1.. SPDX-License-Identifier: (GPL-2.0 OR BSD-2-Clause)
2
3=====================
4BPF sk_lookup program
5=====================
6
7BPF sk_lookup program type (``BPF_PROG_TYPE_SK_LOOKUP``) introduces programmability
8into the socket lookup performed by the transport layer when a packet is to be
9delivered locally.
10
11When invoked BPF sk_lookup program can select a socket that will receive the
12incoming packet by calling the ``bpf_sk_assign()`` BPF helper function.
13
14Hooks for a common attach point (``BPF_SK_LOOKUP``) exist for both TCP and UDP.
15
16Motivation
17==========
18
19BPF sk_lookup program type was introduced to address setup scenarios where
20binding sockets to an address with ``bind()`` socket call is impractical, such
21as:
22
231. receiving connections on a range of IP addresses, e.g. 192.0.2.0/24, when
24   binding to a wildcard address ``INADRR_ANY`` is not possible due to a port
25   conflict,
262. receiving connections on all or a wide range of ports, i.e. an L7 proxy use
27   case.
28
29Such setups would require creating and ``bind()``'ing one socket to each of the
30IP address/port in the range, leading to resource consumption and potential
31latency spikes during socket lookup.
32
33Attachment
34==========
35
36BPF sk_lookup program can be attached to a network namespace with
37``bpf(BPF_LINK_CREATE, ...)`` syscall using the ``BPF_SK_LOOKUP`` attach type and a
38netns FD as attachment ``target_fd``.
39
40Multiple programs can be attached to one network namespace. Programs will be
41invoked in the same order as they were attached.
42
43Hooks
44=====
45
46The attached BPF sk_lookup programs run whenever the transport layer needs to
47find a listening (TCP) or an unconnected (UDP) socket for an incoming packet.
48
49Incoming traffic to established (TCP) and connected (UDP) sockets is delivered
50as usual without triggering the BPF sk_lookup hook.
51
52The attached BPF programs must return with either ``SK_PASS`` or ``SK_DROP``
53verdict code. As for other BPF program types that are network filters,
54``SK_PASS`` signifies that the socket lookup should continue on to regular
55hashtable-based lookup, while ``SK_DROP`` causes the transport layer to drop the
56packet.
57
58A BPF sk_lookup program can also select a socket to receive the packet by
59calling ``bpf_sk_assign()`` BPF helper. Typically, the program looks up a socket
60in a map holding sockets, such as ``SOCKMAP`` or ``SOCKHASH``, and passes a
61``struct bpf_sock *`` to ``bpf_sk_assign()`` helper to record the
62selection. Selecting a socket only takes effect if the program has terminated
63with ``SK_PASS`` code.
64
65When multiple programs are attached, the end result is determined from return
66codes of all the programs according to the following rules:
67
681. If any program returned ``SK_PASS`` and selected a valid socket, the socket
69   is used as the result of the socket lookup.
702. If more than one program returned ``SK_PASS`` and selected a socket, the last
71   selection takes effect.
723. If any program returned ``SK_DROP``, and no program returned ``SK_PASS`` and
73   selected a socket, socket lookup fails.
744. If all programs returned ``SK_PASS`` and none of them selected a socket,
75   socket lookup continues on.
76
77API
78===
79
80In its context, an instance of ``struct bpf_sk_lookup``, BPF sk_lookup program
81receives information about the packet that triggered the socket lookup. Namely:
82
83* IP version (``AF_INET`` or ``AF_INET6``),
84* L4 protocol identifier (``IPPROTO_TCP`` or ``IPPROTO_UDP``),
85* source and destination IP address,
86* source and destination L4 port,
87* the socket that has been selected with ``bpf_sk_assign()``.
88
89Refer to ``struct bpf_sk_lookup`` declaration in ``linux/bpf.h`` user API
90header, and `bpf-helpers(7)
91<https://man7.org/linux/man-pages/man7/bpf-helpers.7.html>`_ man-page section
92for ``bpf_sk_assign()`` for details.
93
94Example
95=======
96
97See ``tools/testing/selftests/bpf/prog_tests/sk_lookup.c`` for the reference
98implementation.
99