]> git.ipfire.org Git - thirdparty/kernel/linux.git/commitdiff
selftests/bpf: cover short IPv4/IPv6 inputs with adjust_room
authorSun Jian <sun.jian.kdev@gmail.com>
Wed, 8 Apr 2026 03:46:23 +0000 (11:46 +0800)
committerAlexei Starovoitov <ast@kernel.org>
Sun, 12 Apr 2026 22:42:57 +0000 (15:42 -0700)
Add a selftest covering ETH_HLEN-sized IPv4/IPv6 EtherType inputs for
bpf_prog_test_run_skb().

Reuse a single zero-initialized struct ethhdr eth_hlen and set
eth_hlen.h_proto from the per-test h_proto field.

Also add a dedicated tc_adjust_room program and route the short
IPv4/IPv6 cases to it, so the selftest actually exercises the
bpf_skb_adjust_room() path from the report.

Signed-off-by: Sun Jian <sun.jian.kdev@gmail.com>
Link: https://lore.kernel.org/r/20260408034623.180320-3-sun.jian.kdev@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
tools/testing/selftests/bpf/prog_tests/empty_skb.c
tools/testing/selftests/bpf/progs/empty_skb.c

index 438583e1f2d12151d9f17b3f096e5a6af1c12dfe..c9fcb70cbafbce571d7826e233920193625dead1 100644 (file)
@@ -10,8 +10,8 @@ void test_empty_skb(void)
        struct empty_skb *bpf_obj = NULL;
        struct nstoken *tok = NULL;
        struct bpf_program *prog;
+       struct ethhdr eth_hlen;
        char eth_hlen_pp[15];
-       char eth_hlen[14];
        int veth_ifindex;
        int ipip_ifindex;
        int err;
@@ -25,7 +25,9 @@ void test_empty_skb(void)
                int err;
                int ret;
                int lwt_egress_ret; /* expected retval at lwt/egress */
+               __be16 h_proto;
                bool success_on_tc;
+               bool adjust_room;
        } tests[] = {
                /* Empty packets are always rejected. */
 
@@ -46,6 +48,28 @@ void test_empty_skb(void)
                        .err = -EINVAL,
                },
 
+               /* ETH_HLEN-sized packets with IPv4/IPv6 EtherType but
+                * no L3 header are rejected.
+                */
+               {
+                       .msg = "veth short IPv4 ingress packet",
+                       .data_in = &eth_hlen,
+                       .data_size_in = sizeof(eth_hlen),
+                       .ifindex = &veth_ifindex,
+                       .err = -EINVAL,
+                       .h_proto = htons(ETH_P_IP),
+                       .adjust_room = true,
+               },
+               {
+                       .msg = "veth short IPv6 ingress packet",
+                       .data_in = &eth_hlen,
+                       .data_size_in = sizeof(eth_hlen),
+                       .ifindex = &veth_ifindex,
+                       .err = -EINVAL,
+                       .h_proto = htons(ETH_P_IPV6),
+                       .adjust_room = true,
+               },
+
                /* ETH_HLEN-sized packets:
                 * - can not be redirected at LWT_XMIT
                 * - can be redirected at TC to non-tunneling dest
@@ -54,7 +78,7 @@ void test_empty_skb(void)
                {
                        /* __bpf_redirect_common */
                        .msg = "veth ETH_HLEN packet ingress",
-                       .data_in = eth_hlen,
+                       .data_in = &eth_hlen,
                        .data_size_in = sizeof(eth_hlen),
                        .ifindex = &veth_ifindex,
                        .ret = -ERANGE,
@@ -68,7 +92,7 @@ void test_empty_skb(void)
                         * tc: skb->len=14 <= skb_network_offset=14
                         */
                        .msg = "ipip ETH_HLEN packet ingress",
-                       .data_in = eth_hlen,
+                       .data_in = &eth_hlen,
                        .data_size_in = sizeof(eth_hlen),
                        .ifindex = &ipip_ifindex,
                        .ret = -ERANGE,
@@ -108,17 +132,27 @@ void test_empty_skb(void)
        SYS(out, "ip addr add 192.168.1.1/16 dev ipip0");
        ipip_ifindex = if_nametoindex("ipip0");
 
+       memset(eth_hlen_pp, 0, sizeof(eth_hlen_pp));
+       memset(&eth_hlen, 0, sizeof(eth_hlen));
+
        bpf_obj = empty_skb__open_and_load();
        if (!ASSERT_OK_PTR(bpf_obj, "open skeleton"))
                goto out;
 
        for (i = 0; i < ARRAY_SIZE(tests); i++) {
+               if (tests[i].data_in == &eth_hlen)
+                       eth_hlen.h_proto = tests[i].h_proto;
+
                bpf_object__for_each_program(prog, bpf_obj->obj) {
                        bool at_egress = strstr(bpf_program__name(prog), "egress") != NULL;
                        bool at_tc = !strncmp(bpf_program__section_name(prog), "tc", 2);
+                       bool is_adjust_room = !strcmp(bpf_program__name(prog), "tc_adjust_room");
                        int expected_ret;
                        char buf[128];
 
+                       if (tests[i].adjust_room != is_adjust_room)
+                               continue;
+
                        expected_ret = at_egress && !at_tc ? tests[i].lwt_egress_ret : tests[i].ret;
 
                        tattr.data_in = tests[i].data_in;
index 4b0cd675325119ad163b29def109b5c10db46864..44326f5cc8bb4b381c3288173a196f1cec3b514b 100644 (file)
@@ -35,3 +35,10 @@ int tc_redirect_egress(struct __sk_buff *skb)
        ret = bpf_clone_redirect(skb, ifindex, 0);
        return 0;
 }
+
+SEC("tc")
+int tc_adjust_room(struct __sk_buff *skb)
+{
+       ret = bpf_skb_adjust_room(skb, 4, BPF_ADJ_ROOM_NET, 0);
+       return 0;
+}