cxgb4: add tc flower match support for tunnel VNI

Adds support for matching flows based on tunnel VNI value.
Introduces fw APIs for allocating/removing MPS entries related
to encapsulation. And uses the same while adding/deleting filters
for offloading flows based on tunnel VNI match.
Signed-off-by: NKumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: NGanesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

drivers/net/ethernet/chelsio/cxgb4/cxgb4.h

@@ -1038,6 +1038,7 @@ struct ch_sched_queue {
 #define VF_BITWIDTH 8
 #define IVLAN_BITWIDTH 16
 #define OVLAN_BITWIDTH 16
+#define ENCAP_VNI_BITWIDTH 24
 
 /* Filter matching rules.  These consist of a set of ingress packet field
  * (value, mask) tuples.  The associated ingress packet field matches the
@@ -1068,6 +1069,7 @@ struct ch_filter_tuple {
 	uint32_t ivlan_vld:1;                   /* inner VLAN valid */
 	uint32_t ovlan_vld:1;                   /* outer VLAN valid */
 	uint32_t pfvf_vld:1;                    /* PF/VF valid */
+	uint32_t encap_vld:1;			/* Encapsulation valid */
 	uint32_t macidx:MACIDX_BITWIDTH;        /* exact match MAC index */
 	uint32_t fcoe:FCOE_BITWIDTH;            /* FCoE packet */
 	uint32_t iport:IPORT_BITWIDTH;          /* ingress port */
@@ -1078,6 +1080,7 @@ struct ch_filter_tuple {
 	uint32_t vf:VF_BITWIDTH;                /* PCI-E VF ID */
 	uint32_t ivlan:IVLAN_BITWIDTH;          /* inner VLAN */
 	uint32_t ovlan:OVLAN_BITWIDTH;          /* outer VLAN */
+	uint32_t vni:ENCAP_VNI_BITWIDTH;	/* VNI of tunnel */
 
 	/* Uncompressed header matching field rules.  These are always
 	 * available for field rules.
@@ -1694,6 +1697,12 @@ int t4_set_rxmode(struct adapter *adap, unsigned int mbox, unsigned int viid,
 int t4_free_raw_mac_filt(struct adapter *adap, unsigned int viid,
 			 const u8 *addr, const u8 *mask, unsigned int idx,
 			 u8 lookup_type, u8 port_id, bool sleep_ok);
+int t4_free_encap_mac_filt(struct adapter *adap, unsigned int viid, int idx,
+			   bool sleep_ok);
+int t4_alloc_encap_mac_filt(struct adapter *adap, unsigned int viid,
+			    const u8 *addr, const u8 *mask, unsigned int vni,
+			    unsigned int vni_mask, u8 dip_hit, u8 lookup_type,
+			    bool sleep_ok);
 int t4_alloc_raw_mac_filt(struct adapter *adap, unsigned int viid,
 			  const u8 *addr, const u8 *mask, unsigned int idx,
 			  u8 lookup_type, u8 port_id, bool sleep_ok);

drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c

@@ -265,6 +265,8 @@ static int validate_filter(struct net_device *dev,
 			fs->mask.pfvf_vld) ||
 	    unsupported(fconf, VNIC_ID_F, fs->val.ovlan_vld,
 			fs->mask.ovlan_vld) ||
+	    unsupported(fconf, VNIC_ID_F, fs->val.encap_vld,
+			fs->mask.encap_vld) ||
 	    unsupported(fconf, VLAN_F, fs->val.ivlan_vld, fs->mask.ivlan_vld))
 		return -EOPNOTSUPP;
 
