[PATCH bpf-next v1 0/3] Optimize bpf selftest to increase CI success rate

[PATCH bpf-next v1 0/3] Optimize bpf selftest to increase CI success rate

[Jiayuan Chen]

1. Optimized some static bound port selftests to avoid port occupation
when running test_progs -j.
2. Optimized the retry logic for test_maps.

Some Failed CI:
https://github.com/kernel-patches/bpf/actions/runs/13275542359/job/37064974076
https://github.com/kernel-patches/bpf/actions/runs/13549227497/job/37868926343
https://github.com/kernel-patches/bpf/actions/runs/13548089029/job/37865812030
https://github.com/kernel-patches/bpf/actions/runs/13553536268/job/37883329296
(Perhaps it's due to the large number of pull requests requiring CI runs?)

Jiayuan Chen (3):
  selftests/bpf: Allow auto port binding for cgroup connect
  selftests/bpf: Allow auto port binding for bpf nf
  selftests/bpf: Fixes for test_maps test

 tools/testing/selftests/bpf/prog_tests/bpf_nf.c     |  9 ++++++---
 .../testing/selftests/bpf/prog_tests/cgroup_v1v2.c  | 13 +++++++++----
 .../testing/selftests/bpf/progs/connect4_dropper.c  |  4 +++-
 tools/testing/selftests/bpf/test_maps.c             |  9 +++++----
 4 files changed, 23 insertions(+), 12 deletions(-)

-- 
2.47.1

[PATCH bpf-next v1 1/3] selftests/bpf: Allow auto port binding for cgroup connect

[Jiayuan Chen]

Allow auto port binding for cgroup connect test to avoid binding conflict.

Result:
./test_progs -a cgroup_v1v2
59      cgroup_v1v2:OK
Summary: 1/0 PASSED, 0 SKIPPED, 0 FAILED

Signed-off-by: Jiayuan Chen <jiayuan.chen@linux.dev>
---
 .../testing/selftests/bpf/prog_tests/cgroup_v1v2.c  | 13 +++++++++----
 .../testing/selftests/bpf/progs/connect4_dropper.c  |  4 +++-
 2 files changed, 12 insertions(+), 5 deletions(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/cgroup_v1v2.c b/tools/testing/selftests/bpf/prog_tests/cgroup_v1v2.c
index 64abba72ac10..37c1cc52ed98 100644
--- a/tools/testing/selftests/bpf/prog_tests/cgroup_v1v2.c
+++ b/tools/testing/selftests/bpf/prog_tests/cgroup_v1v2.c
@@ -10,12 +10,18 @@
 static int run_test(int cgroup_fd, int server_fd, bool classid)
 {
 	struct connect4_dropper *skel;
-	int fd, err = 0;
+	int fd, err = 0, port;
 
 	skel = connect4_dropper__open_and_load();
 	if (!ASSERT_OK_PTR(skel, "skel_open"))
 		return -1;
 
+	port = get_socket_local_port(server_fd);
+	if (!ASSERT_GE(port, 0, "get_socket_local_port"))
+		return -1;
+
+	skel->bss->port = ntohs(port);
+
 	skel->links.connect_v4_dropper =
 		bpf_program__attach_cgroup(skel->progs.connect_v4_dropper,
 					   cgroup_fd);
@@ -48,10 +54,9 @@ void test_cgroup_v1v2(void)
 {
 	struct network_helper_opts opts = {};
 	int server_fd, client_fd, cgroup_fd;
-	static const int port = 60120;
 
