Re: [PATCH bpf-next v3 3/7] bpf: minimize number of allocated lsm slots per program

[Jakub Sitnicki <jakub@cloudflare.com>]

On Mon, Apr 11, 2022 at 11:44 AM -07, Stanislav Fomichev wrote:
> On Sat, Apr 9, 2022 at 11:10 AM Jakub Sitnicki <jakub@cloudflare.com> wrote:

[...]

>> [^1] It looks like we can easily switch from cgroup->bpf.progs[] from
>>      list_head to hlist_head and save some bytes!
>>
>>      We only access the list tail in __cgroup_bpf_attach(). We can
>>      either iterate over the list and eat the cost there or push the new
>>      prog onto the front.
>>
>>      I think we treat cgroup->bpf.progs[] everywhere like an unordered
>>      set. Except for __cgroup_bpf_query, where the user might notice the
>>      order change in the BPF_PROG_QUERY dump.
>
>
> [...]
>
>> [^2] Unrelated, but we would like to propose a
>>      CGROUP_INET[46]_POST_CONNECT hook in the near future to make it
>>      easier to bind UDP sockets to 4-tuple without creating conflicts:
>>
>>      https://github.com/cloudflare/cloudflare-blog/tree/master/2022-02-connectx/ebpf_connect4
>
> Do you think those new lsm hooks can be used instead? If not, what's missing?

Same as for CGROUP_INET hooks, there is no post-connect() LSM hook.

Why are we looking for a post-connect hook?

Having a pre- and a post- connect hook, would allow us to turn the whole
connect() syscall into a critical section with synchronization done in
BPF - lock on pre-connect, unlock on post-connect.

Why do we want to serialize connect() calls?

To check for 4-tuple conflict with an existing unicast UDP socket, in
which case we want fail connect() if there is a conflict.

That said, ideally we would rather have a mechanism like
IP_BIND_ADDRESS_NO_PORT, but for UDP, and one that allows selecting both
an local IP and port.

We're hoping to put together an RFC sometime this quarter.