[PATCH net-next 0/2] selftest/net: Some more TCP-AO selftest post-merge fixups

[PATCH net-next 0/2] selftest/net: Some more TCP-AO selftest post-merge fixups

[Dmitry Safonov]

Note that there's another post-merge fix for TCP-AO selftests, but that
doesn't conflict with these, so I don't resend that:

https://lore.kernel.org/all/20231219-b4-tcp-ao-selftests-out-of-tree-v1-1-0fff92d26eac@arista.com/T/#u

Signed-off-by: Dmitry Safonov <dima@arista.com>
---
Dmitry Safonov (2):
      selftest/tcp-ao: Set routes in a proper VRF table id
      selftest/tcp-ao: Work on namespace-ified sysctl_optmem_max

 tools/testing/selftests/net/tcp_ao/bench-lookups.c |  4 ++-
 tools/testing/selftests/net/tcp_ao/lib/netlink.c   |  4 +--
 tools/testing/selftests/net/tcp_ao/lib/setup.c     | 35 +++++++++++++++++-----
 tools/testing/selftests/net/tcp_ao/unsigned-md5.c  | 11 ++++---
 4 files changed, 36 insertions(+), 18 deletions(-)
---
base-commit: 857647efa9be89a13cf8963c7e167fab062b28bb
change-id: 20231222-selftests-tcp-ao-fixups-ce70a60e6f57

Best regards,
-- 
Dmitry Safonov <dima@arista.com>

[PATCH net-next 1/2] selftest/tcp-ao: Set routes in a proper VRF table id

[Dmitry Safonov]

In unsigned-md5 selftests ip_route_add() is not needed in
client_add_ip(): the route was pre-setup in __test_init() => link_init()
for subnet, rather than a specific ip-address.

Currently, __ip_route_add() mistakenly always sets VRF table
to RT_TABLE_MAIN - this seems to have sneaked in during unsigned-md5
tests debugging. That also explains, why ip_route_add_vrf() ignored
EEXIST, returned by fib6.

Yet, keep EEXIST ignoring in bench-lookups selftests as it's expected
that those selftests may add the same (duplicate) routes.

Reported-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Dmitry Safonov <dima@arista.com>
---
 tools/testing/selftests/net/tcp_ao/bench-lookups.c |  4 +++-
 tools/testing/selftests/net/tcp_ao/lib/netlink.c   |  4 +---
 tools/testing/selftests/net/tcp_ao/unsigned-md5.c  | 11 +++++------
 3 files changed, 9 insertions(+), 10 deletions(-)

diff --git a/tools/testing/selftests/net/tcp_ao/bench-lookups.c b/tools/testing/selftests/net/tcp_ao/bench-lookups.c
index 7be8a7d9308c..a1e6e007c291 100644
--- a/tools/testing/selftests/net/tcp_ao/bench-lookups.c
+++ b/tools/testing/selftests/net/tcp_ao/bench-lookups.c
@@ -46,8 +46,10 @@ static void test_add_routes(union tcp_addr *ips, size_t ips_nr)
 
 	for (i = 0; i < ips_nr; i++) {
 		union tcp_addr *p = (union tcp_addr *)&ips[i];
+		int err;
 
-		if (ip_route_add(veth_name, TEST_FAMILY, this_ip_addr, *p))
+		err = ip_route_add(veth_name, TEST_FAMILY, this_ip_addr, *p);
+		if (err && err != -EEXIST)
 			test_error("Failed to add route");
 	}
 }
diff --git a/tools/testing/selftests/net/tcp_ao/lib/netlink.c b/tools/testing/selftests/net/tcp_ao/lib/netlink.c
index b731f2c84083..7f108493a29a 100644
--- a/tools/testing/selftests/net/tcp_ao/lib/netlink.c
+++ b/tools/testing/selftests/net/tcp_ao/lib/netlink.c
@@ -261,7 +261,7 @@ static int __ip_route_add(int sock, uint32_t seq, const char *intf, int family,
 	req.nh.nlmsg_seq	= seq;
 	req.rt.rtm_family	= family;
 	req.rt.rtm_dst_len	= (family == AF_INET) ? 32 : 128;
-	req.rt.rtm_table	= RT_TABLE_MAIN;
+	req.rt.rtm_table	= vrf;
 	req.rt.rtm_protocol	= RTPROT_BOOT;
 	req.rt.rtm_scope	= RT_SCOPE_UNIVERSE;
 	req.rt.rtm_type		= RTN_UNICAST;
@@ -294,8 +294,6 @@ int ip_route_add_vrf(const char *intf, int family,
 
