[PATCH iproute2 0/3] improve lwtunnel route support

[PATCH iproute2 0/3] improve lwtunnel route support

[Paolo Abeni]

This patch series try to improve the current route based
lwtunnel support in iproute2, namely adding support for the
COLLECT_METADATA flag in vxlan and gre link, and for ip6
encap type in lwtunnel.

Tunnel devices need to have the COLLECT_METADATA flag
set in order to be used for route based lwtunnel.

Paolo Abeni (3):
  vxlan: add support for collect metadata flag
  gre: add support for collect metadata flag
  lwtunnel: implement support for ip6 encap

 ip/iplink_vxlan.c     | 12 ++++++-
 ip/iproute_lwtunnel.c | 91 ++++++++++++++++++++++++++++++++++++++++++++++++++-
 ip/link_gre.c         | 11 +++++++
 3 files changed, 112 insertions(+), 2 deletions(-)

-- 
1.8.3.1

[PATCH iproute2 1/3] vxlan: add support for collect metadata flag

[Paolo Abeni]

This patch add support for IFLA_VXLAN_COLLECT_METADATA via the
'collectmetadata' keyword to the vxlan link.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
 ip/iplink_vxlan.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/ip/iplink_vxlan.c b/ip/iplink_vxlan.c
index db29bf0..a943e03 100644
--- a/ip/iplink_vxlan.c
+++ b/ip/iplink_vxlan.c
@@ -31,7 +31,7 @@ static void print_explain(FILE *f)
 	fprintf(f, "                 [ ageing SECONDS ] [ maxaddress NUMBER ]\n");
 	fprintf(f, "                 [ [no]udpcsum ] [ [no]udp6zerocsumtx ] [ [no]udp6zerocsumrx ]\n");
 	fprintf(f, "                 [ [no]remcsumtx ] [ [no]remcsumrx ]\n");
-	fprintf(f, "                 [ gbp ]\n");
+	fprintf(f, "                 [ [no]collectmetadata ] [ gbp ]\n");
 	fprintf(f, "\n");
 	fprintf(f, "Where: VNI := 0-16777215\n");
 	fprintf(f, "       ADDR := { IP_ADDRESS | any }\n");
@@ -72,6 +72,7 @@ static int vxlan_parse_opt(struct link_util *lu, int argc, char **argv,
 	__u8 udp6zerocsumrx = 0;
 	__u8 remcsumtx = 0;
 	__u8 remcsumrx = 0;
+	__u8 metadata = 0;
 	__u8 gbp = 0;
 	int dst_port_set = 0;
 	struct ifla_vxlan_port_range range = { 0, 0 };
@@ -210,6 +211,10 @@ static int vxlan_parse_opt(struct link_util *lu, int argc, char **argv,
 			remcsumrx = 1;
 		} else if (!matches(*argv, "noremcsumrx")) {
 			remcsumrx = 0;
+		} else if (!matches(*argv, "collectmetadata")) {
+			metadata = 1;
+		} else if (!matches(*argv, "nocollectmetadata")) {
+			metadata = 0;
 		} else if (!matches(*argv, "gbp")) {
 			gbp = 1;
 		} else if (matches(*argv, "help") == 0) {
@@ -272,6 +277,7 @@ static int vxlan_parse_opt(struct link_util *lu, int argc, char **argv,
 	addattr8(n, 1024, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, udp6zerocsumrx);
 	addattr8(n, 1024, IFLA_VXLAN_REMCSUM_TX, remcsumtx);
 	addattr8(n, 1024, IFLA_VXLAN_REMCSUM_RX, remcsumrx);
+	addattr8(n, 1024, IFLA_VXLAN_COLLECT_METADATA, metadata);
 
 	if (noage)
 		addattr32(n, 1024, IFLA_VXLAN_AGEING, 0);
@@ -428,6 +434,10 @@ static void vxlan_print_opt(struct link_util *lu, FILE *f, struct rtattr *tb[])
 	    rta_getattr_u8(tb[IFLA_VXLAN_REMCSUM_RX]))
 		fputs("remcsumrx ", f);
 
+	if (tb[IFLA_VXLAN_COLLECT_METADATA] &&
+	    rta_getattr_u8(tb[IFLA_VXLAN_COLLECT_METADATA]))
+		fputs("collectmetadata ", f);
+
 	if (tb[IFLA_VXLAN_GBP])
 		fputs("gbp ", f);
 }
-- 
1.8.3.1

Re: [PATCH iproute2 1/3] vxlan: add support for collect metadata flag

[Jiri Benc]

On Tue, 15 Dec 2015 16:48:25 +0100, Paolo Abeni wrote:
> This patch add support for IFLA_VXLAN_COLLECT_METADATA via the
> 'collectmetadata' keyword to the vxlan link.

