Merge branch 'mlxsw-dip-sip-mangling'

Ido Schimmel says:

====================
mlxsw: Add SIP and DIP mangling support

Danielle says:

On Spectrum-2 onwards, it is possible to overwrite SIP and DIP address
of an IPv4 or IPv6 packet in the ACL engine. That corresponds to pedit
munges of, respectively, ip src and ip dst fields, and likewise for ip6.
Offload these munges on the systems where they are supported.

Patchset overview:
Patch #1: introduces SIP_DIP_ACTION and its fields.
Patch #2-#3: adds the new pedit fields, and dispatches on them on
Spectrum-2 and above.
Patch #4 adds a selftest.
====================

Signed-off-by: David S. Miller <[e-mail]>

drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.c

···
 }
 EXPORT_SYMBOL(mlxsw_afa_block_append_mcrouter);
 
+/* SIP DIP Action
+ * --------------
+ * The SIP_DIP_ACTION is used for modifying the SIP and DIP fields of the
+ * packet, e.g. for NAT. The L3 checksum is updated. Also, if the L4 is TCP or
+ * if the L4 is UDP and the checksum field is not zero, then the L4 checksum is
+ * updated.
+ */
+
+#define MLXSW_AFA_IP_CODE 0x11
+#define MLXSW_AFA_IP_SIZE 2
+
+enum mlxsw_afa_ip_s_d {
+	/* ip refers to dip */
+	MLXSW_AFA_IP_S_D_DIP,
+	/* ip refers to sip */
+	MLXSW_AFA_IP_S_D_SIP,
+};
+
+/* afa_ip_s_d
+ * Source or destination.
+ */
+MLXSW_ITEM32(afa, ip, s_d, 0x00, 31, 1);
+
+enum mlxsw_afa_ip_m_l {
+	/* LSB: ip[63:0] refers to ip[63:0] */
+	MLXSW_AFA_IP_M_L_LSB,
+	/* MSB: ip[63:0] refers to ip[127:64] */
+	MLXSW_AFA_IP_M_L_MSB,
+};
+
+/* afa_ip_m_l
+ * MSB or LSB.
+ */
+MLXSW_ITEM32(afa, ip, m_l, 0x00, 30, 1);
+
+/* afa_ip_ip_63_32
+ * Bits [63:32] in the IP address to change to.
+ */
+MLXSW_ITEM32(afa, ip, ip_63_32, 0x08, 0, 32);
+
+/* afa_ip_ip_31_0
+ * Bits [31:0] in the IP address to change to.
+ */
+MLXSW_ITEM32(afa, ip, ip_31_0, 0x0C, 0, 32);
+
+static void mlxsw_afa_ip_pack(char *payload, enum mlxsw_afa_ip_s_d s_d,
+			      enum mlxsw_afa_ip_m_l m_l, u32 ip_31_0,
+			      u32 ip_63_32)
+{
+	mlxsw_afa_ip_s_d_set(payload, s_d);
+	mlxsw_afa_ip_m_l_set(payload, m_l);
+	mlxsw_afa_ip_ip_31_0_set(payload, ip_31_0);
+	mlxsw_afa_ip_ip_63_32_set(payload, ip_63_32);
+}
+
+int mlxsw_afa_block_append_ip(struct mlxsw_afa_block *block, bool is_dip,
+			      bool is_lsb, u32 val_31_0, u32 val_63_32,
+			      struct netlink_ext_ack *extack)
+{
+	enum mlxsw_afa_ip_s_d s_d = is_dip ? MLXSW_AFA_IP_S_D_DIP :
+					     MLXSW_AFA_IP_S_D_SIP;
+	enum mlxsw_afa_ip_m_l m_l = is_lsb ? MLXSW_AFA_IP_M_L_LSB :
+					     MLXSW_AFA_IP_M_L_MSB;
+	char *act = mlxsw_afa_block_append_action(block,
+						  MLXSW_AFA_IP_CODE,
+						  MLXSW_AFA_IP_SIZE);
+
+	if (IS_ERR(act)) {
+		NL_SET_ERR_MSG_MOD(extack, "Cannot append IP action");
+		return PTR_ERR(act);
+	}
+
+	mlxsw_afa_ip_pack(act, s_d, m_l, val_31_0, val_63_32);
+	return 0;
+}
+EXPORT_SYMBOL(mlxsw_afa_block_append_ip);
+
 /* L4 Port Action
  * --------------
  * The L4_PORT_ACTION is used for modifying the sport and dport fields of the packet, e.g. for NAT.

drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.h

···
 int mlxsw_afa_block_append_mcrouter(struct mlxsw_afa_block *block,
 				    u16 expected_irif, u16 min_mtu,
 				    bool rmid_valid, u32 kvdl_index);
+int mlxsw_afa_block_append_ip(struct mlxsw_afa_block *block, bool is_dip,
+			      bool is_lsb, u32 val_31_0, u32 val_63_32,
+			      struct netlink_ext_ack *extack);
 int mlxsw_afa_block_append_l4port(struct mlxsw_afa_block *block, bool is_dport, u16 l4_port,
 				  struct netlink_ext_ack *extack);
 int mlxsw_afa_block_append_police(struct mlxsw_afa_block *block,

drivers/net/ethernet/mellanox/mlxsw/spectrum.h

···
 /* spectrum2_kvdl.c */
 extern const struct mlxsw_sp_kvdl_ops mlxsw_sp2_kvdl_ops;
 
+enum mlxsw_sp_acl_mangle_field {
+	MLXSW_SP_ACL_MANGLE_FIELD_IP_DSFIELD,
+	MLXSW_SP_ACL_MANGLE_FIELD_IP_DSCP,
+	MLXSW_SP_ACL_MANGLE_FIELD_IP_ECN,
+	MLXSW_SP_ACL_MANGLE_FIELD_IP_SPORT,
+	MLXSW_SP_ACL_MANGLE_FIELD_IP_DPORT,
+	MLXSW_SP_ACL_MANGLE_FIELD_IP4_SIP,
+	MLXSW_SP_ACL_MANGLE_FIELD_IP4_DIP,
+	MLXSW_SP_ACL_MANGLE_FIELD_IP6_SIP_1,
+	MLXSW_SP_ACL_MANGLE_FIELD_IP6_SIP_2,
+	MLXSW_SP_ACL_MANGLE_FIELD_IP6_SIP_3,
+	MLXSW_SP_ACL_MANGLE_FIELD_IP6_SIP_4,
+	MLXSW_SP_ACL_MANGLE_FIELD_IP6_DIP_1,
+	MLXSW_SP_ACL_MANGLE_FIELD_IP6_DIP_2,
+	MLXSW_SP_ACL_MANGLE_FIELD_IP6_DIP_3,
+	MLXSW_SP_ACL_MANGLE_FIELD_IP6_DIP_4,
+};
+
 struct mlxsw_sp_acl_rule_info {
 	unsigned int priority;
 	struct mlxsw_afk_element_values values;
···
 	   ingress_bind_blocker:1,
 	   egress_bind_blocker:1,
 	   counter_valid:1,
-	   policer_index_valid:1;
+	   policer_index_valid:1,
+	   ipv6_valid:1;
 	unsigned int counter_index;
 	u16 policer_index;
+	struct {
+		u32 prev_val;
+		enum mlxsw_sp_acl_mangle_field prev_field;
+	} ipv6;
 };
 
 /* spectrum_flow.c */

drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c

···
 						      extack);
 }
 
-enum mlxsw_sp_acl_mangle_field {
-	MLXSW_SP_ACL_MANGLE_FIELD_IP_DSFIELD,
-	MLXSW_SP_ACL_MANGLE_FIELD_IP_DSCP,
-	MLXSW_SP_ACL_MANGLE_FIELD_IP_ECN,
-	MLXSW_SP_ACL_MANGLE_FIELD_IP_SPORT,
-	MLXSW_SP_ACL_MANGLE_FIELD_IP_DPORT,
-};
-
 struct mlxsw_sp_acl_mangle_action {
 	enum flow_action_mangle_base htype;
 	/* Offset is u32-aligned. */
···
 
