[PATCH bpf-next v3 0/2] Fix bpf_get/setsockopt failed when TCP over IPv4 via INET6 API

[PATCH bpf-next v3 0/2] Fix bpf_get/setsockopt failed when TCP over IPv4 via INET6 API

[Feng zhou]

From: Feng Zhou <zhoufeng.zf@bytedance.com>

When TCP over IPv4 via INET6 API, sk->sk_family is AF_INET6, but it is a v4 pkt.
inet_csk(sk)->icsk_af_ops is ipv6_mapped and use ip_queue_xmit. Some sockopt did
not take effect, such as tos.

0001: Use sk_is_inet helper to fix it.
0002: Setget_sockopt add a test for tcp over ipv4 via ipv6.

Changelog:
v2->v3: Addressed comments from Eric Dumazet
- Use sk_is_inet() helper
Details in here:
https://lore.kernel.org/bpf/CANn89i+9GmBLCdgsfH=WWe-tyFYpiO27wONyxaxiU6aOBC6G8g@mail.gmail.com/T/

v1->v2: Addressed comments from kernel test robot
- Fix compilation error
Details in here:
https://lore.kernel.org/bpf/202408152058.YXAnhLgZ-lkp@intel.com/T/

Feng Zhou (2):
  bpf: Fix bpf_get/setsockopt to tos not take effect when TCP over IPv4
    via INET6 API
  selftests/bpf: Setget_sockopt add a test for tcp over ipv4 via ipv6

 net/core/filter.c                             |  7 +++-
 .../selftests/bpf/prog_tests/setget_sockopt.c | 33 +++++++++++++++++++
 .../selftests/bpf/progs/setget_sockopt.c      | 13 ++++++--
 3 files changed, 49 insertions(+), 4 deletions(-)

-- 
2.30.2

[PATCH bpf-next v3 1/2] bpf: Fix bpf_get/setsockopt to tos not take effect when TCP over IPv4 via INET6 API

[Feng zhou]

From: Feng Zhou <zhoufeng.zf@bytedance.com>

when TCP over IPv4 via INET6 API, bpf_get/setsockopt with ipv4 will
fail, because sk->sk_family is AF_INET6. With ipv6 will success, not
take effect, because inet_csk(sk)->icsk_af_ops is ipv6_mapped and
use ip_queue_xmit, inet_sk(sk)->tos.

Bpf_get/setsockopt use sk_is_inet() helper to fix this case.

Signed-off-by: Feng Zhou <zhoufeng.zf@bytedance.com>
---
 net/core/filter.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/net/core/filter.c b/net/core/filter.c
index e4a4454df5f9..90f4dbb8d2b5 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -5399,7 +5399,12 @@ static int sol_ip_sockopt(struct sock *sk, int optname,
 			  char *optval, int *optlen,
 			  bool getopt)
 {
-	if (sk->sk_family != AF_INET)
+
+	/*
+	 * SOL_IP socket options are available on AF_INET and AF_INET6, for
+	 * example, TCP over IPv4 via INET6 API.
+	 */
+	if (!sk_is_inet(sk))
 		return -EINVAL;
 
 	switch (optname) {
-- 
2.30.2

Re: [PATCH bpf-next v3 1/2] bpf: Fix bpf_get/setsockopt to tos not take effect when TCP over IPv4 via INET6 API

[Eric Dumazet]

On Sat, Sep 14, 2024 at 12:32 PM Feng zhou <zhoufeng.zf@bytedance.com> wrote:
>
> From: Feng Zhou <zhoufeng.zf@bytedance.com>
>
> when TCP over IPv4 via INET6 API, bpf_get/setsockopt with ipv4 will
> fail, because sk->sk_family is AF_INET6. With ipv6 will success, not
> take effect, because inet_csk(sk)->icsk_af_ops is ipv6_mapped and
> use ip_queue_xmit, inet_sk(sk)->tos.
>
> Bpf_get/setsockopt use sk_is_inet() helper to fix this case.
>
> Signed-off-by: Feng Zhou <zhoufeng.zf@bytedance.com>

Reviewed-by: Eric Dumazet <edumazet@google.com>

Re: [PATCH bpf-next v3 1/2] bpf: Fix bpf_get/setsockopt to tos not take effect when TCP over IPv4 via INET6 API

[Martin KaFai Lau]

On 9/14/24 3:32 AM, Feng zhou wrote:
> From: Feng Zhou <zhoufeng.zf@bytedance.com>
> 
> when TCP over IPv4 via INET6 API, bpf_get/setsockopt with ipv4 will

I think you meant bpf_get/setsockopt with SOL_IP will fail. so s/ipv4/SOL_IP/?

