Re: [PATCH bpf v3 2/2] selftests/bpf: Add protocol check test for bpf_sk_assign_tcp_reqsk()

[Jiayuan Chen <jiayuan.chen@linux.dev>]

On 3/28/26 3:53 AM, Martin KaFai Lau wrote:
>> From: Jiayuan Chen <jiayuan.chen@shopee.com>
>>
>> Add test_tcp_custom_syncookie_protocol_check to verify that
>> bpf_sk_assign_tcp_reqsk() rejects non-TCP skbs. The test sends a UDP
>> packet through TC ingress where a BPF program calls
>> bpf_sk_assign_tcp_reqsk() on it and checks that the kfunc returns an
>> error. A UDP server recv() is used as synchronization to ensure the
>> BPF program has finished processing before checking the result.
>>
>> Without the fix in bpf_sk_assign_tcp_reqsk(), the kfunc succeeds and
>> attaches a TCP reqsk to the UDP skb, which causes a null pointer
>> dereference panic when the kernel processes it through the UDP receive
>> path.
>>
>> Test result:
>>
>>    ./test_progs -a tcp_custom_syncookie_protocol_check -v
>>    setup_netns:PASS:create netns 0 nsec
>>    setup_netns:PASS:ip 0 nsec
>>    write_sysctl:PASS:open sysctl 0 nsec
>>    write_sysctl:PASS:write sysctl 0 nsec
>>    setup_netns:PASS:write_sysctl 0 nsec
>>    test_tcp_custom_syncookie_protocol_check:PASS:open_and_load 0 nsec
>>    setup_tc:PASS:qdisc add dev lo clsact 0 nsec
>>    setup_tc:PASS:filter add dev lo ingress 0 nsec
>>    run_protocol_check:PASS:start tcp_server 0 nsec
>>    run_protocol_check:PASS:start udp_server 0 nsec
>>    run_protocol_check:PASS:connect udp_client 0 nsec
>>    run_protocol_check:PASS:send udp 0 nsec
>>    run_protocol_check:PASS:recv udp 0 nsec
>>    run_protocol_check:PASS:udp_intercepted 0 nsec
>>    run_protocol_check:PASS:assign_ret 0 nsec
>>    #471/1   tcp_custom_syncookie_protocol_check/IPv4 TCP:OK
>>    run_protocol_check:PASS:start tcp_server 0 nsec
>>    run_protocol_check:PASS:start udp_server 0 nsec
>>    run_protocol_check:PASS:connect udp_client 0 nsec
>>    run_protocol_check:PASS:send udp 0 nsec
>>    run_protocol_check:PASS:recv udp 0 nsec
>>    run_protocol_check:PASS:udp_intercepted 0 nsec
>>    run_protocol_check:PASS:assign_ret 0 nsec
>>    #471/2   tcp_custom_syncookie_protocol_check/IPv6 TCP:OK
>>    #471     tcp_custom_syncookie_protocol_check:OK
>>    Summary: 1/2 PASSED, 0 SKIPPED, 0 FAILED
>>
>> Cc: Jiayuan Chen <jiayuan.chen@linux.dev>
>> Signed-off-by: Jiayuan Chen <jiayuan.chen@shopee.com>
>> ---
>>   .../bpf/prog_tests/tcp_custom_syncookie.c     |  94 ++++++++++++++-
>>   .../bpf/progs/test_tcp_custom_syncookie.c     | 109 ++++++++++++++++++
>>   2 files changed, 199 insertions(+), 4 deletions(-)
>>
>> diff --git 
>> a/tools/testing/selftests/bpf/prog_tests/tcp_custom_syncookie.c 
>> b/tools/testing/selftests/bpf/prog_tests/tcp_custom_syncookie.c
>> index eaf441dc7e79..e46031d0786b 100644
>> --- a/tools/testing/selftests/bpf/prog_tests/tcp_custom_syncookie.c
>> +++ b/tools/testing/selftests/bpf/prog_tests/tcp_custom_syncookie.c
>> @@ -5,6 +5,7 @@
>>   #include <sched.h>
>>   #include <stdlib.h>
>>   #include <net/if.h>
>> +#include <netinet/in.h>
>>     #include "test_progs.h"
>>   #include "cgroup_helpers.h"
>> @@ -47,11 +48,10 @@ static int setup_netns(void)
>>       return -1;
>>   }
>>   -static int setup_tc(struct test_tcp_custom_syncookie *skel)
>> +static int setup_tc(int prog_fd)
>>   {
