On Wed, May 27, 2020 at 06:56:55PM +0000, Anton Protopopov wrote:
> This series fixes a bug in the map_lookup_and_delete_elem() function which
> should check for the FMODE_CAN_READ bit, because it returns data to user space.
> The rest of commits fix some typos and comment in selftests and extend the
> test_map_wronly test to cover the new check for the BPF_MAP_TYPE_STACK and
> BPF_MAP_TYPE_QUEUE map types.
>
> Anton Protopopov (5):
> selftests/bpf: fix a typo in test_maps
> selftests/bpf: cleanup some file descriptors in test_maps
> selftests/bpf: cleanup comments in test_maps
> bpf: fix map permissions check
> selftests/bpf: add tests for write-only stacks/queues
>
> kernel/bpf/syscall.c | 3 +-
> tools/testing/selftests/bpf/test_maps.c | 52 ++++++++++++++++++++++---
> 2 files changed, 49 insertions(+), 6 deletions(-)
Looks good to me and is also consistent with what we do for the lookup +
delete batch interface, applied thanks!
Fyi, I've taken it to bpf-next given 5.7 is right around the corner. We
can take the permissions fix to stable once in Linus' tree.