@@ -275,8 +277,12 @@ static int validate_filter(struct net_device *dev,
 	 * carries that overlap, we need to translate any PF/VF
 	 * specification into that internal format below.
 	 */
-	if (is_field_set(fs->val.pfvf_vld, fs->mask.pfvf_vld) &&
-	    is_field_set(fs->val.ovlan_vld, fs->mask.ovlan_vld))
+	if ((is_field_set(fs->val.pfvf_vld, fs->mask.pfvf_vld) &&
+	     is_field_set(fs->val.ovlan_vld, fs->mask.ovlan_vld)) ||
+	    (is_field_set(fs->val.pfvf_vld, fs->mask.pfvf_vld) &&
+	     is_field_set(fs->val.encap_vld, fs->mask.encap_vld)) ||
+	    (is_field_set(fs->val.ovlan_vld, fs->mask.ovlan_vld) &&
+	     is_field_set(fs->val.encap_vld, fs->mask.encap_vld)))
 		return -EOPNOTSUPP;
 	if (unsupported(iconf, VNIC_F, fs->val.pfvf_vld, fs->mask.pfvf_vld) ||
 	    (is_field_set(fs->val.ovlan_vld, fs->mask.ovlan_vld) &&
@@ -306,6 +312,9 @@ static int validate_filter(struct net_device *dev,
 	     fs->newvlan == VLAN_REWRITE))
 		return -EOPNOTSUPP;
 
+	if (fs->val.encap_vld &&
+	    CHELSIO_CHIP_VERSION(adapter->params.chip) < CHELSIO_T6)
+		return -EOPNOTSUPP;
 	return 0;
 }
 
@@ -705,6 +714,8 @@ int delete_filter(struct adapter *adapter, unsigned int fidx)
  */
 void clear_filter(struct adapter *adap, struct filter_entry *f)
 {
+	struct port_info *pi = netdev_priv(f->dev);
+
 	/* If the new or old filter have loopback rewriteing rules then we'll
 	 * need to free any existing L2T, SMT, CLIP entries of filter
 	 * rule.
@@ -715,6 +726,12 @@ void clear_filter(struct adapter *adap, struct filter_entry *f)
 	if (f->smt)
 		cxgb4_smt_release(f->smt);
 
+	if (f->fs.val.encap_vld && f->fs.val.ovlan_vld)
+		if (atomic_dec_and_test(&adap->mps_encap[f->fs.val.ovlan &
+							 0x1ff].refcnt))
+			t4_free_encap_mac_filt(adap, pi->viid,
+					       f->fs.val.ovlan & 0x1ff, 0);
+
 	if ((f->fs.hash || is_t6(adap->params.chip)) && f->fs.type)
 		cxgb4_clip_release(f->dev, (const u32 *)&f->fs.val.lip, 1);
 
@@ -840,6 +857,10 @@ bool is_filter_exact_match(struct adapter *adap,
 	if (!is_hashfilter(adap))
 		return false;
 
+	 /* Keep tunnel VNI match disabled for hash-filters for now */
+	if (fs->mask.encap_vld)
+		return false;
+
 	if (fs->type) {
 		if (is_inaddr_any(fs->val.fip, AF_INET6) ||
 		    !is_addr_all_mask(fs->mask.fip, AF_INET6))
@@ -961,8 +982,12 @@ static u64 hash_filter_ntuple(struct ch_filter_specification *fs,
 		ntuple |= (u64)(fs->val.tos) << tp->tos_shift;
 
 	if (tp->vnic_shift >= 0) {
-		if ((adap->params.tp.ingress_config & VNIC_F) &&
-		    fs->mask.pfvf_vld)
+		if ((adap->params.tp.ingress_config & USE_ENC_IDX_F) &&
+		    fs->mask.encap_vld)
+			ntuple |= (u64)((fs->val.encap_vld << 16) |
+					(fs->val.ovlan)) << tp->vnic_shift;
+		else if ((adap->params.tp.ingress_config & VNIC_F) &&
+			 fs->mask.pfvf_vld)
 			ntuple |= (u64)((fs->val.pfvf_vld << 16) |
 					(fs->val.pf << 13) |
 					(fs->val.vf)) << tp->vnic_shift;
@@ -1076,6 +1101,7 @@ static int cxgb4_set_hash_filter(struct net_device *dev,
 				 struct filter_ctx *ctx)
 {
 	struct adapter *adapter = netdev2adap(dev);
+	struct port_info *pi = netdev_priv(dev);
 	struct tid_info *t = &adapter->tids;
 	struct filter_entry *f;
 	struct sk_buff *skb;
@@ -1142,13 +1168,34 @@ static int cxgb4_set_hash_filter(struct net_device *dev,
 		f->fs.mask.ovlan = (fs->mask.pf << 13) | fs->mask.vf;
 		f->fs.val.ovlan_vld = fs->val.pfvf_vld;
 		f->fs.mask.ovlan_vld = fs->mask.pfvf_vld;
+	} else if (iconf & USE_ENC_IDX_F) {
+		if (f->fs.val.encap_vld) {
+			struct port_info *pi = netdev_priv(f->dev);
+			u8 match_all_mac[] = { 0, 0, 0, 0, 0, 0 };
+
+			/* allocate MPS TCAM entry */
+			ret = t4_alloc_encap_mac_filt(adapter, pi->viid,
+						      match_all_mac,
+						      match_all_mac,
+						      f->fs.val.vni,
+						      f->fs.mask.vni,
+						      0, 1, 1);
+			if (ret < 0)
+				goto free_atid;
+
+			atomic_inc(&adapter->mps_encap[ret].refcnt);
+			f->fs.val.ovlan = ret;
+			f->fs.mask.ovlan = 0xffff;
+			f->fs.val.ovlan_vld = 1;
+			f->fs.mask.ovlan_vld = 1;
+		}
 	}
 