> fail, because sk->sk_family is AF_INET6. With ipv6 will success, not
> take effect, because inet_csk(sk)->icsk_af_ops is ipv6_mapped and
> use ip_queue_xmit, inet_sk(sk)->tos.

Change lgtm.

Patch 2 has a conflict, so can you please reword this commit message to reflect 
the latest change. e.g. afaik, this is no longer specific to mapped address or not.

> 
> Bpf_get/setsockopt use sk_is_inet() helper to fix this case.
> 
> Signed-off-by: Feng Zhou <zhoufeng.zf@bytedance.com>
> ---
>   net/core/filter.c | 7 ++++++-
>   1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/net/core/filter.c b/net/core/filter.c
> index e4a4454df5f9..90f4dbb8d2b5 100644
> --- a/net/core/filter.c
> +++ b/net/core/filter.c
> @@ -5399,7 +5399,12 @@ static int sol_ip_sockopt(struct sock *sk, int optname,
>   			  char *optval, int *optlen,
>   			  bool getopt)
>   {
> -	if (sk->sk_family != AF_INET)
> +
> +	/*
> +	 * SOL_IP socket options are available on AF_INET and AF_INET6, for
> +	 * example, TCP over IPv4 via INET6 API.
> +	 */
> +	if (!sk_is_inet(sk))
>   		return -EINVAL;
>   
>   	switch (optname) {

Re: [External] Re: [PATCH bpf-next v3 1/2] bpf: Fix bpf_get/setsockopt to tos not take effect when TCP over IPv4 via INET6 API

[Feng Zhou]

在 2024/10/1 10:27, Martin KaFai Lau 写道:
> On 9/14/24 3:32 AM, Feng zhou wrote:
>> From: Feng Zhou <zhoufeng.zf@bytedance.com>
>>
>> when TCP over IPv4 via INET6 API, bpf_get/setsockopt with ipv4 will
> 
> I think you meant bpf_get/setsockopt with SOL_IP will fail. so 
> s/ipv4/SOL_IP/?
> 
>> fail, because sk->sk_family is AF_INET6. With ipv6 will success, not
>> take effect, because inet_csk(sk)->icsk_af_ops is ipv6_mapped and
>> use ip_queue_xmit, inet_sk(sk)->tos.
> 
> Change lgtm.
> 
> Patch 2 has a conflict, so can you please reword this commit message to 
> reflect the latest change. e.g. afaik, this is no longer specific to 
> mapped address or not.
> 

Sorry for taking so long to reply.

Will do, thanks.

>>
>> Bpf_get/setsockopt use sk_is_inet() helper to fix this case.
>>
>> Signed-off-by: Feng Zhou <zhoufeng.zf@bytedance.com>
>> ---
>>   net/core/filter.c | 7 ++++++-
>>   1 file changed, 6 insertions(+), 1 deletion(-)
>>
>> diff --git a/net/core/filter.c b/net/core/filter.c
>> index e4a4454df5f9..90f4dbb8d2b5 100644
>> --- a/net/core/filter.c
>> +++ b/net/core/filter.c
>> @@ -5399,7 +5399,12 @@ static int sol_ip_sockopt(struct sock *sk, int 
>> optname,
>>                 char *optval, int *optlen,
>>                 bool getopt)
>>   {
>> -    if (sk->sk_family != AF_INET)
>> +
>> +    /*
>> +     * SOL_IP socket options are available on AF_INET and AF_INET6, for
>> +     * example, TCP over IPv4 via INET6 API.
>> +     */
>> +    if (!sk_is_inet(sk))
>>           return -EINVAL;
>>       switch (optname) {
> 

[PATCH bpf-next v3 2/2] selftests/bpf: Setget_sockopt add a test for tcp over ipv4 via ipv6

[Feng zhou]

From: Feng Zhou <zhoufeng.zf@bytedance.com>

This patch adds a test for TCP over IPv4 via INET6 API.

Signed-off-by: Feng Zhou <zhoufeng.zf@bytedance.com>
---
 .../selftests/bpf/prog_tests/setget_sockopt.c | 33 +++++++++++++++++++
 .../selftests/bpf/progs/setget_sockopt.c      | 13 ++++++--
 2 files changed, 43 insertions(+), 3 deletions(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/setget_sockopt.c b/tools/testing/selftests/bpf/prog_tests/setget_sockopt.c
index 7d4a9b3d3722..3cad92128e60 100644
--- a/tools/testing/selftests/bpf/prog_tests/setget_sockopt.c
+++ b/tools/testing/selftests/bpf/prog_tests/setget_sockopt.c
@@ -15,8 +15,11 @@
 
 #define CG_NAME "/setget-sockopt-test"
 
