[PATCH bpf 0/2] bpf: Fix RHASH batch cursor loss

[Tianyi Chen <hi@tychen.cc>]

RHASH batch lookups can report end-of-map when a key is deleted during
the walk. Return EAGAIN for that case and add a regression test.

Tested in an x86-64 VM: the regression fails without the fix and all
five subtests pass with it. The bpftool batch and recursive tests also
pass, separately and together.

Tianyi Chen (2):
  bpf: Report lost RHASH batch traversal keys as EAGAIN
  selftests/bpf: Cover RHASH batch cursor loss

 kernel/bpf/hashtab.c                          |  13 +-
 .../selftests/bpf/prog_tests/rhash_batch.c    | 181 ++++++++++++++++++
 2 files changed, 188 insertions(+), 6 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/rhash_batch.c


base-commit: 646114a0d2f39e52f2204196ac12a6fe5a5a3fd3
-- 
2.56.0