 	size = sizeof(struct cpl_t6_act_open_req);
 	if (f->fs.type) {
 		ret = cxgb4_clip_get(f->dev, (const u32 *)&f->fs.val.lip, 1);
 		if (ret)
-			goto free_atid;
+			goto free_mps;
 
 		skb = alloc_skb(size, GFP_KERNEL);
 		if (!skb) {
@@ -1163,7 +1210,7 @@ static int cxgb4_set_hash_filter(struct net_device *dev,
 		skb = alloc_skb(size, GFP_KERNEL);
 		if (!skb) {
 			ret = -ENOMEM;
-			goto free_atid;
+			goto free_mps;
 		}
 
 		mk_act_open_req(f, skb,
@@ -1179,6 +1226,10 @@ static int cxgb4_set_hash_filter(struct net_device *dev,
 free_clip:
 	cxgb4_clip_release(f->dev, (const u32 *)&f->fs.val.lip, 1);
 
+free_mps:
+	if (f->fs.val.encap_vld && f->fs.val.ovlan_vld)
+		t4_free_encap_mac_filt(adapter, pi->viid, f->fs.val.ovlan, 1);
+
 free_atid:
 	cxgb4_free_atid(t, atid);
 
@@ -1360,6 +1411,27 @@ int __cxgb4_set_filter(struct net_device *dev, int filter_id,
 		f->fs.mask.ovlan = (fs->mask.pf << 13) | fs->mask.vf;
 		f->fs.val.ovlan_vld = fs->val.pfvf_vld;
 		f->fs.mask.ovlan_vld = fs->mask.pfvf_vld;
+	} else if (iconf & USE_ENC_IDX_F) {
+		if (f->fs.val.encap_vld) {
+			struct port_info *pi = netdev_priv(f->dev);
+			u8 match_all_mac[] = { 0, 0, 0, 0, 0, 0 };
+
+			/* allocate MPS TCAM entry */
+			ret = t4_alloc_encap_mac_filt(adapter, pi->viid,
+						      match_all_mac,
+						      match_all_mac,
+						      f->fs.val.vni,
+						      f->fs.mask.vni,
+						      0, 1, 1);
+			if (ret < 0)
+				goto free_clip;
+
+			atomic_inc(&adapter->mps_encap[ret].refcnt);
+			f->fs.val.ovlan = ret;
+			f->fs.mask.ovlan = 0x1ff;
+			f->fs.val.ovlan_vld = 1;
+			f->fs.mask.ovlan_vld = 1;
+		}
 	}
 
 	/* Attempt to set the filter.  If we don't succeed, we clear
@@ -1376,6 +1448,13 @@ int __cxgb4_set_filter(struct net_device *dev, int filter_id,
 	}
 
 	return ret;
+
+free_clip:
+	if (is_t6(adapter->params.chip) && f->fs.type)
+		cxgb4_clip_release(f->dev, (const u32 *)&f->fs.val.lip, 1);
+	cxgb4_clear_ftid(&adapter->tids, filter_id,
+			 fs->type ? PF_INET6 : PF_INET, chip_ver);
+	return ret;
 }
 
 static int cxgb4_del_hash_filter(struct net_device *dev, int filter_id,

drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c

@@ -194,6 +194,23 @@ static void cxgb4_process_flow_match(struct net_device *dev,
 		fs->mask.tos = mask->tos;
 	}
 