 	ret = __ip_route_add(route_sock, route_seq++, intf,
 			     family, src, dst, vrf);
-	if (ret == -EEXIST) /* ignoring */
-		ret = 0;
 
 	close(route_sock);
 	return ret;
diff --git a/tools/testing/selftests/net/tcp_ao/unsigned-md5.c b/tools/testing/selftests/net/tcp_ao/unsigned-md5.c
index 7cffde02d2be..f5b6d488d501 100644
--- a/tools/testing/selftests/net/tcp_ao/unsigned-md5.c
+++ b/tools/testing/selftests/net/tcp_ao/unsigned-md5.c
@@ -30,7 +30,7 @@ static void setup_vrfs(void)
 	err = ip_route_add_vrf(veth_name, TEST_FAMILY,
 			       this_ip_addr, this_ip_dest, test_vrf_tabid);
 	if (err)
-		test_error("Failed to add a route to VRF");
+		test_error("Failed to add a route to VRF: %d", err);
 }
 
 static void try_accept(const char *tst_name, unsigned int port,
@@ -494,15 +494,14 @@ static void try_to_add(const char *tst_name, unsigned int port,
 
 static void client_add_ip(union tcp_addr *client, const char *ip)
 {
-	int family = TEST_FAMILY;
+	int err, family = TEST_FAMILY;
 
 	if (inet_pton(family, ip, client) != 1)
 		test_error("Can't convert ip address %s", ip);
 
-	if (ip_addr_add(veth_name, family, *client, TEST_PREFIX))
-		test_error("Failed to add ip address");
-	if (ip_route_add(veth_name, family, *client, this_ip_dest))
-		test_error("Failed to add route");
+	err = ip_addr_add(veth_name, family, *client, TEST_PREFIX);
+	if (err)
+		test_error("Failed to add ip address: %d", err);
 }
 
 static void client_add_ips(void)

-- 
2.43.0

[PATCH net-next 2/2] selftest/tcp-ao: Work on namespace-ified sysctl_optmem_max

[Dmitry Safonov]

Since commit f5769faeec36 ("net: Namespace-ify sysctl_optmem_max")
optmem_max is per-netns, so need of switching to root namespace.
It seems trivial to keep the old logic working, so going to keep it for
a while (at least, until kernel with netns-optmem_max will be release).

Currently, there is a test that checks that optmem_max limit applies to
TCP-AO keys and a little benchmark that measures linked-list TCP-AO keys
scaling, those are fixed by this.

Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Dmitry Safonov <dima@arista.com>
---
 tools/testing/selftests/net/tcp_ao/lib/setup.c | 35 ++++++++++++++++++++------
 1 file changed, 27 insertions(+), 8 deletions(-)

diff --git a/tools/testing/selftests/net/tcp_ao/lib/setup.c b/tools/testing/selftests/net/tcp_ao/lib/setup.c
index 374b27c26ebd..92276f916f2f 100644
--- a/tools/testing/selftests/net/tcp_ao/lib/setup.c
+++ b/tools/testing/selftests/net/tcp_ao/lib/setup.c
@@ -277,22 +277,38 @@ void __test_init(unsigned int ntests, int family, unsigned int prefix,
 