+#define INT_PORT	8008
+
 static const char addr4_str[] = "127.0.0.1";
 static const char addr6_str[] = "::1";
+static const char addr6_any_str[] = "::";
 static struct setget_sockopt *skel;
 static int cg_fd;
 
@@ -67,6 +70,35 @@ static void test_tcp(int family)
 	ASSERT_EQ(bss->nr_binddev, 2, "nr_bind");
 }
 
+static void test_tcp_over_ipv4_via_ipv6(void)
+{
+	struct setget_sockopt__bss *bss = skel->bss;
+	int sfd, cfd;
+
+	memset(bss, 0, sizeof(*bss));
+	skel->bss->test_tcp_over_ipv4_via_ipv6 = 1;
+
+	sfd = start_server(AF_INET6, SOCK_STREAM,
+			   addr6_any_str, INT_PORT, 0);
+	if (!ASSERT_GE(sfd, 0, "start_server"))
+		return;
+
+	cfd = connect_to_addr_str(AF_INET, SOCK_STREAM, addr4_str, INT_PORT, NULL);
+	if (!ASSERT_GE(cfd, 0, "connect_to_addr_str")) {
+		close(sfd);
+		return;
+	}
+	close(sfd);
+	close(cfd);
+
+	ASSERT_EQ(bss->nr_listen, 1, "nr_listen");
+	ASSERT_EQ(bss->nr_connect, 1, "nr_connect");
+	ASSERT_EQ(bss->nr_active, 1, "nr_active");
+	ASSERT_EQ(bss->nr_passive, 1, "nr_passive");
+	ASSERT_EQ(bss->nr_socket_post_create, 2, "nr_socket_post_create");
+	ASSERT_EQ(bss->nr_binddev, 2, "nr_bind");
+}
+
 static void test_udp(int family)
 {
 	struct setget_sockopt__bss *bss = skel->bss;
@@ -191,6 +223,7 @@ void test_setget_sockopt(void)
 	test_udp(AF_INET);
 	test_ktls(AF_INET6);
 	test_ktls(AF_INET);
+	test_tcp_over_ipv4_via_ipv6();
 
 done:
 	setget_sockopt__destroy(skel);
diff --git a/tools/testing/selftests/bpf/progs/setget_sockopt.c b/tools/testing/selftests/bpf/progs/setget_sockopt.c
index 60518aed1ffc..ff834d94dd23 100644
--- a/tools/testing/selftests/bpf/progs/setget_sockopt.c
+++ b/tools/testing/selftests/bpf/progs/setget_sockopt.c
@@ -20,6 +20,7 @@ int nr_connect;
 int nr_binddev;
 int nr_socket_post_create;
 int nr_fin_wait1;
+int test_tcp_over_ipv4_via_ipv6;
 
 struct sockopt_test {
 	int opt;
@@ -262,9 +263,15 @@ static int bpf_test_sockopt(void *ctx, struct sock *sk)
 		if (n != ARRAY_SIZE(sol_ip_tests))
 			return -1;
 	} else {
-		n = bpf_loop(ARRAY_SIZE(sol_ipv6_tests), bpf_test_ipv6_sockopt, &lc, 0);
-		if (n != ARRAY_SIZE(sol_ipv6_tests))
-			return -1;
+		if (test_tcp_over_ipv4_via_ipv6) {
+			n = bpf_loop(ARRAY_SIZE(sol_ip_tests), bpf_test_ip_sockopt, &lc, 0);
+			if (n != ARRAY_SIZE(sol_ip_tests))
+				return -1;
+		} else {
+			n = bpf_loop(ARRAY_SIZE(sol_ipv6_tests), bpf_test_ipv6_sockopt, &lc, 0);
+			if (n != ARRAY_SIZE(sol_ipv6_tests))
+				return -1;
+		}
 	}
 
 	return 0;
-- 
2.30.2

Re: [PATCH bpf-next v3 2/2] selftests/bpf: Setget_sockopt add a test for tcp over ipv4 via ipv6

[Martin KaFai Lau]