+	if (dissector_uses_key(cls->dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
+		struct flow_dissector_key_keyid *key, *mask;
+
+		key = skb_flow_dissector_target(cls->dissector,
+						FLOW_DISSECTOR_KEY_ENC_KEYID,
+						cls->key);
+		mask = skb_flow_dissector_target(cls->dissector,
+						 FLOW_DISSECTOR_KEY_ENC_KEYID,
+						 cls->mask);
+		fs->val.vni = be32_to_cpu(key->keyid);
+		fs->mask.vni = be32_to_cpu(mask->keyid);
+		if (fs->mask.vni) {
+			fs->val.encap_vld = 1;
+			fs->mask.encap_vld = 1;
+		}
+	}
+
 	if (dissector_uses_key(cls->dissector, FLOW_DISSECTOR_KEY_VLAN)) {
 		struct flow_dissector_key_vlan *key, *mask;
 		u16 vlan_tci, vlan_tci_mask;
@@ -247,6 +264,7 @@ static int cxgb4_validate_flow_match(struct net_device *dev,
 	      BIT(FLOW_DISSECTOR_KEY_IPV4_ADDRS) |
 	      BIT(FLOW_DISSECTOR_KEY_IPV6_ADDRS) |
 	      BIT(FLOW_DISSECTOR_KEY_PORTS) |
+	      BIT(FLOW_DISSECTOR_KEY_ENC_KEYID) |
 	      BIT(FLOW_DISSECTOR_KEY_VLAN) |
 	      BIT(FLOW_DISSECTOR_KEY_IP))) {
 		netdev_warn(dev, "Unsupported key used: 0x%x\n",

drivers/net/ethernet/chelsio/cxgb4/l2t.c

@@ -491,7 +491,7 @@ u64 cxgb4_select_ntuple(struct net_device *dev,
 	if (tp->protocol_shift >= 0)
 		ntuple |= (u64)IPPROTO_TCP << tp->protocol_shift;
 
-	if (tp->vnic_shift >= 0) {
+	if (tp->vnic_shift >= 0 && (tp->ingress_config & VNIC_F)) {
 		u32 viid = cxgb4_port_viid(dev);
 		u32 vf = FW_VIID_VIN_G(viid);
 		u32 pf = FW_VIID_PFN_G(viid);

drivers/net/ethernet/chelsio/cxgb4/t4_hw.c

@@ -7512,6 +7512,43 @@ int t4_set_rxmode(struct adapter *adap, unsigned int mbox, unsigned int viid,
 	return t4_wr_mbox_meat(adap, mbox, &c, sizeof(c), NULL, sleep_ok);
 }
 
+/**
+ *      t4_free_encap_mac_filt - frees MPS entry at given index
+ *      @adap: the adapter
+ *      @viid: the VI id
+ *      @idx: index of MPS entry to be freed
+ *      @sleep_ok: call is allowed to sleep
+ *
+ *      Frees the MPS entry at supplied index
+ *
+ *      Returns a negative error number or zero on success
+ */
+int t4_free_encap_mac_filt(struct adapter *adap, unsigned int viid,
+			   int idx, bool sleep_ok)
+{
+	struct fw_vi_mac_exact *p;
+	u8 addr[] = {0, 0, 0, 0, 0, 0};
+	struct fw_vi_mac_cmd c;
+	int ret = 0;
+	u32 exact;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_viid = cpu_to_be32(FW_CMD_OP_V(FW_VI_MAC_CMD) |
+				   FW_CMD_REQUEST_F | FW_CMD_WRITE_F |
+				   FW_CMD_EXEC_V(0) |
+				   FW_VI_MAC_CMD_VIID_V(viid));
+	exact = FW_VI_MAC_CMD_ENTRY_TYPE_V(FW_VI_MAC_TYPE_EXACTMAC);
+	c.freemacs_to_len16 = cpu_to_be32(FW_VI_MAC_CMD_FREEMACS_V(0) |
+					  exact |
+					  FW_CMD_LEN16_V(1));
+	p = c.u.exact;
+	p->valid_to_idx = cpu_to_be16(FW_VI_MAC_CMD_VALID_F |
+				      FW_VI_MAC_CMD_IDX_V(idx));
+	memcpy(p->macaddr, addr, sizeof(p->macaddr));
+	ret = t4_wr_mbox_meat(adap, adap->mbox, &c, sizeof(c), &c, sleep_ok);
+	return ret;
+}
+
 /**
  *	t4_free_raw_mac_filt - Frees a raw mac entry in mps tcam
  *	@adap: the adapter
@@ -7562,6 +7599,55 @@ int t4_free_raw_mac_filt(struct adapter *adap, unsigned int viid,
 	return t4_wr_mbox_meat(adap, adap->mbox, &c, sizeof(c), &c, sleep_ok);
 }
 