Could we give this more understandable name? I've never liked the
COLLECT_METADATA name as it describes the actual implementation, not
the intention of the flag.

The name should express that the vxlan interface is driven by an
external control plane instead of the internal fdb. This could be ovs,
a route, tc action, etc. Something like "external" would be better
name, I think.

Also, as a related remark, please note that IFLA_VXLAN_COLLECT_METADATA
is mutually exclusive with many other flags. It's currently not
enforced correctly at the kernel level (instead, you get weird
behavior). I'll fix this on the kernel side but we may want to enforce
this on the iproute2 side, too, to get a better error message. This is
not to be addressed by this patch, though.

 Jiri

-- 
Jiri Benc

Re: [PATCH iproute2 1/3] vxlan: add support for collect metadata flag

[Jiri Benc]

On Wed, 16 Dec 2015 11:50:53 +0100, Jiri Benc wrote:
> Also, as a related remark, please note that IFLA_VXLAN_COLLECT_METADATA
> is mutually exclusive with many other flags. It's currently not
> enforced correctly at the kernel level (instead, you get weird
> behavior). I'll fix this on the kernel side but we may want to enforce
> this on the iproute2 side, too, to get a better error message. This is
> not to be addressed by this patch, though.

Actually, there's one thing this patch must do. Currently, vxlan
enforces vni to be specified. This is not required for metadata based
tunneling. It's actually not permitted.

You'll need to add a hunk similar to this:

--- a/ip/iplink_vxlan.c
+++ b/ip/iplink_vxlan.c
@@ -215,7 +215,12 @@ static int vxlan_parse_opt(struct link_util *lu, int argc, char **argv,
 		argc--, argv++;
 	}
 
-	if (!vni_set) {
+	if (metadata && vni_set) {
+		fprintf(stderr, "vxlan: both metadata and vni cannot be specified\n");
+		return -1;
+	}
+
+	if (!metadata && !vni_set) {
 		fprintf(stderr, "vxlan: missing virtual network identifier\n");
 		return -1;
 	}

-- 
Jiri Benc

Re: [PATCH iproute2 1/3] vxlan: add support for collect metadata flag

[Paolo Abeni]

On Wed, 2015-12-16 at 11:56 +0100, Jiri Benc wrote:
> On Wed, 16 Dec 2015 11:50:53 +0100, Jiri Benc wrote:
> > Also, as a related remark, please note that IFLA_VXLAN_COLLECT_METADATA
> > is mutually exclusive with many other flags. It's currently not
> > enforced correctly at the kernel level (instead, you get weird
> > behavior). I'll fix this on the kernel side but we may want to enforce
> > this on the iproute2 side, too, to get a better error message. This is
> > not to be addressed by this patch, though.
> 
> Actually, there's one thing this patch must do. Currently, vxlan
> enforces vni to be specified. This is not required for metadata based
> tunneling. It's actually not permitted.

Ok, thank you. I'll re-submit incorporating your feedback.

Cheers,

Paolo

[PATCH iproute2 2/3] gre: add support for collect metadata flag

[Paolo Abeni]

This patch add support for IFLA_GRE_COLLECT_METADATA via the
'collectmetadata' keyword to the gre link.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
 ip/link_gre.c | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/ip/link_gre.c b/ip/link_gre.c
index 58f416c..8be157f 100644
--- a/ip/link_gre.c
+++ b/ip/link_gre.c
@@ -74,6 +74,7 @@ static int gre_parse_opt(struct link_util *lu, int argc, char **argv,
 	__u16 encapflags = 0;
 	__u16 encapsport = 0;
 	__u16 encapdport = 0;
+	__u8 metadata = 0;
 
