16 vid = 1;
+
+	wx_vlan_rx_add_vid(wx->netdev, htons(ETH_P_8021Q), 0);
+
+	for_each_set_bit_from(vid, wx->active_vlans, VLAN_N_VID)
+		wx_vlan_rx_add_vid(wx->netdev, htons(ETH_P_8021Q), vid);
+}
+
 /**
  * wx_configure_rx - Configure Receive Unit after Reset
  * @wx: pointer to private structure
@@ -1527,7 +1560,7 @@ void wx_configure(struct wx *wx)
 	wx_configure_port(wx);
 
 	wx_set_rx_mode(wx->netdev);
-
+	wx_restore_vlan(wx);
 	wx_enable_sec_rx_path(wx);
 
 	wx_configure_tx(wx);
@@ -1727,4 +1760,241 @@ int wx_sw_init(struct wx *wx)
 }
 EXPORT_SYMBOL(wx_sw_init);
 
+/**
+ *  wx_find_vlvf_slot - find the vlanid or the first empty slot
+ *  @wx: pointer to hardware structure
+ *  @vlan: VLAN id to write to VLAN filter
+ *
+ *  return the VLVF index where this VLAN id should be placed
+ *
+ **/
+static int wx_find_vlvf_slot(struct wx *wx, u32 vlan)
+{
+	u32 bits = 0, first_empty_slot = 0;
+	int regindex;
+
+	/* short cut the special case */
+	if (vlan == 0)
+		return 0;
+
+	/* Search for the vlan id in the VLVF entries. Save off the first empty
+	 * slot found along the way
+	 */
+	for (regindex = 1; regindex < WX_PSR_VLAN_SWC_ENTRIES; regindex++) {
+		wr32(wx, WX_PSR_VLAN_SWC_IDX, regindex);
+		bits = rd32(wx, WX_PSR_VLAN_SWC);
+		if (!bits && !(first_empty_slot))
+			first_empty_slot = regindex;
+		else if ((bits & 0x0FFF) == vlan)
+			break;
+	}
+
+	if (regindex >= WX_PSR_VLAN_SWC_ENTRIES) {
+		if (first_empty_slot)
+			regindex = first_empty_slot;
+		else
+			regindex = -ENOMEM;
+	}
+
+	return regindex;
+}
+
+/**
+ *  wx_set_vlvf - Set VLAN Pool Filter
+ *  @wx: pointer to hardware structure
+ *  @vlan: VLAN id to write to VLAN filter
+ *  @vind: VMDq output index that maps queue to VLAN id in VFVFB
+ *  @vlan_on: boolean flag to turn on/off VLAN in VFVF
+ *  @vfta_changed: pointer to boolean flag which indicates whether VFTA
+ *                 should be changed
+ *
+ *  Turn on/off specified bit in VLVF table.
+ **/
+static int wx_set_vlvf(struct wx *wx, u32 vlan, u32 vind, bool vlan_on,
+		       bool *vfta_changed)
+{
+	int vlvf_index;
+	u32 vt, bits;
+
+	/* If VT Mode is set
+	 *   Either vlan_on
+	 *     make sure the vlan is in VLVF
+	 *     set the vind bit in the matching VLVFB
+	 *   Or !vlan_on
+	 *     clear the pool bit and possibly the vind
+	 */
+	vt = rd32(wx, WX_CFG_PORT_CTL);
+	if (!(vt & WX_CFG_PORT_CTL_NUM_VT_MASK))
+		return 0;
+
+	vlvf_index = wx_find_vlvf_slot(wx, vlan);
+	if (vlvf_index < 0)
+		return vlvf_index;
+
+	wr32(wx, WX_PSR_VLAN_SWC_IDX, vlvf_index);
+	if (vlan_on) {
+		/* set the pool bit */
+		if (vind < 32) {
+			bits = rd32(wx, WX_PSR_VLAN_SWC_VM_L);
+			bits |= (1 << vind);
+			wr32(wx, WX_PSR_VLAN_SWC_VM_L, bits);
+		} else {
+			bits = rd32(wx, WX_PSR_VLAN_SWC_VM_H);
+			bits |= (1 << (vind - 32));
+			wr32(wx, WX_PSR_VLAN_SWC_VM_H, bits);
+		}
+	} else {
+		/* clear the pool bit */
+		if (vind < 32) {
+			bits = rd32(wx, WX_PSR_VLAN_SWC_VM_L);
+			bits &= ~(1 << vind);
+			wr32(wx, WX_PSR_VLAN_SWC_VM_L, bits);
+			bits |= rd32(wx, WX_PSR_VLAN_SWC_VM_H);
+		} else {
+			bits = rd32(wx, WX_PSR_VLAN_SWC_VM_H);
+			bits &= ~(1 << (vind - 32));
+			wr32(wx, WX_PSR_VLAN_SWC_VM_H, bits);
+			bits |= rd32(wx, WX_PSR_VLAN_SWC_VM_L);
+		}
+	}
+
+	if (bits) {
+		wr32(wx, WX_PSR_VLAN_SWC, (WX_PSR_VLAN_SWC_VIEN | vlan));
+		if (!vlan_on && vfta_changed)
+			*vfta_changed = false;
+	} else {
+		wr32(wx, WX_PSR_VLAN_SWC, 0);
+	}
+
+	return 0;
+}
+
+/**
+ *  wx_set_vfta - Set VLAN filter table
+ *  @wx: pointer to hardware structure
+ *  @vlan: VLAN id to write to VLAN filter
+ *  @vind: VMDq output index that maps queue to VLAN id in VFVFB
+ *  @vlan_on: boolean flag to turn on/off VLAN in VFVF
+ *
+ *  Turn on/off specified VLAN in the VLAN filter table.
+ **/
+static int wx_set_vfta(struct wx *wx, u32 vlan, u32 vind, bool vlan_on)
+{
+	u32 bitindex, vfta, targetbit;
+	bool vfta_changed = false;
+	int regindex, ret;
+
+	/* this is a 2 part operation - first the VFTA, then the
+	 * VLVF and VLVFB if VT Mode is set
+	 * We don't write the VFTA until we know the VLVF part succeeded.
+	 */
+
+	/* Part 1
+	 * The VFTA is a bitstring made up of 128 32-bit registers
+	 * that enable the particular VLAN id, much like the MTA:
+	 *    bits[11-5]: which register
+	 *    bits[4-0]:  which bit in the register
+	 */
+	regindex = (vlan >> 5) & 0x7F;
+	bitindex = vlan & 0x1F;
+	targetbit = (1 << bitindex);
+	/* errata 5 */
+	vfta = wx->mac.vft_shadow[regindex];
+	if (vlan_on) {
+		if (!(vfta & targetbit)) {
+			vfta |= targetbit;
+			vfta_changed = true;
+		}
+	} else {
+		if ((vfta & targetbit)) {
+			vfta &= ~targetbit;
+			vfta_changed = true;
+		}
+	}
+	/* Part 2
+	 * Call wx_set_vlvf to set VLVFB and VLVF
+	 */
+	ret = wx_set_vlvf(wx, vlan, vind, vlan_on, &vfta_changed);
+	if (ret != 0)
+		return ret;
+
+	if (vfta_changed)
+		wr32(wx, WX_PSR_VLAN_TBL(regindex), vfta);
+	wx->mac.vft_shadow[regindex] = vfta;
+
+	return 0;
+}
+
+/**
+ *  wx_clear_vfta - Clear VLAN filter table
+ *  @wx: pointer to hardware structure
+ *
+ *  Clears the VLAN filer table, and the VMDq index associated with the filter
+ **/
+static void wx_clear_vfta(struct wx *wx)
+{
+	u32 offset;
+
+	for (offset = 0; offset < wx->mac.vft_size; offset++) {
+		wr32(wx, WX_PSR_VLAN_TBL(offset), 0);
+		wx->mac.vft_shadow[offset] = 0;
+	}
+
+	for (offset = 0; offset < WX_PSR_VLAN_SWC_ENTRIES; offset++) {
+		wr32(wx, WX_PSR_VLAN_SWC_IDX, offset);
+		wr32(wx, WX_PSR_VLAN_SWC, 0);
+		wr32(wx, WX_PSR_VLAN_SWC_VM_L, 0);
+		wr32(wx, WX_PSR_VLAN_SWC_VM_H, 0);
+	}
+}
+
+int wx_vlan_rx_add_vid(struct net_device *netdev,
+		       __be16 proto, u16 vid)
+{
+	struct wx *wx = netdev_priv(netdev);
+
+	/* add VID to filter table */
+	wx_set_vfta(wx, vid, VMDQ_P(0), true);
+	set_bit(vid, wx->active_vlans);
+
+	return 0;
+}
+EXPORT_SYMBOL(wx_vlan_rx_add_vid);
+
+int wx_vlan_rx_kill_vid(struct net_device *netdev, __be16 proto, u16 vid)
+{
+	struct wx *wx = netdev_priv(netdev);
+
+	/* remove VID from filter table */
+	if (vid)
+		wx_set_vfta(wx, vid, VMDQ_P(0), false);
+	clear_bit(vid, wx->active_vlans);
+
+	return 0;
+}
+EXPORT_SYMBOL(wx_vlan_rx_kill_vid);
+
+/**
+ *  wx_start_hw - Prepare hardware for Tx/Rx
+ *  @wx: pointer to hardware structure
+ *
+ *  Starts the hardware using the generic start_hw function
+ *  and the generation start_hw function.
+ *  Then performs revision-specific operations, if any.
+ **/
+void wx_start_hw(struct wx *wx)
+{
+	int i;
+
+	/* Clear the VLAN filter table */
+	wx_clear_vfta(wx);
+	WX_WRITE_FLUSH(wx);
+	/* Clear the rate limiters */
+	for (i = 0; i < wx->mac.max_tx_queues; i++) {
+		wr32(wx, WX_TDM_RP_IDX, i);
+		wr32(wx, WX_TDM_RP_RATE, 0);
+	}
+}
+EXPORT_SYMBOL(wx_start_hw);
+
 MODULE_LICENSE("GPL");
diff --git a/drivers/net/ethernet/wangxun/libwx/wx_hw.h b/drivers/net/ethernet/wangxun/libwx/wx_hw.h
index c173c56f0ab5..1f93ca32c921 100644
--- a/drivers/net/ethernet/wangxun/libwx/wx_hw.h
+++ b/drivers/net/ethernet/wangxun/libwx/wx_hw.h
@@ -26,10 +26,13 @@ void wx_set_rx_mode(struct net_device *netdev);
 int wx_change_mtu(struct net_device *netdev, int new_mtu);
 void wx_disable_rx_queue(struct wx *wx, struct wx_ring *ring);
 void wx_configure(struct wx *wx);
+void wx_start_hw(struct wx *wx);
 int wx_disable_pcie_master(struct wx *wx);
 int wx_stop_adapter(struct wx *wx);
 void wx_reset_misc(struct wx *wx);
 int wx_get_pcie_msix_counts(struct wx *wx, u16 *msix_count, u16 max_msix_count);
 int wx_sw_init(struct wx *wx);
+int wx_vlan_rx_add_vid(struct net_device *netdev, __be16 proto, u16 vid);
+int wx_vlan_rx_kill_vid(struct net_device *netdev, __be16 proto, u16 vid);
 