>>       LIBBPF_OPTS(bpf_tc_hook, qdisc_lo, .attach_point = 
>> BPF_TC_INGRESS);
>> -    LIBBPF_OPTS(bpf_tc_opts, tc_attach,
>> -            .prog_fd = 
>> bpf_program__fd(skel->progs.tcp_custom_syncookie));
>> +    LIBBPF_OPTS(bpf_tc_opts, tc_attach, .prog_fd = prog_fd);
>>         qdisc_lo.ifindex = if_nametoindex("lo");
>>       if (!ASSERT_OK(bpf_tc_hook_create(&qdisc_lo), "qdisc add dev lo 
>> clsact"))
>> @@ -127,7 +127,7 @@ void test_tcp_custom_syncookie(void)
>>       if (!ASSERT_OK_PTR(skel, "open_and_load"))
>>           return;
>>   -    if (setup_tc(skel))
>> +    if (setup_tc(bpf_program__fd(skel->progs.tcp_custom_syncookie)))
>>           goto destroy_skel;
>>         for (i = 0; i < ARRAY_SIZE(test_cases); i++) {
>> @@ -145,6 +145,92 @@ void test_tcp_custom_syncookie(void)
>>     destroy_skel:
>>       system("tc qdisc del dev lo clsact");
>> +    test_tcp_custom_syncookie__destroy(skel);
>> +}
>> +
>> +/* Test: bpf_sk_assign_tcp_reqsk() should reject non-TCP skb.
>> + *
>> + * Send a UDP packet through TC ingress where a BPF program calls
>> + * bpf_sk_assign_tcp_reqsk() on it. The kfunc should return an error
>> + * because the skb carries UDP, not TCP.
>> + *
>> + * TCP and UDP servers share the same port. The BPF program intercepts
>> + * the UDP packet, looks up the TCP listener via the dest port, and
>> + * attempts to assign a TCP reqsk to the UDP skb.
>> + */
>> +static void run_protocol_check(struct test_tcp_custom_syncookie *skel,
>> +                   int family, const char *addr)
>> +{
>> +    int tcp_server = -1, udp_server = -1, udp_client = -1;
>> +    char buf[32] = "test";
>> +    int port, ret;
>> +
>> +    tcp_server = start_server(family, SOCK_STREAM, addr, 0, 0);
>> +    if (!ASSERT_NEQ(tcp_server, -1, "start tcp_server"))
>> +        return;
>> +
>> +    port = ntohs(get_socket_local_port(tcp_server));
>> +
>> +    /* UDP server on same port for synchronization and port sharing */
>> +    udp_server = start_server(family, SOCK_DGRAM, addr, port, 0);
>> +    if (!ASSERT_NEQ(udp_server, -1, "start udp_server"))
>> +        goto close_tcp;
>> +
>> +    skel->bss->udp_intercepted = false;
>> +    skel->data->assign_ret = -1;
>
> assign_ret is init to -1
>
>> +
>> +    udp_client = connect_to_fd(udp_server, 0);
>> +    if (!ASSERT_NEQ(udp_client, -1, "connect udp_client"))
>> +        goto close_udp_server;
>>   +    ret = send(udp_client, buf, sizeof(buf), 0);
>> +    if (!ASSERT_EQ(ret, sizeof(buf), "send udp"))
>> +        goto close_udp_client;
>> +
>> +    /* recv() ensures TC ingress BPF has processed the skb */
>> +    ret = recv(udp_server, buf, sizeof(buf), 0);
>> +    if (!ASSERT_EQ(ret, sizeof(buf), "recv udp"))
>> +        goto close_udp_client;
>> +
>> +    ASSERT_EQ(skel->bss->udp_intercepted, true, "udp_intercepted");
>> +
>> +    /* assign_ret == 0 means kfunc accepted UDP skb (bug).
>> +     * assign_ret < 0 means kfunc correctly rejected it (fixed).
>> +     */
>> +    ASSERT_NEQ(skel->data->assign_ret, 0, "assign_ret");
>
> and here it checks NEQ 0....
>
> Maybe init assign_ret to 0?
>
> Also, why not specifically check for -EINVAL?


You're right, checking for -EINVAL is more precise and avoids hiding a 
silent failure in the lookup path.

Will init assign_ret to 0 and check for -EINVAL instead.