 	/* Step 1: Check base connectivity works without any BPF. */
-	server_fd = start_server(AF_INET, SOCK_STREAM, NULL, port, 0);
+	server_fd = start_server(AF_INET, SOCK_STREAM, NULL, 0, 0);
 	if (!ASSERT_GE(server_fd, 0, "server_fd"))
 		return;
 	client_fd = connect_to_fd_opts(server_fd, &opts);
@@ -66,7 +71,7 @@ void test_cgroup_v1v2(void)
 	cgroup_fd = test__join_cgroup("/connect_dropper");
 	if (!ASSERT_GE(cgroup_fd, 0, "cgroup_fd"))
 		return;
-	server_fd = start_server(AF_INET, SOCK_STREAM, NULL, port, 0);
+	server_fd = start_server(AF_INET, SOCK_STREAM, NULL, 0, 0);
 	if (!ASSERT_GE(server_fd, 0, "server_fd")) {
 		close(cgroup_fd);
 		return;
diff --git a/tools/testing/selftests/bpf/progs/connect4_dropper.c b/tools/testing/selftests/bpf/progs/connect4_dropper.c
index d3f4c5e4fb69..a3819a5d09c8 100644
--- a/tools/testing/selftests/bpf/progs/connect4_dropper.c
+++ b/tools/testing/selftests/bpf/progs/connect4_dropper.c
@@ -13,12 +13,14 @@
 #define VERDICT_REJECT	0
 #define VERDICT_PROCEED	1
 
+int port;
+
 SEC("cgroup/connect4")
 int connect_v4_dropper(struct bpf_sock_addr *ctx)
 {
 	if (ctx->type != SOCK_STREAM)
 		return VERDICT_PROCEED;
-	if (ctx->user_port == bpf_htons(60120))
+	if (ctx->user_port == bpf_htons(port))
 		return VERDICT_REJECT;
 	return VERDICT_PROCEED;
 }
-- 
2.47.1

[PATCH bpf-next v1 2/3] selftests/bpf: Allow auto port binding for bpf nf

[Jiayuan Chen]

Allow auto port binding for bpf nf test to avoid binding conflict.

./test_progs -a bpf_nf
24/1    bpf_nf/xdp-ct:OK
24/2    bpf_nf/tc-bpf-ct:OK
24/3    bpf_nf/alloc_release:OK
24/4    bpf_nf/insert_insert:OK
24/5    bpf_nf/lookup_insert:OK
24/6    bpf_nf/set_timeout_after_insert:OK
24/7    bpf_nf/set_status_after_insert:OK
24/8    bpf_nf/change_timeout_after_alloc:OK
24/9    bpf_nf/change_status_after_alloc:OK
24/10   bpf_nf/write_not_allowlisted_field:OK
24      bpf_nf:OK
Summary: 1/10 PASSED, 0 SKIPPED, 0 FAILED

Signed-off-by: Jiayuan Chen <jiayuan.chen@linux.dev>
---
 tools/testing/selftests/bpf/prog_tests/bpf_nf.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/bpf_nf.c b/tools/testing/selftests/bpf/prog_tests/bpf_nf.c
index a4a1f93878d4..dbd13f8e42a7 100644
--- a/tools/testing/selftests/bpf/prog_tests/bpf_nf.c
+++ b/tools/testing/selftests/bpf/prog_tests/bpf_nf.c
@@ -72,11 +72,14 @@ static void test_bpf_nf_ct(int mode)
 	if (!ASSERT_OK(system(cmd), cmd))
 		goto end;
 
-	srv_port = (mode == TEST_XDP) ? 5005 : 5006;
-	srv_fd = start_server(AF_INET, SOCK_STREAM, "127.0.0.1", srv_port, TIMEOUT_MS);
+	srv_fd = start_server(AF_INET, SOCK_STREAM, "127.0.0.1", 0, TIMEOUT_MS);
 	if (!ASSERT_GE(srv_fd, 0, "start_server"))
 		goto end;
 
+	srv_port = get_socket_local_port(srv_fd);
+	if (!ASSERT_GE(srv_port, 0, "get_sock_local_port"))
+		goto end;
+
 	client_fd = connect_to_server(srv_fd);
 	if (!ASSERT_GE(client_fd, 0, "connect_to_server"))
 		goto end;
@@ -91,7 +94,7 @@ static void test_bpf_nf_ct(int mode)
 	skel->bss->saddr = peer_addr.sin_addr.s_addr;
 	skel->bss->sport = peer_addr.sin_port;
 	skel->bss->daddr = peer_addr.sin_addr.s_addr;
-	skel->bss->dport = htons(srv_port);
+	skel->bss->dport = srv_port;
 
 	if (mode == TEST_XDP)
 		prog_fd = bpf_program__fd(skel->progs.nf_xdp_ct_test);
-- 
2.47.1

