nfp: offload flow steering to the nfp

This is the second part to implement flow steering. Mailbox is used
for the communication between driver and HW.

Signed-off-by: Yinjun Zhang <yinjun.zhang@corigine.com>
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20231117071114.10667-3-louis.peens@corigine.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Author: Yinjun Zhang

drivers/net/ethernet/netronome/nfp/nfp_net_common.c

@@ -1763,14 +1763,173 @@ nfp_net_vlan_rx_kill_vid(struct net_device *netdev, __be16 proto, u16 vid)
 	return nfp_net_mbox_reconfig_and_unlock(nn, cmd);
 }
 
+static void
+nfp_net_fs_fill_v4(struct nfp_net *nn, struct nfp_fs_entry *entry, u32 op, u32 *addr)
+{
+	unsigned int i;
+
+	union {
+		struct {
+			__be16 loc;
+			u8 k_proto, m_proto;
+			__be32 k_sip, m_sip, k_dip, m_dip;
+			__be16 k_sport, m_sport, k_dport, m_dport;
+		};
+		__be32 val[7];
+	} v4_rule;
+
+	nn_writel(nn, *addr, op);
+	*addr += sizeof(u32);
+
+	v4_rule.loc     = cpu_to_be16(entry->loc);
+	v4_rule.k_proto = entry->key.l4_proto;
+	v4_rule.m_proto = entry->msk.l4_proto;
+	v4_rule.k_sip   = entry->key.sip4;
+	v4_rule.m_sip   = entry->msk.sip4;
+	v4_rule.k_dip   = entry->key.dip4;
+	v4_rule.m_dip   = entry->msk.dip4;
+	v4_rule.k_sport = entry->key.sport;
+	v4_rule.m_sport = entry->msk.sport;
+	v4_rule.k_dport = entry->key.dport;
+	v4_rule.m_dport = entry->msk.dport;
+
+	for (i = 0; i < ARRAY_SIZE(v4_rule.val); i++, *addr += sizeof(__be32))
+		nn_writel(nn, *addr, be32_to_cpu(v4_rule.val[i]));
+}
+
+static void
+nfp_net_fs_fill_v6(struct nfp_net *nn, struct nfp_fs_entry *entry, u32 op, u32 *addr)
+{
+	unsigned int i;
+
+	union {
+		struct {
+			__be16 loc;
+			u8 k_proto, m_proto;
+			__be32 k_sip[4], m_sip[4], k_dip[4], m_dip[4];
+			__be16 k_sport, m_sport, k_dport, m_dport;
+		};
+		__be32 val[19];
+	} v6_rule;
+
+	nn_writel(nn, *addr, op);
+	*addr += sizeof(u32);
+
+	v6_rule.loc     = cpu_to_be16(entry->loc);
+	v6_rule.k_proto = entry->key.l4_proto;
+	v6_rule.m_proto = entry->msk.l4_proto;
+	for (i = 0; i < 4; i++) {
+		v6_rule.k_sip[i] = entry->key.sip6[i];
+		v6_rule.m_sip[i] = entry->msk.sip6[i];
+		v6_rule.k_dip[i] = entry->key.dip6[i];
+		v6_rule.m_dip[i] = entry->msk.dip6[i];
+	}
+	v6_rule.k_sport = entry->key.sport;
+	v6_rule.m_sport = entry->msk.sport;
+	v6_rule.k_dport = entry->key.dport;
+	v6_rule.m_dport = entry->msk.dport;
+
+	for (i = 0; i < ARRAY_SIZE(v6_rule.val); i++, *addr += sizeof(__be32))
+		nn_writel(nn, *addr, be32_to_cpu(v6_rule.val[i]));
+}
+
+#define NFP_FS_QUEUE_ID	GENMASK(22, 16)
+#define NFP_FS_ACT	GENMASK(15, 0)
+#define NFP_FS_ACT_DROP	BIT(0)
+#define NFP_FS_ACT_Q	BIT(1)
+static void
+nfp_net_fs_fill_act(struct nfp_net *nn, struct nfp_fs_entry *entry, u32 addr)
+{
+	u32 action = 0; /* 0 means default passthrough */
+
+	if (entry->action == RX_CLS_FLOW_DISC)
+		action = NFP_FS_ACT_DROP;
+	else if (!(entry->flow_type & FLOW_RSS))
+		action = FIELD_PREP(NFP_FS_QUEUE_ID, entry->action) | NFP_FS_ACT_Q;
+
+	nn_writel(nn, addr, action);
+}
+
 int nfp_net_fs_add_hw(struct nfp_net *nn, struct nfp_fs_entry *entry)
 {
-	return -EOPNOTSUPP;
+	u32 addr = nn->tlv_caps.mbox_off + NFP_NET_CFG_MBOX_SIMPLE_VAL;
+	int err;
+
+	err = nfp_net_mbox_lock(nn, NFP_NET_CFG_FS_SZ);
+	if (err)
+		return err;
+
+	switch (entry->flow_type & ~FLOW_RSS) {
+	case TCP_V4_FLOW:
+	case UDP_V4_FLOW:
+	case SCTP_V4_FLOW:
+	case IPV4_USER_FLOW:
+		nfp_net_fs_fill_v4(nn, entry, NFP_NET_CFG_MBOX_CMD_FS_ADD_V4, &addr);
+		break;
+	case TCP_V6_FLOW:
+	case UDP_V6_FLOW:
+	case SCTP_V6_FLOW:
+	case IPV6_USER_FLOW:
+		nfp_net_fs_fill_v6(nn, entry, NFP_NET_CFG_MBOX_CMD_FS_ADD_V6, &addr);
+		break;
+	case ETHER_FLOW:
+		nn_writel(nn, addr, NFP_NET_CFG_MBOX_CMD_FS_ADD_ETHTYPE);
+		addr += sizeof(u32);
+		nn_writew(nn, addr, be16_to_cpu(entry->key.l3_proto));
+		addr += sizeof(u32);
+		break;
+	}
+
+	nfp_net_fs_fill_act(nn, entry, addr);
+
+	err = nfp_net_mbox_reconfig_and_unlock(nn, NFP_NET_CFG_MBOX_CMD_FLOW_STEER);
+	if (err) {
+		nn_err(nn, "Add new fs rule failed with %d\n", err);
+		return -EIO;
+	}
+
+	return 0;
 }
 