 	MLXSW_SP_ACL_MANGLE_ACTION_UDP(0, 0x0000ffff, 16, IP_SPORT),
 	MLXSW_SP_ACL_MANGLE_ACTION_UDP(0, 0xffff0000, 0,  IP_DPORT),
+
+	MLXSW_SP_ACL_MANGLE_ACTION_IP4(12, 0x00000000, 0, IP4_SIP),
+	MLXSW_SP_ACL_MANGLE_ACTION_IP4(16, 0x00000000, 0, IP4_DIP),
+
+	MLXSW_SP_ACL_MANGLE_ACTION_IP6(8, 0x00000000, 0, IP6_SIP_1),
+	MLXSW_SP_ACL_MANGLE_ACTION_IP6(12, 0x00000000, 0, IP6_SIP_2),
+	MLXSW_SP_ACL_MANGLE_ACTION_IP6(16, 0x00000000, 0, IP6_SIP_3),
+	MLXSW_SP_ACL_MANGLE_ACTION_IP6(20, 0x00000000, 0, IP6_SIP_4),
+	MLXSW_SP_ACL_MANGLE_ACTION_IP6(24, 0x00000000, 0, IP6_DIP_1),
+	MLXSW_SP_ACL_MANGLE_ACTION_IP6(28, 0x00000000, 0, IP6_DIP_2),
+	MLXSW_SP_ACL_MANGLE_ACTION_IP6(32, 0x00000000, 0, IP6_DIP_3),
+	MLXSW_SP_ACL_MANGLE_ACTION_IP6(36, 0x00000000, 0, IP6_DIP_4),
 };
 
 static int
···
 	return err;
 }
 
+static int
+mlxsw_sp2_acl_rulei_act_mangle_field_ip_odd(struct mlxsw_sp_acl_rule_info *rulei,
+					    enum mlxsw_sp_acl_mangle_field field,
+					    u32 val, struct netlink_ext_ack *extack)
+{
+	if (!rulei->ipv6_valid) {
+		rulei->ipv6.prev_val = val;
+		rulei->ipv6_valid = true;
+		rulei->ipv6.prev_field = field;
+		return 0;
+	}
+
+	NL_SET_ERR_MSG_MOD(extack, "Unsupported mangle field order");
+	return -EOPNOTSUPP;
+}
+
 static int mlxsw_sp2_acl_rulei_act_mangle_field(struct mlxsw_sp *mlxsw_sp,
 						struct mlxsw_sp_acl_rule_info *rulei,
 						struct mlxsw_sp_acl_mangle_action *mact,
···
 		return mlxsw_afa_block_append_l4port(rulei->act_block, false, val, extack);
 	case MLXSW_SP_ACL_MANGLE_FIELD_IP_DPORT:
 		return mlxsw_afa_block_append_l4port(rulei->act_block, true, val, extack);
+	/* IPv4 fields */
+	case MLXSW_SP_ACL_MANGLE_FIELD_IP4_SIP:
+		return mlxsw_afa_block_append_ip(rulei->act_block, false,
+						 true, val, 0, extack);
+	case MLXSW_SP_ACL_MANGLE_FIELD_IP4_DIP:
+		return mlxsw_afa_block_append_ip(rulei->act_block, true,
+						 true, val, 0, extack);
+	/* IPv6 fields */
+	case MLXSW_SP_ACL_MANGLE_FIELD_IP6_SIP_1:
+	case MLXSW_SP_ACL_MANGLE_FIELD_IP6_SIP_3:
+	case MLXSW_SP_ACL_MANGLE_FIELD_IP6_DIP_1:
+	case MLXSW_SP_ACL_MANGLE_FIELD_IP6_DIP_3:
+		return mlxsw_sp2_acl_rulei_act_mangle_field_ip_odd(rulei,
+								   mact->field,
+								   val, extack);
+	case MLXSW_SP_ACL_MANGLE_FIELD_IP6_SIP_2:
+		if (rulei->ipv6_valid &&
+		    rulei->ipv6.prev_field == MLXSW_SP_ACL_MANGLE_FIELD_IP6_SIP_1) {
+			rulei->ipv6_valid = false;
+			return mlxsw_afa_block_append_ip(rulei->act_block,
+							 false, false, val,
+							 rulei->ipv6.prev_val,
+							 extack);
+		}
+		break;
+	case MLXSW_SP_ACL_MANGLE_FIELD_IP6_SIP_4:
+		if (rulei->ipv6_valid &&
+		    rulei->ipv6.prev_field == MLXSW_SP_ACL_MANGLE_FIELD_IP6_SIP_3) {
+			rulei->ipv6_valid = false;
+			return mlxsw_afa_block_append_ip(rulei->act_block,
+							 false, true, val,
+							 rulei->ipv6.prev_val,
+							 extack);
+		}
+		break;
+	case MLXSW_SP_ACL_MANGLE_FIELD_IP6_DIP_2:
+		if (rulei->ipv6_valid &&
+		    rulei->ipv6.prev_field == MLXSW_SP_ACL_MANGLE_FIELD_IP6_DIP_1) {
+			rulei->ipv6_valid = false;
+			return mlxsw_afa_block_append_ip(rulei->act_block,
+							 true, false, val,
+							 rulei->ipv6.prev_val,
+							 extack);
+		}
+		break;
+	case MLXSW_SP_ACL_MANGLE_FIELD_IP6_DIP_4:
+		if (rulei->ipv6_valid &&
+		    rulei->ipv6.prev_field == MLXSW_SP_ACL_MANGLE_FIELD_IP6_DIP_3) {
+			rulei->ipv6_valid = false;
+			return mlxsw_afa_block_append_ip(rulei->act_block,
+							 true, true, val,
+							 rulei->ipv6.prev_val,
+							 extack);
+		}
+		break;
 	default:
 		break;
 	}