 #endif /* _WX_HW_H_ */
diff --git a/drivers/net/ethernet/wangxun/libwx/wx_lib.c b/drivers/net/ethernet/wangxun/libwx/wx_lib.c
index 84107208401e..680f1ad36240 100644
--- a/drivers/net/ethernet/wangxun/libwx/wx_lib.c
+++ b/drivers/net/ethernet/wangxun/libwx/wx_lib.c
@@ -634,6 +634,23 @@ static void wx_rx_checksum(struct wx_ring *ring,
 	ring->rx_stats.csum_good_cnt++;
 }
 
+static void wx_rx_vlan(struct wx_ring *ring, union wx_rx_desc *rx_desc,
+		       struct sk_buff *skb)
+{
+	u16 ethertype;
+	u8 idx = 0;
+
+	if ((ring->netdev->features &
+	     (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_STAG_RX)) &&
+	    wx_test_staterr(rx_desc, WX_RXD_STAT_VP)) {
+		idx = (le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
+		       0x1c0) >> 6;
+		ethertype = ring->q_vector->wx->tpid[idx];
+		__vlan_hwaccel_put_tag(skb, htons(ethertype),
+				       le16_to_cpu(rx_desc->wb.upper.vlan));
+	}
+}
+
 /**
  * wx_process_skb_fields - Populate skb header fields from Rx descriptor
  * @rx_ring: rx descriptor ring packet is being transacted on
@@ -650,6 +667,7 @@ static void wx_process_skb_fields(struct wx_ring *rx_ring,
 {
 	wx_rx_hash(rx_ring, rx_desc, skb);
 	wx_rx_checksum(rx_ring, rx_desc, skb);
+	wx_rx_vlan(rx_ring, rx_desc, skb);
 	skb_record_rx_queue(skb, rx_ring->queue_index);
 	skb->protocol = eth_type_trans(skb, rx_ring->netdev);
 }
diff --git a/drivers/net/ethernet/wangxun/libwx/wx_type.h b/drivers/net/ethernet/wangxun/libwx/wx_type.h
index 47abefb5a4fc..a7ff6c6749d0 100644
--- a/drivers/net/ethernet/wangxun/libwx/wx_type.h
+++ b/drivers/net/ethernet/wangxun/libwx/wx_type.h
@@ -6,6 +6,7 @@
 