[PATCH bpf-next v1 3/3] selftests/bpf: Fixes for test_maps test

[Jiayuan Chen]

BPF CI has failed 3 times in the last 24 hours. Add retry for ENOMEM.
It's similar to the optimization plan:
commit 2f553b032cad ("selftsets/bpf: Retry map update for non-preallocated per-cpu map")

Failed CI:
https://github.com/kernel-patches/bpf/actions/runs/13549227497/job/37868926343
https://github.com/kernel-patches/bpf/actions/runs/13548089029/job/37865812030
https://github.com/kernel-patches/bpf/actions/runs/13553536268/job/37883329296

selftests/bpf: Fixes for test_maps test
Fork 100 tasks to 'test_update_delete'
Fork 100 tasks to 'test_update_delete'
Fork 100 tasks to 'test_update_delete'
Fork 100 tasks to 'test_update_delete'
......
test_task_storage_map_stress_lookup:PASS
test_maps: OK, 0 SKIPPED

Signed-off-by: Jiayuan Chen <jiayuan.chen@linux.dev>
---
 tools/testing/selftests/bpf/test_maps.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/tools/testing/selftests/bpf/test_maps.c b/tools/testing/selftests/bpf/test_maps.c
index 8b40e9496af1..986ce32b113a 100644
--- a/tools/testing/selftests/bpf/test_maps.c
+++ b/tools/testing/selftests/bpf/test_maps.c
@@ -1396,9 +1396,10 @@ static void test_map_stress(void)
 #define MAX_DELAY_US 50000
 #define MIN_DELAY_RANGE_US 5000
 
-static bool retry_for_again_or_busy(int err)
+static bool can_retry(int err)
 {
-	return (err == EAGAIN || err == EBUSY);
+	return (err == EAGAIN || err == EBUSY ||
+		(err == ENOMEM && map_opts.map_flags == BPF_F_NO_PREALLOC));
 }
 
 int map_update_retriable(int map_fd, const void *key, const void *value, int flags, int attempts,
@@ -1451,12 +1452,12 @@ static void test_update_delete(unsigned int fn, void *data)
 
 		if (do_update) {
 			err = map_update_retriable(fd, &key, &value, BPF_NOEXIST, MAP_RETRIES,
-						   retry_for_again_or_busy);
+						   can_retry);
 			if (err)
 				printf("error %d %d\n", err, errno);
 			assert(err == 0);
 			err = map_update_retriable(fd, &key, &value, BPF_EXIST, MAP_RETRIES,
-						   retry_for_again_or_busy);
+						   can_retry);
 			if (err)
 				printf("error %d %d\n", err, errno);
 			assert(err == 0);
-- 
2.47.1

Re: [PATCH bpf-next v1 0/3] Optimize bpf selftest to increase CI success rate

[Alexei Starovoitov]

On Thu, Feb 27, 2025 at 6:27 AM Jiayuan Chen <jiayuan.chen@linux.dev> wrote:
>
> 1. Optimized some static bound port selftests to avoid port occupation
> when running test_progs -j.
> 2. Optimized the retry logic for test_maps.

Looks great. Applied.
Thank you for fixing them.

Re: [PATCH bpf-next v1 0/3] Optimize bpf selftest to increase CI success rate

[patchwork-bot+netdevbpf]

Hello:

This series was applied to bpf/bpf-next.git (master)
by Alexei Starovoitov <ast@kernel.org>:

On Thu, 27 Feb 2025 22:26:43 +0800 you wrote:
> 1. Optimized some static bound port selftests to avoid port occupation
> when running test_progs -j.
> 2. Optimized the retry logic for test_maps.
> 
> Some Failed CI:
> https://github.com/kernel-patches/bpf/actions/runs/13275542359/job/37064974076
> https://github.com/kernel-patches/bpf/actions/runs/13549227497/job/37868926343
> https://github.com/kernel-patches/bpf/actions/runs/13548089029/job/37865812030
> https://github.com/kernel-patches/bpf/actions/runs/13553536268/job/37883329296
> (Perhaps it's due to the large number of pull requests requiring CI runs?)
> 
> [...]