 int nfp_net_fs_del_hw(struct nfp_net *nn, struct nfp_fs_entry *entry)
 {
-	return -EOPNOTSUPP;
+	u32 addr = nn->tlv_caps.mbox_off + NFP_NET_CFG_MBOX_SIMPLE_VAL;
+	int err;
+
+	err = nfp_net_mbox_lock(nn, NFP_NET_CFG_FS_SZ);
+	if (err)
+		return err;
+
+	switch (entry->flow_type & ~FLOW_RSS) {
+	case TCP_V4_FLOW:
+	case UDP_V4_FLOW:
+	case SCTP_V4_FLOW:
+	case IPV4_USER_FLOW:
+		nfp_net_fs_fill_v4(nn, entry, NFP_NET_CFG_MBOX_CMD_FS_DEL_V4, &addr);
+		break;
+	case TCP_V6_FLOW:
+	case UDP_V6_FLOW:
+	case SCTP_V6_FLOW:
+	case IPV6_USER_FLOW:
+		nfp_net_fs_fill_v6(nn, entry, NFP_NET_CFG_MBOX_CMD_FS_DEL_V6, &addr);
+		break;
+	case ETHER_FLOW:
+		nn_writel(nn, addr, NFP_NET_CFG_MBOX_CMD_FS_DEL_ETHTYPE);
+		addr += sizeof(u32);
+		nn_writew(nn, addr, be16_to_cpu(entry->key.l3_proto));
+		addr += sizeof(u32);
+		break;
+	}
+
+	nfp_net_fs_fill_act(nn, entry, addr);
+
+	err = nfp_net_mbox_reconfig_and_unlock(nn, NFP_NET_CFG_MBOX_CMD_FLOW_STEER);
+	if (err) {
+		nn_err(nn, "Delete fs rule failed with %d\n", err);
+		return -EIO;
+	}
+
+	return 0;
 }
 
 static void nfp_net_fs_clean(struct nfp_net *nn)

drivers/net/ethernet/netronome/nfp/nfp_net_ctrl.h

@@ -419,6 +419,8 @@
 #define NFP_NET_CFG_MBOX_CMD_MULTICAST_ADD		8
 #define NFP_NET_CFG_MBOX_CMD_MULTICAST_DEL		9
 
+#define NFP_NET_CFG_MBOX_CMD_FLOW_STEER			10
+
 /* VLAN filtering using general use mailbox
  * %NFP_NET_CFG_VLAN_FILTER:		Base address of VLAN filter mailbox
  * %NFP_NET_CFG_VLAN_FILTER_VID:	VLAN ID to filter
@@ -441,6 +443,18 @@
 #define NFP_NET_CFG_MULTICAST_MAC_LO	(NFP_NET_CFG_MULTICAST + 6)
 #define NFP_NET_CFG_MULTICAST_SZ	0x0006
 
+/* Max size of FS rules in bytes */
+#define NFP_NET_CFG_FS_SZ		0x0054
+/* Sub commands for FS */
+enum {
+	NFP_NET_CFG_MBOX_CMD_FS_ADD_V4,
+	NFP_NET_CFG_MBOX_CMD_FS_DEL_V4,
+	NFP_NET_CFG_MBOX_CMD_FS_ADD_V6,
+	NFP_NET_CFG_MBOX_CMD_FS_DEL_V6,
+	NFP_NET_CFG_MBOX_CMD_FS_ADD_ETHTYPE,
+	NFP_NET_CFG_MBOX_CMD_FS_DEL_ETHTYPE,
+};
+
 /* TLV capabilities
  * %NFP_NET_CFG_TLV_TYPE:	Offset of type within the TLV
  * %NFP_NET_CFG_TLV_TYPE_REQUIRED: Driver must be able to parse the TLV