 #include <linux/bitfield.h>
 #include <linux/netdevice.h>
+#include <linux/if_vlan.h>
 #include <net/ip.h>
 
 #define WX_NCSI_SUP                             0x8000
@@ -65,6 +66,8 @@
 #define WX_CFG_PORT_CTL_QINQ         BIT(2)
 #define WX_CFG_PORT_CTL_D_VLAN       BIT(0) /* double vlan*/
 #define WX_CFG_TAG_TPID(_i)          (0x14430 + ((_i) * 4))
+#define WX_CFG_PORT_CTL_NUM_VT_MASK  GENMASK(13, 12) /* number of TVs */
+
 
 /* GPIO Registers */
 #define WX_GPIO_DR                   0x14800
@@ -88,6 +91,8 @@
 /* TDM CTL BIT */
 #define WX_TDM_CTL_TE                BIT(0) /* Transmit Enable */
 #define WX_TDM_PB_THRE(_i)           (0x18020 + ((_i) * 4))
+#define WX_TDM_RP_IDX                0x1820C
+#define WX_TDM_RP_RATE               0x18404
 
 /***************************** RDB registers *********************************/
 /* receive packet buffer */
@@ -151,6 +156,9 @@
 #define WX_PSR_LAN_FLEX_DW_H(_i)     (0x15C04 + ((_i) * 16))
 #define WX_PSR_LAN_FLEX_MSK(_i)      (0x15C08 + ((_i) * 16))
 
+/* vlan tbl */
+#define WX_PSR_VLAN_TBL(_i)          (0x16000 + ((_i) * 4))
+
 /* mac switcher */
 #define WX_PSR_MAC_SWC_AD_L          0x16200
 #define WX_PSR_MAC_SWC_AD_H          0x16204
@@ -162,6 +170,15 @@
 #define WX_PSR_MAC_SWC_IDX           0x16210
 #define WX_CLEAR_VMDQ_ALL            0xFFFFFFFFU
 
+/* vlan switch */
+#define WX_PSR_VLAN_SWC              0x16220
+#define WX_PSR_VLAN_SWC_VM_L         0x16224
+#define WX_PSR_VLAN_SWC_VM_H         0x16228
+#define WX_PSR_VLAN_SWC_IDX          0x16230         /* 64 vlan entries */
+/* VLAN pool filtering masks */
+#define WX_PSR_VLAN_SWC_VIEN         BIT(31)  /* filter is valid */
+#define WX_PSR_VLAN_SWC_ENTRIES      64
+
 /********************************* RSEC **************************************/
 /* general rsec */
 #define WX_RSC_CTL                   0x17000
@@ -256,6 +273,7 @@
 #define WX_PX_RR_RP(_i)              (0x0100C + ((_i) * 0x40))
 #define WX_PX_RR_CFG(_i)             (0x01010 + ((_i) * 0x40))
 /* PX_RR_CFG bit definitions */
+#define WX_PX_RR_CFG_VLAN            BIT(31)
 #define WX_PX_RR_CFG_SPLIT_MODE      BIT(26)
 #define WX_PX_RR_CFG_RR_THER_SHIFT   16
 #define WX_PX_RR_CFG_RR_HDR_SZ       GENMASK(15, 12)
@@ -297,6 +315,7 @@
 #define WX_MAX_TXD                   8192
 
 #define WX_MAX_JUMBO_FRAME_SIZE      9432 /* max payload 9414 */
+#define VMDQ_P(p)                    p
 
 /* Supported Rx Buffer Sizes */
 #define WX_RXBUFFER_256      256    /* Used for skb receive header */
@@ -321,6 +340,7 @@
 /******************* Receive Descriptor bit definitions **********************/
 #define WX_RXD_STAT_DD               BIT(0) /* Done */
 #define WX_RXD_STAT_EOP              BIT(1) /* End of Packet */
+#define WX_RXD_STAT_VP               BIT(5) /* IEEE VLAN Pkt */
 #define WX_RXD_STAT_L4CS             BIT(7) /* L4 xsum calculated */
 #define WX_RXD_STAT_IPCS             BIT(8) /* IP xsum calculated */
 #define WX_RXD_STAT_OUTERIPCS        BIT(10) /* Cloud IP xsum calculated*/
@@ -566,6 +586,8 @@ struct wx_mac_info {
 	u32 mta_shadow[128];
 	s32 mc_filter_type;
 	u32 mcft_size;
+	u32 vft_shadow[128];
+	u32 vft_size;
 	u32 num_rar_entries;
 	u32 rx_pb_size;
 	u32 tx_pb_size;
@@ -726,7 +748,6 @@ struct wx_ring_container {
 	u8 count;                       /* total number of rings in vector */
 	u8 itr;                         /* current ITR setting for ring */
 };
-
 struct wx_ring {
 	struct wx_ring *next;           /* pointer to next ring in q_vector */
 	struct wx_q_vector *q_vector;   /* backpointer to host q_vector */
@@ -789,6 +810,8 @@ enum wx_isb_idx {
 };
 
 struct wx {
+	unsigned long active_vlans[BITS_TO_LONGS(VLAN_N_VID)];
+
 	u8 __iomem *hw_addr;
 	struct pci_dev *pdev;
 	struct net_device *netdev;
-- 
2.40.1

[RESEND,PATCH net-next v7 4/8] net: libwx: Implement xx_set_features ops

[Mengyuan Lou]

Implement wx_set_features function which to support
ndo_set_features.

