On Wed, Jul 1, 2020 at 1:14 PM Stanislav Fomichev <sdf@xxxxxxxxxx> wrote:
>
> Simple test that enforces a single SOCK_DGRAM socker per cgroup.
>
> Signed-off-by: Stanislav Fomichev <sdf@xxxxxxxxxx>
> ---
> .../selftests/bpf/prog_tests/udp_limit.c | 72 +++++++++++++++++++
> tools/testing/selftests/bpf/progs/udp_limit.c | 42 +++++++++++
> 2 files changed, 114 insertions(+)
> create mode 100644 tools/testing/selftests/bpf/prog_tests/udp_limit.c
> create mode 100644 tools/testing/selftests/bpf/progs/udp_limit.c
>
[...]
> + fd1 = socket(AF_INET, SOCK_DGRAM, 0);
> + if (CHECK(fd1 < 0, "fd1", "errno %d", errno))
> + goto close_skeleton;
> +
> + fd2 = socket(AF_INET, SOCK_DGRAM, 0);
> + if (CHECK(fd2 >= 0, "fd2", "errno %d", errno))
close(fd2);
> + goto close_fd1;
> +
[...]
> +close_fd1:
> + close(fd1);
> +close_skeleton:
> + udp_limit__destroy(skel);
> +close_cgroup_fd:
> + close(cgroup_fd);
nit: choosing between close_fd1 and close_skeleton (which also
alternates!) is hard to keep track of. When clean up gets one step
beyond trivial, I usually just initialize variables properly and do
clean up for all exit scenario in one block:
if (fd1 >= 0)
close(fd1);
udp_limit__destroy(skel);
close(cgroup_fd);
It also makes later extensions simpler.
> +}
> diff --git a/tools/testing/selftests/bpf/progs/udp_limit.c b/tools/testing/selftests/bpf/progs/udp_limit.c
> new file mode 100644
> index 000000000000..af1154bfb946
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/progs/udp_limit.c
> @@ -0,0 +1,42 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +
> +#include <sys/socket.h>
> +#include <linux/bpf.h>
> +#include <bpf/bpf_helpers.h>
> +
> +int invocations, in_use;
uninitialized globals can leads to libbpf refusing to load them (due
to COM section), Daniel just recently had this problem. So better to
always zero-initialize them.
[...]