 	if (!(n->nlmsg_flags & NLM_F_CREATE)) {
 		memset(&req, 0, sizeof(req));
@@ -148,6 +149,9 @@ get_failed:
 			encapsport = rta_getattr_u16(greinfo[IFLA_GRE_ENCAP_SPORT]);
 		if (greinfo[IFLA_GRE_ENCAP_DPORT])
 			encapdport = rta_getattr_u16(greinfo[IFLA_GRE_ENCAP_DPORT]);
+
+		if (greinfo[IFLA_GRE_COLLECT_METADATA])
+			metadata = 1;
 	}
 
 	while (argc > 0) {
@@ -291,6 +295,8 @@ get_failed:
 			encapflags |= TUNNEL_ENCAP_FLAG_REMCSUM;
 		} else if (strcmp(*argv, "noencap-remcsum") == 0) {
 			encapflags |= ~TUNNEL_ENCAP_FLAG_REMCSUM;
+		} else if (strcmp(*argv, "collectmetadata") == 0) {
+			metadata = 1;
 		} else
 			usage();
 		argc--; argv++;
@@ -325,6 +331,8 @@ get_failed:
 	addattr16(n, 1024, IFLA_GRE_ENCAP_FLAGS, encapflags);
 	addattr16(n, 1024, IFLA_GRE_ENCAP_SPORT, htons(encapsport));
 	addattr16(n, 1024, IFLA_GRE_ENCAP_DPORT, htons(encapdport));
+	if (metadata)
+		addattr_l(n, 1024, IFLA_GRE_COLLECT_METADATA, NULL, 0);
 
 	return 0;
 }
@@ -413,6 +421,9 @@ static void gre_print_opt(struct link_util *lu, FILE *f, struct rtattr *tb[])
 	if (oflags & GRE_CSUM)
 		fputs("ocsum ", f);
 
+	if (tb[IFLA_GRE_COLLECT_METADATA])
+		fputs("collectmetadata ", f);
+
 	if (tb[IFLA_GRE_ENCAP_TYPE] &&
 	    *(__u16 *)RTA_DATA(tb[IFLA_GRE_ENCAP_TYPE]) != TUNNEL_ENCAP_NONE) {
 		__u16 type = rta_getattr_u16(tb[IFLA_GRE_ENCAP_TYPE]);
-- 
1.8.3.1

[PATCH iproute2 3/3] lwtunnel: implement support for ip6 encap

[Paolo Abeni]

Currently ip6 encap support for lwtunnel is missing.
This patch implement it, mostly duplicating the ipv4 parts.

Also be sure to insert a space after the encap type, when
showing lwtunnel, to avoid the tunnel type and the following
argument being merged into a single word.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
 ip/iproute_lwtunnel.c | 91 ++++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 90 insertions(+), 1 deletion(-)

diff --git a/ip/iproute_lwtunnel.c b/ip/iproute_lwtunnel.c
index c825b82..b34e02c 100644
--- a/ip/iproute_lwtunnel.c
+++ b/ip/iproute_lwtunnel.c
@@ -95,6 +95,37 @@ static void print_encap_ip(FILE *fp, struct rtattr *encap)
 		fprintf(fp, "tos %d ", rta_getattr_u8(tb[LWTUNNEL_IP_TOS]));
 }
 
+static void print_encap_ip6(FILE *fp, struct rtattr *encap)
+{
+	struct rtattr *tb[LWTUNNEL_IP6_MAX+1];
+	char abuf[256];
+
+	parse_rtattr_nested(tb, LWTUNNEL_IP6_MAX, encap);
+
+	if (tb[LWTUNNEL_IP6_ID])
+		fprintf(fp, "id %llu ", ntohll(rta_getattr_u64(tb[LWTUNNEL_IP6_ID])));
+
+	if (tb[LWTUNNEL_IP6_SRC])
+		fprintf(fp, "src %s ",
+			rt_addr_n2a(AF_INET6,
+				    RTA_PAYLOAD(tb[LWTUNNEL_IP6_SRC]),
+				    RTA_DATA(tb[LWTUNNEL_IP6_SRC]),
+				    abuf, sizeof(abuf)));
+
+	if (tb[LWTUNNEL_IP6_DST])
+		fprintf(fp, "dst %s ",
+			rt_addr_n2a(AF_INET6,
+				    RTA_PAYLOAD(tb[LWTUNNEL_IP6_DST]),
+				    RTA_DATA(tb[LWTUNNEL_IP6_DST]),
+				    abuf, sizeof(abuf)));
+
+	if (tb[LWTUNNEL_IP6_HOPLIMIT])
+		fprintf(fp, "hoplimit %d ", rta_getattr_u8(tb[LWTUNNEL_IP6_HOPLIMIT]));
+
+	if (tb[LWTUNNEL_IP6_TC])
+		fprintf(fp, "tc %d ", rta_getattr_u8(tb[LWTUNNEL_IP6_TC]));
+}
+
 void lwt_print_encap(FILE *fp, struct rtattr *encap_type,
 			  struct rtattr *encap)
 {
@@ -105,7 +136,7 @@ void lwt_print_encap(FILE *fp, struct rtattr *encap_type,
 