Signed-off-by: Mengyuan Lou <mengyuanlou@net-swift.com>
---
 drivers/net/ethernet/wangxun/libwx/wx_lib.c  | 20 ++++++++++++++++++++
 drivers/net/ethernet/wangxun/libwx/wx_lib.h  |  1 +
 drivers/net/ethernet/wangxun/libwx/wx_type.h |  2 ++
 3 files changed, 23 insertions(+)

diff --git a/drivers/net/ethernet/wangxun/libwx/wx_lib.c b/drivers/net/ethernet/wangxun/libwx/wx_lib.c
index 680f1ad36240..3dd328d33fcc 100644
--- a/drivers/net/ethernet/wangxun/libwx/wx_lib.c
+++ b/drivers/net/ethernet/wangxun/libwx/wx_lib.c
@@ -2704,4 +2704,24 @@ void wx_get_stats64(struct net_device *netdev,
 }
 EXPORT_SYMBOL(wx_get_stats64);
 
+int wx_set_features(struct net_device *netdev, netdev_features_t features)
+{
+	netdev_features_t changed = netdev->features ^ features;
+	struct wx *wx = netdev_priv(netdev);
+
+	if (changed & NETIF_F_RXHASH)
+		wr32m(wx, WX_RDB_RA_CTL, WX_RDB_RA_CTL_RSS_EN,
+		      WX_RDB_RA_CTL_RSS_EN);
+	else
+		wr32m(wx, WX_RDB_RA_CTL, WX_RDB_RA_CTL_RSS_EN, 0);
+
+	if (changed &
+	    (NETIF_F_HW_VLAN_CTAG_RX |
+	     NETIF_F_HW_VLAN_STAG_RX))
+		wx_set_rx_mode(netdev);
+
+	return 1;
+}
+EXPORT_SYMBOL(wx_set_features);
+
 MODULE_LICENSE("GPL");
diff --git a/drivers/net/ethernet/wangxun/libwx/wx_lib.h b/drivers/net/ethernet/wangxun/libwx/wx_lib.h
index 50ee41f1fa10..df1f4a5951f0 100644
--- a/drivers/net/ethernet/wangxun/libwx/wx_lib.h
+++ b/drivers/net/ethernet/wangxun/libwx/wx_lib.h
@@ -28,5 +28,6 @@ void wx_free_resources(struct wx *wx);
 int wx_setup_resources(struct wx *wx);
 void wx_get_stats64(struct net_device *netdev,
 		    struct rtnl_link_stats64 *stats);
+int wx_set_features(struct net_device *netdev, netdev_features_t features);
 
 #endif /* _NGBE_LIB_H_ */
diff --git a/drivers/net/ethernet/wangxun/libwx/wx_type.h b/drivers/net/ethernet/wangxun/libwx/wx_type.h
index a7ff6c6749d0..a52cc5ac1db6 100644
--- a/drivers/net/ethernet/wangxun/libwx/wx_type.h
+++ b/drivers/net/ethernet/wangxun/libwx/wx_type.h
@@ -111,6 +111,8 @@
 #define WX_RDB_PL_CFG_L2HDR          BIT(3)
 #define WX_RDB_PL_CFG_TUN_TUNHDR     BIT(4)
 #define WX_RDB_PL_CFG_TUN_OUTL2HDR   BIT(5)
+#define WX_RDB_RA_CTL                0x194F4
+#define WX_RDB_RA_CTL_RSS_EN         BIT(2) /* RSS Enable */
 
 /******************************* PSR Registers *******************************/
 /* psr control */
-- 
2.40.1

[RESEND,PATCH net-next v7 5/8] net: ngbe: Add netdev features support

[Mengyuan Lou]

Add features and hw_features that ngbe can support.

Signed-off-by: Mengyuan Lou <mengyuanlou@net-swift.com>
---
 drivers/net/ethernet/wangxun/ngbe/ngbe_main.c | 17 ++++++++++++-----
 1 file changed, 12 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/wangxun/ngbe/ngbe_main.c b/drivers/net/ethernet/wangxun/ngbe/ngbe_main.c
index df6b870aa871..f234c9c4b942 100644
--- a/drivers/net/ethernet/wangxun/ngbe/ngbe_main.c
+++ b/drivers/net/ethernet/wangxun/ngbe/ngbe_main.c
@@ -473,6 +473,7 @@ static const struct net_device_ops ngbe_netdev_ops = {
 	.ndo_change_mtu         = wx_change_mtu,
 	.ndo_start_xmit         = wx_xmit_frame,
 	.ndo_set_rx_mode        = wx_set_rx_mode,
+	.ndo_set_features       = wx_set_features,
 	.ndo_validate_addr      = eth_validate_addr,
 	.ndo_set_mac_address    = wx_set_mac,
 	.ndo_get_stats64        = wx_get_stats64,
@@ -551,12 +552,18 @@ static int ngbe_probe(struct pci_dev *pdev,
 	ngbe_set_ethtool_ops(netdev);
 	netdev->netdev_ops = &ngbe_netdev_ops;
 
-	netdev->features |= NETIF_F_HIGHDMA;
-	netdev->features = NETIF_F_SG;
-
+	netdev->features = NETIF_F_SG | NETIF_F_IP_CSUM |
+			   NETIF_F_TSO | NETIF_F_TSO6 |
+			   NETIF_F_RXHASH | NETIF_F_RXCSUM;
+	netdev->features |= NETIF_F_SCTP_CRC | NETIF_F_TSO_MANGLEID;
+	netdev->vlan_features |= netdev->features;
+	netdev->features |= NETIF_F_IPV6_CSUM | NETIF_F_VLAN_FEATURES;
 	/* copy netdev features into list of user selectable features */
-	netdev->hw_features |= netdev->features |
-			       NETIF_F_RXALL;
+	netdev->hw_features |= netdev->features | NETIF_F_RXALL;
+	netdev->hw_features |= NETIF_F_NTUPLE | NETIF_F_HW_TC;
+	netdev->features |= NETIF_F_HIGHDMA;
+	netdev->hw_features |= NETIF_F_GRO;
+	netdev->features |= NETIF_F_GRO;
 
 	netdev->priv_flags |= IFF_UNICAST_FLT;
 	netdev->priv_flags |= IFF_SUPP_NOFCS;
-- 
2.40.1

[RESEND,PATCH net-next v7 6/8] net: ngbe: Implement vlan add and remove ops

[Mengyuan Lou]

ngbe add ndo_vlan_rx_add_vid and ndo_vlan_rx_kill_vid.

