netdev - Re: [PATCH bpf-next 0/5] add bpf cgroup hooks that trigger on socket close

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 18 Jan 2019 02:36:58 +0000
From:   Andrey Ignatov <rdna@...com>
To:     Stanislav Fomichev <sdf@...gle.com>
CC:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "davem@...emloft.net" <davem@...emloft.net>,
        "ast@...nel.org" <ast@...nel.org>,
        "daniel@...earbox.net" <daniel@...earbox.net>
Subject: Re: [PATCH bpf-next 0/5] add bpf cgroup hooks that trigger on socket
 close

Stanislav Fomichev <sdf@...gle.com> [Thu, 2019-01-17 16:41 -0800]:
> Currently, we have BPF_CGROUP_INET_SOCK_CREATE hook that triggers on
> socket creation and there is no way to know when the socket is being
> closed. Add new set of hooks BPF_CGROUP_INET{4,6}_SOCK_RELEASE
> that trigger when the socket is closed.
> 
> Initial intended usecase is to cleanup statistics after POST{4,6}_BIND.
> Hooks have read-only access to all fields of struct bpf_sock.

Do you need it for both TCP and UDP?

I was thinking about this hook earlier but since in my case only TCP was
needed I ended up using TCP-BPF. E.g. be BPF_SOCK_OPS_TCP_LISTEN_CB or
BPF_SOCK_OPS_TCP_CONNECT_CB can be used instead of POST{4,6}_BIND to
enable something, and then BPF_SOCK_OPS_STATE_CB can be used instead of
SOCK_RELEASE to disable that something when socket transisions to
BPF_TCP_CLOSE (e.g. BPF_TCP_LISTEN -> BPF_TCP_CLOSE).

That turned out to be much cleaner than POST{4,6}_BIND and also works
fine when socket is disconnected with AF_UNSPEC and then connected again
(what Eric mentioned).


> First patch adds hooks, the rest of the patches add uapi and tests to make
> sure these hooks work.
> 
> Stanislav Fomichev (5):
>   bpf: add BPF_CGROUP_INET{4,6}_SOCK_RELEASE hooks
>   tools: bpf: support BPF_CGROUP_INET{4,6}_SOCK_RELEASE in
>     libbpf/bpftool
>   selftests/bpf: add BPF_CGROUP_INET{4,6}_SOCK_RELEASE to
>     test_section_names.c
>   selftests/bpf: add BPF_CGROUP_INET{4,6}_SOCK_RELEASE to test_sock.c
>   selftests/bpf: add BPF_CGROUP_INET{4,6}_SOCK_RELEASE to
>     test_sock_addr.c
> 
>  include/linux/bpf-cgroup.h                    |   6 +
>  include/net/inet_common.h                     |   1 +
>  include/uapi/linux/bpf.h                      |   2 +
>  kernel/bpf/syscall.c                          |   8 ++
>  net/core/filter.c                             |   7 +
>  net/ipv4/af_inet.c                            |  13 +-
>  net/ipv6/af_inet6.c                           |   5 +-
>  tools/bpf/bpftool/cgroup.c                    |   2 +
>  tools/include/uapi/linux/bpf.h                |   2 +
>  tools/lib/bpf/libbpf.c                        |   4 +
>  .../selftests/bpf/test_section_names.c        |  10 ++
>  tools/testing/selftests/bpf/test_sock.c       | 119 ++++++++++++++++
>  tools/testing/selftests/bpf/test_sock_addr.c  | 131 +++++++++++++++++-
>  13 files changed, 307 insertions(+), 3 deletions(-)
> 
> -- 
> 2.20.1.321.g9e740568ce-goog
> 

-- 
Andrey Ignatov