+/**
+ *      t4_alloc_encap_mac_filt - Adds a mac entry in mps tcam with VNI support
+ *      @adap: the adapter
+ *      @viid: the VI id
+ *      @mac: the MAC address
+ *      @mask: the mask
+ *      @vni: the VNI id for the tunnel protocol
+ *      @vni_mask: mask for the VNI id
+ *      @dip_hit: to enable DIP match for the MPS entry
+ *      @lookup_type: MAC address for inner (1) or outer (0) header
+ *      @sleep_ok: call is allowed to sleep
+ *
+ *      Allocates an MPS entry with specified MAC address and VNI value.
+ *
+ *      Returns a negative error number or the allocated index for this mac.
+ */
+int t4_alloc_encap_mac_filt(struct adapter *adap, unsigned int viid,
+			    const u8 *addr, const u8 *mask, unsigned int vni,
+			    unsigned int vni_mask, u8 dip_hit, u8 lookup_type,
+			    bool sleep_ok)
+{
+	struct fw_vi_mac_cmd c;
+	struct fw_vi_mac_vni *p = c.u.exact_vni;
+	int ret = 0;
+	u32 val;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_viid = cpu_to_be32(FW_CMD_OP_V(FW_VI_MAC_CMD) |
+				   FW_CMD_REQUEST_F | FW_CMD_WRITE_F |
+				   FW_VI_MAC_CMD_VIID_V(viid));
+	val = FW_CMD_LEN16_V(1) |
+	      FW_VI_MAC_CMD_ENTRY_TYPE_V(FW_VI_MAC_TYPE_EXACTMAC_VNI);
+	c.freemacs_to_len16 = cpu_to_be32(val);
+	p->valid_to_idx = cpu_to_be16(FW_VI_MAC_CMD_VALID_F |
+				      FW_VI_MAC_CMD_IDX_V(FW_VI_MAC_ADD_MAC));
+	memcpy(p->macaddr, addr, sizeof(p->macaddr));
+	memcpy(p->macaddr_mask, mask, sizeof(p->macaddr_mask));
+
+	p->lookup_type_to_vni =
+		cpu_to_be32(FW_VI_MAC_CMD_VNI_V(vni) |
+			    FW_VI_MAC_CMD_DIP_HIT_V(dip_hit) |
+			    FW_VI_MAC_CMD_LOOKUP_TYPE_V(lookup_type));
+	p->vni_mask_pkd = cpu_to_be32(FW_VI_MAC_CMD_VNI_MASK_V(vni_mask));
+	ret = t4_wr_mbox_meat(adap, adap->mbox, &c, sizeof(c), &c, sleep_ok);
+	if (ret == 0)
+		ret = FW_VI_MAC_CMD_IDX_G(be16_to_cpu(p->valid_to_idx));
+	return ret;
+}
+
 /**
  *	t4_alloc_raw_mac_filt - Adds a mac entry in mps tcam
  *	@adap: the adapter

drivers/net/ethernet/chelsio/cxgb4/t4_regs.h

@@ -1598,6 +1598,10 @@
 #define VNIC_V(x) ((x) << VNIC_S)
 #define VNIC_F    VNIC_V(1U)
 