Signed-off-by: Mengyuan Lou <mengyuanlou@net-swift.com>
---
 drivers/net/ethernet/wangxun/ngbe/ngbe_main.c | 3 +++
 drivers/net/ethernet/wangxun/ngbe/ngbe_type.h | 1 +
 2 files changed, 4 insertions(+)

diff --git a/drivers/net/ethernet/wangxun/ngbe/ngbe_main.c b/drivers/net/ethernet/wangxun/ngbe/ngbe_main.c
index f234c9c4b942..c99a5d3de72e 100644
--- a/drivers/net/ethernet/wangxun/ngbe/ngbe_main.c
+++ b/drivers/net/ethernet/wangxun/ngbe/ngbe_main.c
@@ -115,6 +115,7 @@ static int ngbe_sw_init(struct wx *wx)
 	wx->mac.max_rx_queues = NGBE_MAX_RX_QUEUES;
 	wx->mac.max_tx_queues = NGBE_MAX_TX_QUEUES;
 	wx->mac.mcft_size = NGBE_MC_TBL_SIZE;
+	wx->mac.vft_size = NGBE_SP_VFT_TBL_SIZE;
 	wx->mac.rx_pb_size = NGBE_RX_PB_SIZE;
 	wx->mac.tx_pb_size = NGBE_TDB_PB_SZ;
 
@@ -477,6 +478,8 @@ static const struct net_device_ops ngbe_netdev_ops = {
 	.ndo_validate_addr      = eth_validate_addr,
 	.ndo_set_mac_address    = wx_set_mac,
 	.ndo_get_stats64        = wx_get_stats64,
+	.ndo_vlan_rx_add_vid    = wx_vlan_rx_add_vid,
+	.ndo_vlan_rx_kill_vid   = wx_vlan_rx_kill_vid,
 };
 
 /**
diff --git a/drivers/net/ethernet/wangxun/ngbe/ngbe_type.h b/drivers/net/ethernet/wangxun/ngbe/ngbe_type.h
index 373d5af628cd..b70eca397b67 100644
--- a/drivers/net/ethernet/wangxun/ngbe/ngbe_type.h
+++ b/drivers/net/ethernet/wangxun/ngbe/ngbe_type.h
@@ -136,6 +136,7 @@ enum NGBE_MSCA_CMD_value {
 #define NGBE_RAR_ENTRIES			32
 #define NGBE_RX_PB_SIZE				42
 #define NGBE_MC_TBL_SIZE			128
+#define NGBE_SP_VFT_TBL_SIZE			128
 #define NGBE_TDB_PB_SZ				(20 * 1024) /* 160KB Packet Buffer */
 
 /* TX/RX descriptor defines */
-- 
2.40.1

[RESEND,PATCH net-next v7 7/8] net: txgbe: Add netdev features support

[Mengyuan Lou]

Add features and hw_features that ngbe can support.

Signed-off-by: Mengyuan Lou <mengyuanlou@net-swift.com>
---
 .../net/ethernet/wangxun/txgbe/txgbe_main.c   | 21 ++++++++++++++++---
 1 file changed, 18 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/wangxun/txgbe/txgbe_main.c b/drivers/net/ethernet/wangxun/txgbe/txgbe_main.c
index 5b8a121fb496..bcc9c2959177 100644
--- a/drivers/net/ethernet/wangxun/txgbe/txgbe_main.c
+++ b/drivers/net/ethernet/wangxun/txgbe/txgbe_main.c
@@ -491,6 +491,7 @@ static const struct net_device_ops txgbe_netdev_ops = {
 	.ndo_change_mtu         = wx_change_mtu,
 	.ndo_start_xmit         = wx_xmit_frame,
 	.ndo_set_rx_mode        = wx_set_rx_mode,
+	.ndo_set_features       = wx_set_features,
 	.ndo_validate_addr      = eth_validate_addr,
 	.ndo_set_mac_address    = wx_set_mac,
 	.ndo_get_stats64        = wx_get_stats64,
@@ -596,11 +597,25 @@ static int txgbe_probe(struct pci_dev *pdev,
 		goto err_free_mac_table;
 	}
 
-	netdev->features |= NETIF_F_HIGHDMA;
-	netdev->features = NETIF_F_SG;
-
+	netdev->features = NETIF_F_SG |
+			   NETIF_F_TSO |
+			   NETIF_F_TSO6 |
+			   NETIF_F_RXHASH |
+			   NETIF_F_RXCSUM |
+			   NETIF_F_HW_CSUM;
+
+	netdev->gso_partial_features =  NETIF_F_GSO_ENCAP_ALL;
+	netdev->features |= netdev->gso_partial_features;
+	netdev->features |= NETIF_F_SCTP_CRC;
+	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
+	netdev->hw_enc_features |= netdev->vlan_features;
+	netdev->features |= NETIF_F_VLAN_FEATURES;
 	/* copy netdev features into list of user selectable features */
 	netdev->hw_features |= netdev->features | NETIF_F_RXALL;
+	netdev->hw_features |= NETIF_F_NTUPLE | NETIF_F_HW_TC;
+	netdev->features |= NETIF_F_HIGHDMA;
+	netdev->hw_features |= NETIF_F_GRO;
+	netdev->features |= NETIF_F_GRO;
 
 	netdev->priv_flags |= IFF_UNICAST_FLT;
 	netdev->priv_flags |= IFF_SUPP_NOFCS;
-- 
2.40.1

[RESEND,PATCH net-next v7 8/8] net: txgbe: Implement vlan add and remove ops

[Mengyuan Lou]

txgbe add ndo_vlan_rx_add_vid and ndo_vlan_rx_kill_vid.

Signed-off-by: Mengyuan Lou <mengyuanlou@net-swift.com>
---
 drivers/net/ethernet/wangxun/txgbe/txgbe_main.c | 4 ++++
 drivers/net/ethernet/wangxun/txgbe/txgbe_type.h | 1 +
 2 files changed, 5 insertions(+)

diff --git a/drivers/net/ethernet/wangxun/txgbe/txgbe_main.c b/drivers/net/ethernet/wangxun/txgbe/txgbe_main.c
index bcc9c2959177..0f0d9fa1cde1 100644
--- a/drivers/net/ethernet/wangxun/txgbe/txgbe_main.c
+++ b/drivers/net/ethernet/wangxun/txgbe/txgbe_main.c
@@ -258,6 +258,7 @@ static void txgbe_reset(struct wx *wx)
 	if (err != 0)
 		wx_err(wx, "Hardware Error: %d\n", err);
 
+	wx_start_hw(wx);
 	/* do not flush user set addresses */
 	memcpy(old_addr, &wx->mac_table[0].addr, netdev->addr_len);
 	wx_flush_sw_mac_table(wx);
@@ -330,6 +331,7 @@ static int txgbe_sw_init(struct wx *wx)
 	wx->mac.max_tx_queues = TXGBE_SP_MAX_TX_QUEUES;
 	wx->mac.max_rx_queues = TXGBE_SP_MAX_RX_QUEUES;
 	wx->mac.mcft_size = TXGBE_SP_MC_TBL_SIZE;
+	wx->mac.vft_size = TXGBE_SP_VFT_TBL_SIZE;
 	wx->mac.rx_pb_size = TXGBE_SP_RX_PB_SIZE;
 	wx->mac.tx_pb_size = TXGBE_SP_TDB_PB_SZ;
 