On 9/14/24 3:32 AM, Feng zhou wrote:
> From: Feng Zhou <zhoufeng.zf@bytedance.com>
> 
> This patch adds a test for TCP over IPv4 via INET6 API.
> 
> Signed-off-by: Feng Zhou <zhoufeng.zf@bytedance.com>
> ---
>   .../selftests/bpf/prog_tests/setget_sockopt.c | 33 +++++++++++++++++++
>   .../selftests/bpf/progs/setget_sockopt.c      | 13 ++++++--
>   2 files changed, 43 insertions(+), 3 deletions(-)
> 
> diff --git a/tools/testing/selftests/bpf/prog_tests/setget_sockopt.c b/tools/testing/selftests/bpf/prog_tests/setget_sockopt.c
> index 7d4a9b3d3722..3cad92128e60 100644
> --- a/tools/testing/selftests/bpf/prog_tests/setget_sockopt.c
> +++ b/tools/testing/selftests/bpf/prog_tests/setget_sockopt.c
> @@ -15,8 +15,11 @@
>   
>   #define CG_NAME "/setget-sockopt-test"
>   
> +#define INT_PORT	8008
> +
>   static const char addr4_str[] = "127.0.0.1";
>   static const char addr6_str[] = "::1";
> +static const char addr6_any_str[] = "::";
>   static struct setget_sockopt *skel;
>   static int cg_fd;
>   
> @@ -67,6 +70,35 @@ static void test_tcp(int family)
>   	ASSERT_EQ(bss->nr_binddev, 2, "nr_bind");
>   }
>   
> +static void test_tcp_over_ipv4_via_ipv6(void)
> +{
> +	struct setget_sockopt__bss *bss = skel->bss;
> +	int sfd, cfd;
> +
> +	memset(bss, 0, sizeof(*bss));
> +	skel->bss->test_tcp_over_ipv4_via_ipv6 = 1;
> +
> +	sfd = start_server(AF_INET6, SOCK_STREAM,
> +			   addr6_any_str, INT_PORT, 0);
> +	if (!ASSERT_GE(sfd, 0, "start_server"))
> +		return;
> +
> +	cfd = connect_to_addr_str(AF_INET, SOCK_STREAM, addr4_str, INT_PORT, NULL);
> +	if (!ASSERT_GE(cfd, 0, "connect_to_addr_str")) {
> +		close(sfd);
> +		return;
> +	}
> +	close(sfd);
> +	close(cfd);
> +
> +	ASSERT_EQ(bss->nr_listen, 1, "nr_listen");
> +	ASSERT_EQ(bss->nr_connect, 1, "nr_connect");
> +	ASSERT_EQ(bss->nr_active, 1, "nr_active");
> +	ASSERT_EQ(bss->nr_passive, 1, "nr_passive");
> +	ASSERT_EQ(bss->nr_socket_post_create, 2, "nr_socket_post_create");
> +	ASSERT_EQ(bss->nr_binddev, 2, "nr_bind");
> +}
> +
>   static void test_udp(int family)
>   {
>   	struct setget_sockopt__bss *bss = skel->bss;
> @@ -191,6 +223,7 @@ void test_setget_sockopt(void)
>   	test_udp(AF_INET);
>   	test_ktls(AF_INET6);
>   	test_ktls(AF_INET);
> +	test_tcp_over_ipv4_via_ipv6();

This has a conflict with commit d53050934e66.

pw-bot: cr

>   
>   done:
>   	setget_sockopt__destroy(skel);
> diff --git a/tools/testing/selftests/bpf/progs/setget_sockopt.c b/tools/testing/selftests/bpf/progs/setget_sockopt.c
> index 60518aed1ffc..ff834d94dd23 100644
> --- a/tools/testing/selftests/bpf/progs/setget_sockopt.c
> +++ b/tools/testing/selftests/bpf/progs/setget_sockopt.c
> @@ -20,6 +20,7 @@ int nr_connect;
>   int nr_binddev;
>   int nr_socket_post_create;
>   int nr_fin_wait1;
> +int test_tcp_over_ipv4_via_ipv6;
>   
>   struct sockopt_test {
>   	int opt;
> @@ -262,9 +263,15 @@ static int bpf_test_sockopt(void *ctx, struct sock *sk)
>   		if (n != ARRAY_SIZE(sol_ip_tests))
>   			return -1;
>   	} else {
> -		n = bpf_loop(ARRAY_SIZE(sol_ipv6_tests), bpf_test_ipv6_sockopt, &lc, 0);
> -		if (n != ARRAY_SIZE(sol_ipv6_tests))
> -			return -1;
> +		if (test_tcp_over_ipv4_via_ipv6) {
> +			n = bpf_loop(ARRAY_SIZE(sol_ip_tests), bpf_test_ip_sockopt, &lc, 0);

Can this bpf_loop(..., bpf_test_ip_sockopt, ...) be always run? Then the above 
test_tcp_over_ipv4_via_ipv6() addition will not be needed.


> +			if (n != ARRAY_SIZE(sol_ip_tests))
> +				return -1;
> +		} else {
> +			n = bpf_loop(ARRAY_SIZE(sol_ipv6_tests), bpf_test_ipv6_sockopt, &lc, 0);
> +			if (n != ARRAY_SIZE(sol_ipv6_tests))
> +				return -1;
> +		}
>   	}
>   
>   	return 0;