+#define USE_ENC_IDX_S		13
+#define USE_ENC_IDX_V(x)	((x) << USE_ENC_IDX_S)
+#define USE_ENC_IDX_F		USE_ENC_IDX_V(1U)
+
 #define CSUM_HAS_PSEUDO_HDR_S    10
 #define CSUM_HAS_PSEUDO_HDR_V(x) ((x) << CSUM_HAS_PSEUDO_HDR_S)
 #define CSUM_HAS_PSEUDO_HDR_F    CSUM_HAS_PSEUDO_HDR_V(1U)

drivers/net/ethernet/chelsio/cxgb4/t4fw_api.h

@@ -2158,6 +2158,14 @@ struct fw_vi_mac_cmd {
 			__be64 data0m_pkd;
 			__be32 data1m[2];
 		} raw;
+		struct fw_vi_mac_vni {
+			__be16 valid_to_idx;
+			__u8 macaddr[6];
+			__be16 r7;
+			__u8 macaddr_mask[6];
+			__be32 lookup_type_to_vni;
+			__be32 vni_mask_pkd;
+		} exact_vni[2];
 	} u;
 };
 
@@ -2205,6 +2213,32 @@ struct fw_vi_mac_cmd {
 #define FW_VI_MAC_CMD_RAW_IDX_G(x)      \
 	(((x) >> FW_VI_MAC_CMD_RAW_IDX_S) & FW_VI_MAC_CMD_RAW_IDX_M)
 
+#define FW_VI_MAC_CMD_LOOKUP_TYPE_S	31
+#define FW_VI_MAC_CMD_LOOKUP_TYPE_M	0x1
+#define FW_VI_MAC_CMD_LOOKUP_TYPE_V(x)	((x) << FW_VI_MAC_CMD_LOOKUP_TYPE_S)
+#define FW_VI_MAC_CMD_LOOKUP_TYPE_G(x)	\
+	(((x) >> FW_VI_MAC_CMD_LOOKUP_TYPE_S) & FW_VI_MAC_CMD_LOOKUP_TYPE_M)
+#define FW_VI_MAC_CMD_LOOKUP_TYPE_F	FW_VI_MAC_CMD_LOOKUP_TYPE_V(1U)
+
+#define FW_VI_MAC_CMD_DIP_HIT_S		30
+#define FW_VI_MAC_CMD_DIP_HIT_M		0x1
+#define FW_VI_MAC_CMD_DIP_HIT_V(x)	((x) << FW_VI_MAC_CMD_DIP_HIT_S)
+#define FW_VI_MAC_CMD_DIP_HIT_G(x)	\
+	(((x) >> FW_VI_MAC_CMD_DIP_HIT_S) & FW_VI_MAC_CMD_DIP_HIT_M)
+#define FW_VI_MAC_CMD_DIP_HIT_F		FW_VI_MAC_CMD_DIP_HIT_V(1U)
+
+#define FW_VI_MAC_CMD_VNI_S		0
+#define FW_VI_MAC_CMD_VNI_M		0xffffff
+#define FW_VI_MAC_CMD_VNI_V(x)		((x) << FW_VI_MAC_CMD_VNI_S)
+#define FW_VI_MAC_CMD_VNI_G(x)		\
+	(((x) >> FW_VI_MAC_CMD_VNI_S) & FW_VI_MAC_CMD_VNI_M)
+
+#define FW_VI_MAC_CMD_VNI_MASK_S	0
+#define FW_VI_MAC_CMD_VNI_MASK_M	0xffffff
+#define FW_VI_MAC_CMD_VNI_MASK_V(x)	((x) << FW_VI_MAC_CMD_VNI_MASK_S)
+#define FW_VI_MAC_CMD_VNI_MASK_G(x)	\
+	(((x) >> FW_VI_MAC_CMD_VNI_MASK_S) & FW_VI_MAC_CMD_VNI_MASK_M)
+
 #define FW_RXMODE_MTU_NO_CHG	65535
 
 struct fw_vi_rxmode_cmd {