@@ -495,6 +497,8 @@ static const struct net_device_ops txgbe_netdev_ops = {
 	.ndo_validate_addr      = eth_validate_addr,
 	.ndo_set_mac_address    = wx_set_mac,
 	.ndo_get_stats64        = wx_get_stats64,
+	.ndo_vlan_rx_add_vid    = wx_vlan_rx_add_vid,
+	.ndo_vlan_rx_kill_vid   = wx_vlan_rx_kill_vid,
 };
 
 /**
diff --git a/drivers/net/ethernet/wangxun/txgbe/txgbe_type.h b/drivers/net/ethernet/wangxun/txgbe/txgbe_type.h
index 63a1c733718d..032972369965 100644
--- a/drivers/net/ethernet/wangxun/txgbe/txgbe_type.h
+++ b/drivers/net/ethernet/wangxun/txgbe/txgbe_type.h
@@ -77,6 +77,7 @@
 #define TXGBE_SP_MAX_RX_QUEUES  128
 #define TXGBE_SP_RAR_ENTRIES    128
 #define TXGBE_SP_MC_TBL_SIZE    128
+#define TXGBE_SP_VFT_TBL_SIZE   128
 #define TXGBE_SP_RX_PB_SIZE     512
 #define TXGBE_SP_TDB_PB_SZ      (160 * 1024) /* 160KB Packet Buffer */
 
-- 
2.40.1

Re: [RESEND,PATCH net-next v7 0/8] Wangxun netdev features support

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Tue, 30 May 2023 10:26:24 +0800 you wrote:
> Implement tx_csum and rx_csum to support hardware checksum offload.
> Implement ndo_vlan_rx_add_vid and ndo_vlan_rx_kill_vid.
> Implement ndo_set_features.
> Enable macros in netdev features which wangxun can support.
> 
> changes v7:
> - Remove fragmented packets parsing.
> - Jakub Kicinski:
> https://lore.kernel.org/netdev/20230523210454.12963d67@kernel.org/
> changes v6:
> - Fix some code spelling errors.
> changes v5:
> - Add ndo_set_features support.
> - Move wx_decode_ptype() and wx_ptype_lookup to C file.
> - Remove wx_fwd_adapter.
> changes v4:
> - Yunsheng Lin:
> https://lore.kernel.org/netdev/c4b9765d-7213-2718-5de3-5e8231753b95@huawei.com/
> changes v3:
> - Yunsheng Lin: Tidy up logic for wx_encode_tx_desc_ptype.
> changes v2:
> - Andrew Lunn:
> Add ETH_P_CNM Congestion Notification Message to if_ether.h.
> Remove lro support.
> - Yunsheng Lin:
> https://lore.kernel.org/netdev/eb75ae23-8c19-bbc5-e99a-9b853511affa@huawei.com/
> 
> [...]

Here is the summary with links:
  - [RESEND,net-next,v7,1/8] net: wangxun: libwx add tx offload functions
    https://git.kernel.org/netdev/net-next/c/3403960cdf86
  - [RESEND,net-next,v7,2/8] net: wangxun: libwx add rx offload functions
    https://git.kernel.org/netdev/net-next/c/ef4f3c19f912
  - [RESEND,net-next,v7,3/8] net: wangxun: Implement vlan add and kill functions
    https://git.kernel.org/netdev/net-next/c/f3b03c655f67
  - [RESEND,net-next,v7,4/8] net: libwx: Implement xx_set_features ops
    https://git.kernel.org/netdev/net-next/c/6dbedcffcf54
  - [RESEND,net-next,v7,5/8] net: ngbe: Add netdev features support
    https://git.kernel.org/netdev/net-next/c/50a908a0bd8b
  - [RESEND,net-next,v7,6/8] net: ngbe: Implement vlan add and remove ops
    https://git.kernel.org/netdev/net-next/c/361bf4f47cee
  - [RESEND,net-next,v7,7/8] net: txgbe: Add netdev features support
    https://git.kernel.org/netdev/net-next/c/6670f1ece2c8
  - [RESEND,net-next,v7,8/8] net: txgbe: Implement vlan add and remove ops
    https://git.kernel.org/netdev/net-next/c/7df4af51deb3

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html

Re: [RESEND,PATCH net-next v7 7/8] net: txgbe: Add netdev features support

[Paolo Abeni]

Hi,

I just stumbled upon the following while working on something slightly
related. Adding a bunch of people hopefully interested.

On 5/30/23 4:26 AM, Mengyuan Lou wrote:
> Add features and hw_features that ngbe can support.
> 
> Signed-off-by: Mengyuan Lou <mengyuanlou@net-swift.com>
> ---
> @@ -596,11 +597,25 @@ static int txgbe_probe(struct pci_dev *pdev,
>  		goto err_free_mac_table;
>  	}
>  
> -	netdev->features |= NETIF_F_HIGHDMA;
> -	netdev->features = NETIF_F_SG;
> -
> +	netdev->features = NETIF_F_SG |
> +			   NETIF_F_TSO |
> +			   NETIF_F_TSO6 |
> +			   NETIF_F_RXHASH |
> +			   NETIF_F_RXCSUM |
> +			   NETIF_F_HW_CSUM;
> +
> +	netdev->gso_partial_features =  NETIF_F_GSO_ENCAP_ALL;
> +	netdev->features |= netdev->gso_partial_features;
> +	netdev->features |= NETIF_F_SCTP_CRC;
> +	netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
> +	netdev->hw_enc_features |= netdev->vlan_features;

This driver does not implement the .ndo_features_check callback, meaning
it will happily accept TSO_V4 over any IP tunnel, even when ID mangling
is explicitly not allowed.

The above in turn looks inconsistent. If the device is able to update
the (outer) IP (and IP csum) while performing TSO, than it should be
able to fully offload NETIF_F_GSO_GRE: such offload should not be
included in the partial ones.

Otherwise, if the device is not able to perform the mentioned tasks, the
driver should implement a suitable ndo_features_check op, stripping
NETIF_F_TSO for tunneled packet that could be potentially fragmented,
that is, when `features` lacks the `NETIF_F_TSO_MANGLEID` bit.

Alike what several intel drivers (ixgbe, igc, etc.) do.

Assuming I did not misread something relevant, and the above is somewhat
correct, I'm wondering if we should move the mentioned check into the
core; preventing future similar errors and avoiding some driver code
duplication.

