test_maps from ubuntu_bpf failed with "FAIL:failed to create hashmap" on B-hwe i386

[Po-Hsu Lin]

Test failed with:
  Fork 100 tasks to 'test_hashmap_walk'
  helper_fill_hashmap(270):FAIL:failed to create hashmap err: Cannot allocate memory, flags: 0x0
  ...
  Fork 100 tasks to 'test_hashmap_walk'
  test_maps: test_maps.c:1286: __run_parallel: Assertion `status == 0' failed.
  Fork 1024 tasks to 'test_update_delete'

Complete test log could be found here:
https://pastebin.ubuntu.com/p/nQBvZJH5Qb/

ProblemType: Bug
DistroRelease: Ubuntu 18.04
Package: linux-image-5.0.0-21-lowlatency 5.0.0-21.22~18.04.1
ProcVersionSignature: User Name 5.0.0-21.22~18.04.1-lowlatency 5.0.15
Uname: Linux 5.0.0-21-lowlatency i686
ApportVersion: 2.20.9-0ubuntu7.7
Architecture: i386
Date: Mon Jul 22 11:30:53 2019
SourcePackage: linux-hwe
UpgradeStatus: No upgrade log present (probably fresh install)

[Po-Hsu Lin (cypressyew)]

Spotted on Bionic 5.4.0-42.46~18.04.1 node fozzie:

  Fork 100 tasks to 'test_hashmap_walk'
  helper_fill_hashmap(261):FAIL:failed to create hashmap err: Cannot allocate memory, flags: 0x0

[Po-Hsu Lin (cypressyew)]

BTW
there is an error before this failure mentioned in comment #2:
  Fork 100 tasks to 'test_hashmap'
  test_maps: test_maps.c:1300: __run_parallel: Assertion `status == 0' failed

Maybe it's this failure that cause the "failed to create hashmap"

Please find attachment for the complete test log

[Luke Nowakowski-Krijger (lukenow)]

Very similar issue found on bionic/gcp-4.15 on 4.15.0-1109.123 only for n1-highcpu-4 instance.

18:45:34 DEBUG| Running './test_maps'
18:45:36 DEBUG| [stdout] Failed to create hashmap 'Cannot allocate memory'!
18:45:36 DEBUG| [stdout] Failed to create hashmap 'Cannot allocate memory'!
18:45:36 DEBUG| [stdout] Failed to create hashmap 'Cannot allocate memory'!
18:45:36 DEBUG| [stdout] Failed to create hashmap 'Cannot allocate memory'!
18:46:06 ERROR| [stderr] test_maps: test_maps.c:967: run_parallel: Assertion `status == 0' failed.

With the rest of the threads reporting the same error.

This issue also seems to be found in previous cycles on other instances as well.

[Juerg Haefliger (juergh)]

Also flaky with b/oracle and VM.Standard.E3.Flex-4c.8m instance types. That instance has 'only' 8GB so maybe at the hairy edge?