Re: [PATCH bpf-next 4/4] selftests/bpf: convert test_skb_cgroup_id_user to test_progs

On 8/1/24 10:49, Alan Maguire wrote:

On 31/07/2024 11:38, Alexis Lothoré (eBPF Foundation) wrote:

[...]

...

+static int wait_local_ip(void) +{

• char *ping_cmd = ping_command(AF_INET6);
• int i, err;
• for (i = 0; i < WAIT_AUTO_IP_MAX_ATTEMPT; i++) {

• err = SYS_NOFAIL("%s -c 1 -W 1 %s%%%s", ping_cmd, DST_ADDR,
  

• 		 VETH_1);
  

• if (!err)
  

• 	break;
  

• }

thinking about the risks of CI flakiness, would a small sleep between checks be worth doing here?

I assumed that adding -W 1 (ping timeout duration) to the command would be enough to make sure that there is a proper wait between each attempt (so currently, waiting at most 10s for network configuration between the 2 veths). Don't you think it is enough to prevent issues in CI ?

...

[...]

...

• expected_ids[0] = get_cgroup_id("/.."); /* root cgroup */
• expected_ids[1] = get_cgroup_id("");
• expected_ids[2] = get_cgroup_id(CGROUP_PATH);
• expected_ids[3] = 0; /* non-existent cgroup */
• for (level = 0; level < NUM_CGROUP_LEVELS; level++) {

• err = bpf_map__lookup_elem(t->skel->maps.cgroup_ids, &level,
  

• 			   sizeof(level), &actual_ids[level],
  

• 			   sizeof(__u64), 0);
  

could probably simplify this + the BPF prog using a global array of actual_ids[], then compare it to the expected values using skel->bss->actual_ids

ACK, I'll update this.

Thanks,

Alexis