drivers/net/ethernet/mellanox/mlxsw/spectrum_flower.c

···
 			return -EOPNOTSUPP;
 		}
 	}
+
+	if (rulei->ipv6_valid) {
+		NL_SET_ERR_MSG_MOD(extack, "Unsupported mangle field");
+		return -EOPNOTSUPP;
+	}
+
 	return 0;
 }
 

tools/testing/selftests/net/forwarding/pedit_ip.sh

···
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+
+# This test sends traffic from H1 to H2. Either on ingress of $swp1, or on
+# egress of $swp2, the traffic is acted upon by a pedit action. An ingress
+# filter installed on $h2 verifies that the packet looks like expected.
+#
+# +----------------------+                             +----------------------+
+# | H1                   |                             |                   H2 |
+# |    + $h1             |                             |            $h2 +     |
+# |    | 192.0.2.1/28    |                             |   192.0.2.2/28 |     |
+# +----|-----------------+                             +----------------|-----+
+#      |                                                                |
+# +----|----------------------------------------------------------------|-----+
+# | SW |                                                                |     |
+# |  +-|----------------------------------------------------------------|-+   |
+# |  | + $swp1                       BR                           $swp2 + |   |
+# |  +--------------------------------------------------------------------+   |
+# +---------------------------------------------------------------------------+
+
+ALL_TESTS="
+	ping_ipv4
+	ping_ipv6
+	test_ip4_src
+	test_ip4_dst
+	test_ip6_src
+	test_ip6_dst
+"
+
+NUM_NETIFS=4
+source lib.sh
+source tc_common.sh
+
+h1_create()
+{
+	simple_if_init $h1 192.0.2.1/28 2001:db8:1::1/64
+}
+
+h1_destroy()
+{
+	simple_if_fini $h1 192.0.2.1/28 2001:db8:1::1/64
+}
+
+h2_create()
+{
+	simple_if_init $h2 192.0.2.2/28 2001:db8:1::2/64
+	tc qdisc add dev $h2 clsact
+}
+
+h2_destroy()
+{
+	tc qdisc del dev $h2 clsact
+	simple_if_fini $h2 192.0.2.2/28 2001:db8:1::2/64
+}
+
+switch_create()
+{
+	ip link add name br1 up type bridge vlan_filtering 1
+	ip link set dev $swp1 master br1
+	ip link set dev $swp1 up
+	ip link set dev $swp2 master br1
+	ip link set dev $swp2 up
+
+	tc qdisc add dev $swp1 clsact
+	tc qdisc add dev $swp2 clsact
+}
+
+switch_destroy()
+{
+	tc qdisc del dev $swp2 clsact
+	tc qdisc del dev $swp1 clsact
+
+	ip link set dev $swp2 down
+	ip link set dev $swp2 nomaster
+	ip link set dev $swp1 down