Here is the summary with links:
  - [bpf-next,v1,1/3] selftests/bpf: Allow auto port binding for cgroup connect
    https://git.kernel.org/bpf/bpf-next/c/27e3162a0364
  - [bpf-next,v1,2/3] selftests/bpf: Allow auto port binding for bpf nf
    https://git.kernel.org/bpf/bpf-next/c/dbe7d46ed109
  - [bpf-next,v1,3/3] selftests/bpf: Fixes for test_maps test
    https://git.kernel.org/bpf/bpf-next/c/09de329523c8

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

Re: [PATCH bpf-next v1 3/3] selftests/bpf: Fixes for test_maps test

[Hou Tao]

Hi,

On 2/27/2025 10:26 PM, Jiayuan Chen wrote:
> BPF CI has failed 3 times in the last 24 hours. Add retry for ENOMEM.
> It's similar to the optimization plan:
> commit 2f553b032cad ("selftsets/bpf: Retry map update for non-preallocated per-cpu map")
>
> Failed CI:
> https://github.com/kernel-patches/bpf/actions/runs/13549227497/job/37868926343
> https://github.com/kernel-patches/bpf/actions/runs/13548089029/job/37865812030
> https://github.com/kernel-patches/bpf/actions/runs/13553536268/job/37883329296

Er, it is suspicious that the allocation of htab_elem failed, because
the size of key and value is small in these cases. It usually means that
there is something wrong with bpf mem allocator or the slub, therefore,
I think it is a bad idea to retry for ENOMEM in this case. Will try to
reproduce the problem first when get some time.
>
> selftests/bpf: Fixes for test_maps test
> Fork 100 tasks to 'test_update_delete'
> Fork 100 tasks to 'test_update_delete'
> Fork 100 tasks to 'test_update_delete'
> Fork 100 tasks to 'test_update_delete'
> ......
> test_task_storage_map_stress_lookup:PASS
> test_maps: OK, 0 SKIPPED
>
> Signed-off-by: Jiayuan Chen <jiayuan.chen@linux.dev>
> ---
>  tools/testing/selftests/bpf/test_maps.c | 9 +++++----
>  1 file changed, 5 insertions(+), 4 deletions(-)
>
> diff --git a/tools/testing/selftests/bpf/test_maps.c b/tools/testing/selftests/bpf/test_maps.c
> index 8b40e9496af1..986ce32b113a 100644
> --- a/tools/testing/selftests/bpf/test_maps.c
> +++ b/tools/testing/selftests/bpf/test_maps.c
> @@ -1396,9 +1396,10 @@ static void test_map_stress(void)
>  #define MAX_DELAY_US 50000
>  #define MIN_DELAY_RANGE_US 5000
>  
> -static bool retry_for_again_or_busy(int err)
> +static bool can_retry(int err)
>  {
> -	return (err == EAGAIN || err == EBUSY);
> +	return (err == EAGAIN || err == EBUSY ||
> +		(err == ENOMEM && map_opts.map_flags == BPF_F_NO_PREALLOC));
>  }
>  
>  int map_update_retriable(int map_fd, const void *key, const void *value, int flags, int attempts,
> @@ -1451,12 +1452,12 @@ static void test_update_delete(unsigned int fn, void *data)
>  
>  		if (do_update) {
>  			err = map_update_retriable(fd, &key, &value, BPF_NOEXIST, MAP_RETRIES,
> -						   retry_for_again_or_busy);
> +						   can_retry);
>  			if (err)
>  				printf("error %d %d\n", err, errno);
>  			assert(err == 0);
>  			err = map_update_retriable(fd, &key, &value, BPF_EXIST, MAP_RETRIES,
> -						   retry_for_again_or_busy);
> +						   can_retry);
>  			if (err)
>  				printf("error %d %d\n", err, errno);
>  			assert(err == 0);