Something alike the following, completely untested:
---
diff --git a/net/core/dev.c b/net/core/dev.c
index d5ab9a4b318e..2fdfcddf9c3b 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -3668,15 +3668,6 @@ static netdev_features_t gso_features_check(const
struct sk_buff *skb,
 		return features & ~NETIF_F_GSO_MASK;
 	}

-	/* Support for GSO partial features requires software
-	 * intervention before we can actually process the packets
-	 * so we need to strip support for any partial features now
-	 * and we can pull them back in after we have partially
-	 * segmented the frame.
-	 */
-	if (!(skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL))
-		features &= ~dev->gso_partial_features;
-
 	/* Make sure to clear the IPv4 ID mangling feature if the
 	 * IPv4 header has the potential to be fragmented.
 	 */
@@ -3684,10 +3675,24 @@ static netdev_features_t
gso_features_check(const struct sk_buff *skb,
 		struct iphdr *iph = skb->encapsulation ?
 				    inner_ip_hdr(skb) : ip_hdr(skb);

-		if (!(iph->frag_off & htons(IP_DF)))
+		if (!(iph->frag_off & htons(IP_DF))) {
 			features &= ~NETIF_F_TSO_MANGLEID;
+			if (features & dev->gso_partial_features &
+			    (NETIF_F_GSO_GRE | NETIF_F_GSO_IPXIP4 |
+			     NETIF_F_GSO_IPXIP6 | NETIF_F_GSO_UDP_TUNNEL))
+				features &= ~NETIF_F_TSO;
+		}
 	}

+	/* Support for GSO partial features requires software
+	 * intervention before we can actually process the packets
+	 * so we need to strip support for any partial features now
+	 * and we can pull them back in after we have partially
+	 * segmented the frame.
+	 */
+	if (!(skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL))
+		features &= ~dev->gso_partial_features;
+
 	return features;
 }

Re: [RESEND,PATCH net-next v7 7/8] net: txgbe: Add netdev features support

[Alexander Duyck]

On Mon, Feb 24, 2025 at 2:17 PM Paolo Abeni <pabeni@redhat.com> wrote:
>
> Hi,
>
> I just stumbled upon the following while working on something slightly
> related. Adding a bunch of people hopefully interested.
>
> On 5/30/23 4:26 AM, Mengyuan Lou wrote:
> > Add features and hw_features that ngbe can support.
> >
> > Signed-off-by: Mengyuan Lou <mengyuanlou@net-swift.com>
> > ---
> > @@ -596,11 +597,25 @@ static int txgbe_probe(struct pci_dev *pdev,
> >               goto err_free_mac_table;
> >       }
> >
> > -     netdev->features |= NETIF_F_HIGHDMA;
> > -     netdev->features = NETIF_F_SG;
> > -
> > +     netdev->features = NETIF_F_SG |
> > +                        NETIF_F_TSO |
> > +                        NETIF_F_TSO6 |
> > +                        NETIF_F_RXHASH |
> > +                        NETIF_F_RXCSUM |
> > +                        NETIF_F_HW_CSUM;
> > +
> > +     netdev->gso_partial_features =  NETIF_F_GSO_ENCAP_ALL;
> > +     netdev->features |= netdev->gso_partial_features;
> > +     netdev->features |= NETIF_F_SCTP_CRC;
> > +     netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
> > +     netdev->hw_enc_features |= netdev->vlan_features;
>
> This driver does not implement the .ndo_features_check callback, meaning
> it will happily accept TSO_V4 over any IP tunnel, even when ID mangling
> is explicitly not allowed.
>
> The above in turn looks inconsistent. If the device is able to update
> the (outer) IP (and IP csum) while performing TSO, than it should be
> able to fully offload NETIF_F_GSO_GRE: such offload should not be
> included in the partial ones.
>
> Otherwise, if the device is not able to perform the mentioned tasks, the
> driver should implement a suitable ndo_features_check op, stripping
> NETIF_F_TSO for tunneled packet that could be potentially fragmented,
> that is, when `features` lacks the `NETIF_F_TSO_MANGLEID` bit.
>
> Alike what several intel drivers (ixgbe, igc, etc.) do.
>
> Assuming I did not misread something relevant, and the above is somewhat
> correct, I'm wondering if we should move the mentioned check into the
> core; preventing future similar errors and avoiding some driver code
> duplication.
>
> Something alike the following, completely untested:
> ---
> diff --git a/net/core/dev.c b/net/core/dev.c
> index d5ab9a4b318e..2fdfcddf9c3b 100644
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -3668,15 +3668,6 @@ static netdev_features_t gso_features_check(const
> struct sk_buff *skb,
>                 return features & ~NETIF_F_GSO_MASK;
>         }
>
> -       /* Support for GSO partial features requires software
> -        * intervention before we can actually process the packets
> -        * so we need to strip support for any partial features now
> -        * and we can pull them back in after we have partially
> -        * segmented the frame.
> -        */
> -       if (!(skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL))
> -               features &= ~dev->gso_partial_features;
> -
>         /* Make sure to clear the IPv4 ID mangling feature if the
>          * IPv4 header has the potential to be fragmented.
>          */
> @@ -3684,10 +3675,24 @@ static netdev_features_t
> gso_features_check(const struct sk_buff *skb,
>                 struct iphdr *iph = skb->encapsulation ?
>                                     inner_ip_hdr(skb) : ip_hdr(skb);
>
> -               if (!(iph->frag_off & htons(IP_DF)))
> +               if (!(iph->frag_off & htons(IP_DF))) {
>                         features &= ~NETIF_F_TSO_MANGLEID;
> +                       if (features & dev->gso_partial_features &
> +                           (NETIF_F_GSO_GRE | NETIF_F_GSO_IPXIP4 |
> +                            NETIF_F_GSO_IPXIP6 | NETIF_F_GSO_UDP_TUNNEL))
> +                               features &= ~NETIF_F_TSO;
> +               }
>         }
>

My main concern would be that there might be implementations of
partial that are capable of doing a partial offload without the
MANGLEID feature that would be impacted.

If I recall I think for the i40e it may have supported some logic to
do the proper things while not knowing how to deal with the tunnel
headers but correctly handling the IP headers.

> +       /* Support for GSO partial features requires software
> +        * intervention before we can actually process the packets
> +        * so we need to strip support for any partial features now
> +        * and we can pull them back in after we have partially
> +        * segmented the frame.
> +        */
> +       if (!(skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL))
> +               features &= ~dev->gso_partial_features;
> +

Why move this? I'm not sure it gains you anything if it is either
before or after the result should be the same.