+	ip link set dev $swp1 nomaster
+	ip link del dev br1
+}
+
+setup_prepare()
+{
+	h1=${NETIFS[p1]}
+	swp1=${NETIFS[p2]}
+
+	swp2=${NETIFS[p3]}
+	h2=${NETIFS[p4]}
+
+	h2mac=$(mac_get $h2)
+
+	vrf_prepare
+	h1_create
+	h2_create
+	switch_create
+}
+
+cleanup()
+{
+	pre_cleanup
+
+	switch_destroy
+	h2_destroy
+	h1_destroy
+	vrf_cleanup
+}
+
+ping_ipv4()
+{
+	ping_test $h1 192.0.2.2
+}
+
+ping_ipv6()
+{
+	ping6_test $h1 2001:db8:1::2
+}
+
+do_test_pedit_ip()
+{
+	local pedit_locus=$1; shift
+	local pedit_action=$1; shift
+	local match_prot=$1; shift
+	local match_flower=$1; shift
+	local mz_flags=$1; shift
+
+	tc filter add $pedit_locus handle 101 pref 1 \
+	   flower action pedit ex munge $pedit_action
+	tc filter add dev $h2 ingress handle 101 pref 1 prot $match_prot \
+	   flower skip_hw $match_flower action pass
+
+	RET=0
+
+	$MZ $mz_flags $h1 -c 10 -d 20msec -p 100 -a own -b $h2mac -q -t ip
+
+	local pkts
+	pkts=$(busywait "$TC_HIT_TIMEOUT" until_counter_is ">= 10" \
+			tc_rule_handle_stats_get "dev $h2 ingress" 101)
+	check_err $? "Expected to get 10 packets, but got $pkts."
+
+	pkts=$(tc_rule_handle_stats_get "$pedit_locus" 101)
+	((pkts >= 10))
+	check_err $? "Expected to get 10 packets on pedit rule, but got $pkts."
+
+	log_test "$pedit_locus pedit $pedit_action"
+
+	tc filter del dev $h2 ingress pref 1
+	tc filter del $pedit_locus pref 1
+}
+
+do_test_pedit_ip6()
+{
+	local locus=$1; shift
+	local pedit_addr=$1; shift
+	local flower_addr=$1; shift
+
+	do_test_pedit_ip "$locus" "$pedit_addr set 2001:db8:2::1" ipv6	\
+			 "$flower_addr 2001:db8:2::1"			\
+			 "-6 -A 2001:db8:1::1 -B 2001:db8:1::2"
+}
+
+do_test_pedit_ip4()
+{
+	local locus=$1; shift
+	local pedit_addr=$1; shift
+	local flower_addr=$1; shift
+
+	do_test_pedit_ip "$locus" "$pedit_addr set 198.51.100.1" ip	\
+			 "$flower_addr 198.51.100.1"			\
+			 "-A 192.0.2.1 -B 192.0.2.2"
+}
+
+test_ip4_src()
+{
+	do_test_pedit_ip4 "dev $swp1 ingress" "ip src" src_ip
+	do_test_pedit_ip4 "dev $swp2 egress"  "ip src" src_ip
+}
+
+test_ip4_dst()
+{
+	do_test_pedit_ip4 "dev $swp1 ingress" "ip dst" dst_ip
+	do_test_pedit_ip4 "dev $swp2 egress"  "ip dst" dst_ip
+}
+
+test_ip6_src()
+{
+	do_test_pedit_ip6 "dev $swp1 ingress" "ip6 src" src_ip
+	do_test_pedit_ip6 "dev $swp2 egress"  "ip6 src" src_ip
+}
+
+test_ip6_dst()
+{
+	do_test_pedit_ip6 "dev $swp1 ingress" "ip6 dst" dst_ip
+	do_test_pedit_ip6 "dev $swp2 egress"  "ip6 dst" dst_ip
+}
+
+trap cleanup EXIT
+
+setup_prepare
+setup_wait
+
+tests_run
+
+exit $EXIT_STATUS