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/370649740... https://github.com/kernel-patches/bpf/actions/runs/13549227497/job/378689263... https://github.com/kernel-patches/bpf/actions/runs/13548089029/job/378658120... https://github.com/kernel-patches/bpf/actions/runs/13553536268/job/378833292... (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(-)
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; }
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);
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/378689263... https://github.com/kernel-patches/bpf/actions/runs/13548089029/job/378658120... https://github.com/kernel-patches/bpf/actions/runs/13553536268/job/378833292...
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);
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/378689263... https://github.com/kernel-patches/bpf/actions/runs/13548089029/job/378658120... https://github.com/kernel-patches/bpf/actions/runs/13553536268/job/378833292...
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);
On Thu, Feb 27, 2025 at 6:27 AM Jiayuan Chen jiayuan.chen@linux.dev wrote:
- 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.
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:
- 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/370649740... https://github.com/kernel-patches/bpf/actions/runs/13549227497/job/378689263... https://github.com/kernel-patches/bpf/actions/runs/13548089029/job/378658120... https://github.com/kernel-patches/bpf/actions/runs/13553536268/job/378833292... (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!
linux-kselftest-mirror@lists.linaro.org
-
Alexei Starovoitov -
Hou Tao -
Jiayuan Chen -
patchwork-bot+netdevbpf＠kernel.org