>         return features;
>  }

Re: [RESEND,PATCH net-next v7 7/8] net: txgbe: Add netdev features support

[Paolo Abeni]

Adding Joe, which authored a change possibly relevant here.

On 2/25/25 12:59 AM, Alexander Duyck wrote:
> On Mon, Feb 24, 2025 at 2:17 PM Paolo Abeni <pabeni@redhat.com> wrote:
>> I just stumbled upon the following while working on something slightly
>> related. Adding a bunch of people hopefully interested.
>>
>> On 5/30/23 4:26 AM, Mengyuan Lou wrote:
>>> Add features and hw_features that ngbe can support.
>>>
>>> Signed-off-by: Mengyuan Lou <mengyuanlou@net-swift.com>
>>> ---
>>> @@ -596,11 +597,25 @@ static int txgbe_probe(struct pci_dev *pdev,
>>>               goto err_free_mac_table;
>>>       }
>>>
>>> -     netdev->features |= NETIF_F_HIGHDMA;
>>> -     netdev->features = NETIF_F_SG;
>>> -
>>> +     netdev->features = NETIF_F_SG |
>>> +                        NETIF_F_TSO |
>>> +                        NETIF_F_TSO6 |
>>> +                        NETIF_F_RXHASH |
>>> +                        NETIF_F_RXCSUM |
>>> +                        NETIF_F_HW_CSUM;
>>> +
>>> +     netdev->gso_partial_features =  NETIF_F_GSO_ENCAP_ALL;
>>> +     netdev->features |= netdev->gso_partial_features;
>>> +     netdev->features |= NETIF_F_SCTP_CRC;
>>> +     netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
>>> +     netdev->hw_enc_features |= netdev->vlan_features;
>>
>> This driver does not implement the .ndo_features_check callback, meaning
>> it will happily accept TSO_V4 over any IP tunnel, even when ID mangling
>> is explicitly not allowed.
>>
>> The above in turn looks inconsistent. If the device is able to update
>> the (outer) IP (and IP csum) while performing TSO, than it should be
>> able to fully offload NETIF_F_GSO_GRE: such offload should not be
>> included in the partial ones.
>>
>> Otherwise, if the device is not able to perform the mentioned tasks, the
>> driver should implement a suitable ndo_features_check op, stripping
>> NETIF_F_TSO for tunneled packet that could be potentially fragmented,
>> that is, when `features` lacks the `NETIF_F_TSO_MANGLEID` bit.
>>
>> Alike what several intel drivers (ixgbe, igc, etc.) do.
>>
>> Assuming I did not misread something relevant, and the above is somewhat
>> correct, I'm wondering if we should move the mentioned check into the
>> core; preventing future similar errors and avoiding some driver code
>> duplication.
>>
>> Something alike the following, completely untested:
>> ---
>> diff --git a/net/core/dev.c b/net/core/dev.c
>> index d5ab9a4b318e..2fdfcddf9c3b 100644
>> --- a/net/core/dev.c
>> +++ b/net/core/dev.c
>> @@ -3668,15 +3668,6 @@ static netdev_features_t gso_features_check(const
>> struct sk_buff *skb,
>>                 return features & ~NETIF_F_GSO_MASK;
>>         }
>>
>> -       /* Support for GSO partial features requires software
>> -        * intervention before we can actually process the packets
>> -        * so we need to strip support for any partial features now
>> -        * and we can pull them back in after we have partially
>> -        * segmented the frame.
>> -        */
>> -       if (!(skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL))
>> -               features &= ~dev->gso_partial_features;
>> -
>>         /* Make sure to clear the IPv4 ID mangling feature if the
>>          * IPv4 header has the potential to be fragmented.
>>          */
>> @@ -3684,10 +3675,24 @@ static netdev_features_t
>> gso_features_check(const struct sk_buff *skb,
>>                 struct iphdr *iph = skb->encapsulation ?
>>                                     inner_ip_hdr(skb) : ip_hdr(skb);
>>
>> -               if (!(iph->frag_off & htons(IP_DF)))
>> +               if (!(iph->frag_off & htons(IP_DF))) {
>>                         features &= ~NETIF_F_TSO_MANGLEID;
>> +                       if (features & dev->gso_partial_features &
>> +                           (NETIF_F_GSO_GRE | NETIF_F_GSO_IPXIP4 |
>> +                            NETIF_F_GSO_IPXIP6 | NETIF_F_GSO_UDP_TUNNEL))
>> +                               features &= ~NETIF_F_TSO;
>> +               }
>>         }
>>
> 
> My main concern would be that there might be implementations of
> partial that are capable of doing a partial offload without the
> MANGLEID feature that would be impacted.
> 
> If I recall I think for the i40e it may have supported some logic to
> do the proper things while not knowing how to deal with the tunnel
> headers but correctly handling the IP headers.

Thanks for the pointer, I missed the i40e case. It's not clear to me how
such driver handles the special case.

If I understand correctly, such NIC is able to touch inner and outer IP
id and csum correctly, without being able to parse the encap header. If
so, I guess it needs explicit notion of the inner and outer header
offsets, provided by the driver.

But I don't see the driver passing to the H/W the offsets, so possibly
the current code is actually bugged for this specific case? What am I
missing here?

If i40e can handle the IP headers correctly, why it need the partial
feature for NETIF_F_GSO_GRE? Perhaps commit b4fb2d33514 ("i40e: Add
support for MPLS + TSO") increased the partial features set too
broadly?!? It looks like originally none of
NETIF_F_GSO_UDP_TUNNEL, NETIF_F_GSO_GRE, NETIF_F_GSO_IPXIP4,
NETIF_F_GSO_IPXIP6
belonged to the partial features set.

>> +       /* Support for GSO partial features requires software
>> +        * intervention before we can actually process the packets
>> +        * so we need to strip support for any partial features now
>> +        * and we can pull them back in after we have partially
>> +        * segmented the frame.
>> +        */
>> +       if (!(skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL))
>> +               features &= ~dev->gso_partial_features;
>> +
> 
> Why move this? I'm not sure it gains you anything if it is either
> before or after the result should be the same.

I'm sorry, I rushed the code out-of-the-door to try to get some
feedback; I moved that chunk so that in the previous block `features`
still contains the partials bits and the code can check for tunnels
looking at `features` contents.

Thinking again about it, it's not needed, as we could (and should) check
instead the skb wanted features (skb_shinfo(skb)->gso_type <<
NETIF_F_GSO_SHIFT).

Thanks,

Paolo