[PATCH v4] net: ipv6: seg6: report lwtunnel setup errors via extack

[PATCH v4] net: ipv6: seg6: report lwtunnel setup errors via extack

[Gabriel Goller]

seg6_build_state() rejected invalid configurations with a generic -EINVAL,
so "ip route add ... encap seg6 ..." always reported "Invalid argument".

Attach an extack message to each of them. The checks themselves and their
return values are unchanged.

Signed-off-by: Gabriel Goller <g.goller@proxmox.com>
---

v4 (https://lore.kernel.org/all/20260923120228.172643-1-g.goller@proxmox.com/):
 * use NL_SET_ERR_MSG when the SRH attr is missing
 * use NL_SET_ERR_MSG_ATTR when we have an invalid attr

v3 (https://lore.kernel.org/all/20260922090851.38978-1-g.goller@proxmox.com/):
 * format fix
 * use NL_SET_ERR_MSG_ATTR

v2 (https://lore.kernel.org/all/20260918153544.1178884-1-g.goller@proxmox.com/):
 * dropped tests

 net/ipv6/seg6_iptunnel.c | 27 ++++++++++++++++++++++-----
 1 file changed, 22 insertions(+), 5 deletions(-)

diff --git a/net/ipv6/seg6_iptunnel.c b/net/ipv6/seg6_iptunnel.c
index 69a6ef5c5317..73160179a545 100644
--- a/net/ipv6/seg6_iptunnel.c
+++ b/net/ipv6/seg6_iptunnel.c
@@ -762,8 +762,11 @@ static int seg6_build_state(struct net *net, struct nlattr *nla,
 	struct seg6_lwt *slwt;
 	int err;
 
-	if (family != AF_INET && family != AF_INET6)
+	if (family != AF_INET && family != AF_INET6) {
+		NL_SET_ERR_MSG(extack,
+			       "unsupported address family for SRv6 encapsulation");
 		return -EINVAL;
+	}
 
 	err = nla_parse_nested_deprecated(tb, SEG6_IPTUNNEL_MAX, nla,
 					  seg6_iptunnel_policy, extack);
@@ -771,8 +774,10 @@ static int seg6_build_state(struct net *net, struct nlattr *nla,
 	if (err < 0)
 		return err;
 
-	if (!tb[SEG6_IPTUNNEL_SRH])
+	if (!tb[SEG6_IPTUNNEL_SRH]) {
+		NL_SET_ERR_MSG(extack, "missing SRv6 SRH attribute");
 		return -EINVAL;
+	}
 
 	tuninfo = nla_data(tb[SEG6_IPTUNNEL_SRH]);
 	tuninfo_len = nla_len(tb[SEG6_IPTUNNEL_SRH]);
@@ -782,13 +787,19 @@ static int seg6_build_state(struct net *net, struct nlattr *nla,
 	 */
 	min_size = sizeof(*tuninfo) + sizeof(struct ipv6_sr_hdr) +
 		   sizeof(struct in6_addr);
-	if (tuninfo_len < min_size)
+	if (tuninfo_len < min_size) {
+		NL_SET_ERR_MSG_ATTR(extack, tb[SEG6_IPTUNNEL_SRH],
+				    "truncated SRv6 SRH attribute");
 		return -EINVAL;
+	}
 
 	switch (tuninfo->mode) {
 	case SEG6_IPTUN_MODE_INLINE:
-		if (family != AF_INET6)
+		if (family != AF_INET6) {
+			NL_SET_ERR_MSG_ATTR(extack, tb[SEG6_IPTUNNEL_SRH],
+					    "inline mode requires an IPv6 route");
 			return -EINVAL;
+		}
 
 		if (tb[SEG6_IPTUNNEL_SRC]) {
 			NL_SET_ERR_MSG(extack, "incompatible mode for tunsrc");
@@ -804,12 +815,18 @@ static int seg6_build_state(struct net *net, struct nlattr *nla,
 	case SEG6_IPTUN_MODE_L2ENCAP_RED:
 		break;
 	default:
+		NL_SET_ERR_MSG_ATTR(extack, tb[SEG6_IPTUNNEL_SRH],
+				    "invalid SRv6 encapsulation mode");
 		return -EINVAL;
 	}
 
 	/* verify that SRH is consistent */
-	if (!seg6_validate_srh(tuninfo->srh, tuninfo_len - sizeof(*tuninfo), false))
+	if (!seg6_validate_srh(tuninfo->srh, tuninfo_len - sizeof(*tuninfo),
+			       false)) {
+		NL_SET_ERR_MSG_ATTR(extack, tb[SEG6_IPTUNNEL_SRH],
+				    "invalid SRv6 segment routing header");
 		return -EINVAL;
+	}
 
 	newts = lwtunnel_state_alloc(tuninfo_len + sizeof(*slwt));
 	if (!newts)
-- 
2.47.3

Re: [PATCH v4] net: ipv6: seg6: report lwtunnel setup errors via extack

[Jakub Kicinski]

On Mon,  5 Oct 2026 16:06:34 +0200 Gabriel Goller wrote:
> seg6_build_state() rejected invalid configurations with a generic -EINVAL,
> so "ip route add ... encap seg6 ..." always reported "Invalid argument".
> 
> Attach an extack message to each of them. The checks themselves and their
> return values are unchanged.

Why not use _MISS were appropriate?

https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20261005140637.644840-1-g.goller@proxmox.com
-- 
pw-bot: cr