[RFC next-net 2/3] ipv4: accept the reported route type when deleting a blackhole nexthop route

From: Stephen Hemminger

Date: Fri Nov 20 2026 - 15:06:52 EST


fib_dump_info() rewrites rtm_type to RTN_BLACKHOLE for a blackhole nexthop
object, before the nexthop_compat_mode bail-out:

if (fi->nh) {
if (nla_put_u32(skb, RTA_NH_ID, fi->nh->id))
goto nla_put_failure;
if (nexthop_is_blackhole(fi->nh))
rtm->rtm_type = RTN_BLACKHOLE;

So rtm_type describes the nexthop object, not the alias, which keeps the
type it was created with. fib_table_delete() matches the stored type, so
the kernel won't delete a route using the type it just reported:

# ip nexthop add id 301 blackhole
# ip route add 192.0.2.0/24 nhid 301
# ip route show 192.0.2.0/24
blackhole 192.0.2.0/24 nhid 301 dev lo
# ip route del blackhole 192.0.2.0/24 nhid 301
RTNETLINK answers: No such process
# ip route del unicast 192.0.2.0/24 nhid 301
#

"ip route flush" echoes the dumped message back, rtmsg header included, so
it carries the rewritten type and fails with ESRCH. Because the rewrite
happens before the compat-mode check, this breaks flush even with
nexthop_compat_mode off, where there are no legacy attributes at all.

Accept RTN_BLACKHOLE on delete when the alias uses a blackhole nexthop
object, so a route can be removed using the type that was advertised.
"ip route flush type blackhole" works too: the type filter is applied to
the dump in userspace, so the kernel only ever sees the type it reported.

Dropping the rewrite in fib_dump_info() instead would change what every
existing dump reader sees.

IPv6 is unaffected, ip6_route_del() doesn't compare rtm_type.

Reported-by: Yuen Su <zypulsecore@xxxxxxxxx>
Link: https://lore.kernel.org/netdev/CAC9WPdwmM4=64DH5nY85srnLhRBS7PkjR9vrsALpob-GutbmZw@xxxxxxxxxxxxxx/
Fixes: 4c7e8084fd46 ("ipv4: Plumb support for nexthop object in a fib_info")
Signed-off-by: Stephen Hemminger <stephen@xxxxxxxxxxxxxxxxxx>
---
net/ipv4/fib_lookup.h | 1 +
net/ipv4/fib_semantics.c | 14 ++++++++++++++
net/ipv4/fib_trie.c | 2 +-
3 files changed, 16 insertions(+), 1 deletion(-)

diff --git a/net/ipv4/fib_lookup.h b/net/ipv4/fib_lookup.h
index 0b72796dd1ad..5109cc197de2 100644
--- a/net/ipv4/fib_lookup.h
+++ b/net/ipv4/fib_lookup.h
@@ -40,6 +40,7 @@ struct fib_info *fib_create_info(struct fib_config *cfg,
struct netlink_ext_ack *extack);
int fib_nh_match(struct net *net, struct fib_config *cfg, struct fib_info *fi,
struct netlink_ext_ack *extack);
+bool fib_type_match(u8 fc_type, const struct fib_alias *fa);
bool fib_metrics_match(struct fib_config *cfg, struct fib_info *fi);
int fib_dump_info(struct sk_buff *skb, u32 pid, u32 seq, int event,
const struct fib_rt_info *fri, unsigned int flags);
diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c
index db030bc26b6d..a089761e2104 100644
--- a/net/ipv4/fib_semantics.c
+++ b/net/ipv4/fib_semantics.c
@@ -1061,6 +1061,20 @@ int fib_nh_match(struct net *net, struct fib_config *cfg, struct fib_info *fi,
return 0;
}

+/* fib_dump_info() reports a blackhole nexthop object as RTN_BLACKHOLE
+ * whatever type the alias has, so accept that type back on delete.
+ */
+bool fib_type_match(u8 fc_type, const struct fib_alias *fa)
+{
+ const struct fib_info *fi = fa->fa_info;
+
+ if (!fc_type || fa->fa_type == fc_type)
+ return true;
+
+ return fc_type == RTN_BLACKHOLE && fi->nh &&
+ nexthop_is_blackhole(fi->nh);
+}
+
bool fib_metrics_match(struct fib_config *cfg, struct fib_info *fi)
{
struct nlattr *nla;
diff --git a/net/ipv4/fib_trie.c b/net/ipv4/fib_trie.c
index 523a5a4228d0..c67df5aa1b96 100644
--- a/net/ipv4/fib_trie.c
+++ b/net/ipv4/fib_trie.c
@@ -1726,7 +1726,7 @@ int fib_table_delete(struct net *net, struct fib_table *tb,
(fa->fa_dscp != dscp))
break;

- if ((!cfg->fc_type || fa->fa_type == cfg->fc_type) &&
+ if (fib_type_match(cfg->fc_type, fa) &&
(cfg->fc_scope == RT_SCOPE_NOWHERE ||
fa->fa_info->fib_scope == cfg->fc_scope) &&
(!cfg->fc_prefsrc ||
--
2.53.0