 /* /proc/sys/net/core/optmem_max artifically limits the amount of memory
  * that can be allocated with sock_kmalloc() on each socket in the system.
- * It is not virtualized, so it has to written outside test namespaces.
- * To be nice a test will revert optmem back to the old value.
+ * It is not virtualized in v6.7, so it has to written outside test
+ * namespaces. To be nice a test will revert optmem back to the old value.
  * Keeping it simple without any file lock, which means the tests that
  * need to set/increase optmem value shouldn't run in parallel.
  * Also, not re-entrant.
+ * Since commit f5769faeec36 ("net: Namespace-ify sysctl_optmem_max")
+ * it is per-namespace, keeping logic for non-virtualized optmem_max
+ * for v6.7, which supports TCP-AO.
  */
 static const char *optmem_file = "/proc/sys/net/core/optmem_max";
 static size_t saved_optmem;
+static int optmem_ns = -1;
+
+static bool is_optmem_namespaced(void)
+{
+	if (optmem_ns == -1) {
+		int old_ns = switch_save_ns(nsfd_child);
+
+		optmem_ns = !access(optmem_file, F_OK);
+		switch_ns(old_ns);
+	}
+	return !!optmem_ns;
+}
 
 size_t test_get_optmem(void)
 {
+	int old_ns = 0;
 	FILE *foptmem;
-	int old_ns;
 	size_t ret;
 
-	old_ns = switch_save_ns(nsfd_outside);
+	if (!is_optmem_namespaced())
+		old_ns = switch_save_ns(nsfd_outside);
 	foptmem = fopen(optmem_file, "r");
 	if (!foptmem)
 		test_error("failed to open %s", optmem_file);
@@ -300,19 +316,21 @@ size_t test_get_optmem(void)
 	if (fscanf(foptmem, "%zu", &ret) != 1)
 		test_error("can't read from %s", optmem_file);
 	fclose(foptmem);
-	switch_ns(old_ns);
+	if (!is_optmem_namespaced())
+		switch_ns(old_ns);
 	return ret;
 }
 
 static void __test_set_optmem(size_t new, size_t *old)
 {
+	int old_ns = 0;
 	FILE *foptmem;
-	int old_ns;
 
 	if (old != NULL)
 		*old = test_get_optmem();
 
-	old_ns = switch_save_ns(nsfd_outside);
+	if (!is_optmem_namespaced())
+		old_ns = switch_save_ns(nsfd_outside);
 	foptmem = fopen(optmem_file, "w");
 	if (!foptmem)
 		test_error("failed to open %s", optmem_file);
@@ -320,7 +338,8 @@ static void __test_set_optmem(size_t new, size_t *old)
 	if (fprintf(foptmem, "%zu", new) <= 0)
 		test_error("can't write %zu to %s", new, optmem_file);
 	fclose(foptmem);
-	switch_ns(old_ns);
+	if (!is_optmem_namespaced())
+		switch_ns(old_ns);
 }
 
 static void test_revert_optmem(void)

-- 
2.43.0

Re: [PATCH net-next 0/2] selftest/net: Some more TCP-AO selftest post-merge fixups

[Hangbin Liu]

On Fri, Dec 22, 2023 at 01:59:05AM +0000, Dmitry Safonov wrote:
> Note that there's another post-merge fix for TCP-AO selftests, but that
> doesn't conflict with these, so I don't resend that:
> 
> https://lore.kernel.org/all/20231219-b4-tcp-ao-selftests-out-of-tree-v1-1-0fff92d26eac@arista.com/T/#u
> 
> Signed-off-by: Dmitry Safonov <dima@arista.com>

Tested-by: Hangbin Liu <liuhangbin@gmail.com>
Reviewed-by: Hangbin Liu <liuhangbin@gmail.com>

Re: [PATCH net-next 0/2] selftest/net: Some more TCP-AO selftest post-merge fixups

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net-next.git (main)
by David S. Miller <davem@davemloft.net>:

On Fri, 22 Dec 2023 01:59:05 +0000 you wrote:
> Note that there's another post-merge fix for TCP-AO selftests, but that
> doesn't conflict with these, so I don't resend that:
> 
> https://lore.kernel.org/all/20231219-b4-tcp-ao-selftests-out-of-tree-v1-1-0fff92d26eac@arista.com/T/#u
> 
> Signed-off-by: Dmitry Safonov <dima@arista.com>
> 
> [...]

Here is the summary with links:
  - [net-next,1/2] selftest/tcp-ao: Set routes in a proper VRF table id
    https://git.kernel.org/netdev/net-next/c/72cd9f8d5a99
  - [net-next,2/2] selftest/tcp-ao: Work on namespace-ified sysctl_optmem_max
    https://git.kernel.org/netdev/net-next/c/80057b2080a8

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html