 	et = rta_getattr_u16(encap_type);
 
-	fprintf(fp, " encap %s", format_encap_type(et));
+	fprintf(fp, " encap %s ", format_encap_type(et));
 
 	switch (et) {
 	case LWTUNNEL_ENCAP_MPLS:
@@ -114,6 +145,9 @@ void lwt_print_encap(FILE *fp, struct rtattr *encap_type,
 	case LWTUNNEL_ENCAP_IP:
 		print_encap_ip(fp, encap);
 		break;
+	case LWTUNNEL_ENCAP_IP6:
+		print_encap_ip6(fp, encap);
+		break;
 	}
 }
 
@@ -190,6 +224,58 @@ static int parse_encap_ip(struct rtattr *rta, size_t len, int *argcp, char ***ar
 	return 0;
 }
 
+static int parse_encap_ip6(struct rtattr *rta, size_t len, int *argcp, char ***argvp)
+{
+	int id_ok = 0, dst_ok = 0, tos_ok = 0, ttl_ok = 0;
+	char **argv = *argvp;
+	int argc = *argcp;
+
+	while (argc > 0) {
+		if (strcmp(*argv, "id") == 0) {
+			__u64 id;
+			NEXT_ARG();
+			if (id_ok++)
+				duparg2("id", *argv);
+			if (get_u64(&id, *argv, 0))
+				invarg("\"id\" value is invalid\n", *argv);
+			rta_addattr64(rta, len, LWTUNNEL_IP6_ID, htonll(id));
+		} else if (strcmp(*argv, "dst") == 0) {
+			inet_prefix addr;
+			NEXT_ARG();
+			if (dst_ok++)
+				duparg2("dst", *argv);
+			get_addr(&addr, *argv, AF_INET6);
+			rta_addattr_l(rta, len, LWTUNNEL_IP6_DST, &addr.data, addr.bytelen);
+		} else if (strcmp(*argv, "tc") == 0) {
+			__u32 tc;
+			NEXT_ARG();
+			if (tos_ok++)
+				duparg2("tc", *argv);
+			if (rtnl_dsfield_a2n(&tc, *argv))
+				invarg("\"tc\" value is invalid\n", *argv);
+			rta_addattr8(rta, len, LWTUNNEL_IP6_TC, tc);
+		} else if (strcmp(*argv, "hoplimit") == 0) {
+			__u8 hoplimit;
+			NEXT_ARG();
+			if (ttl_ok++)
+				duparg2("hoplimit", *argv);
+			if (get_u8(&hoplimit, *argv, 0))
+				invarg("\"hoplimit\" value is invalid\n", *argv);
+			rta_addattr8(rta, len, LWTUNNEL_IP6_HOPLIMIT, hoplimit);
+		} else {
+			break;
+		}
+		argc--; argv++;
+	}
+
+	/* argv is currently the first unparsed argument,
+	 * but the lwt_parse_encap() caller will move to the next,
+	 * so step back */
+	*argcp = argc + 1;
+	*argvp = argv - 1;
+
+	return 0;
+}
 
 int lwt_parse_encap(struct rtattr *rta, size_t len, int *argcp, char ***argvp)
 {
@@ -217,6 +303,9 @@ int lwt_parse_encap(struct rtattr *rta, size_t len, int *argcp, char ***argvp)
 	case LWTUNNEL_ENCAP_IP:
 		parse_encap_ip(rta, len, &argc, &argv);
 		break;
+	case LWTUNNEL_ENCAP_IP6:
+		parse_encap_ip6(rta, len, &argc, &argv);
+		break;
 	default:
 		fprintf(stderr, "Error: unsupported encap type\n");
 		break;
-- 
1.8.3.1