1 .. SPDX-License-Identifier: GPL-2.0
3 ============================
4 BPF_PROG_TYPE_FLOW_DISSECTOR
5 ============================
10 Flow dissector is a routine that parses metadata out of the packets. It's
11 used in the various places in the networking subsystem (RFS, flow hash, etc).
13 BPF flow dissector is an attempt to reimplement C-based flow dissector logic
14 in BPF to gain all the benefits of BPF verifier (namely, limits on the
15 number of instructions and tail calls).
20 BPF flow dissector programs operate on an ``__sk_buff``. However, only the
21 limited set of fields is allowed: ``data``, ``data_end`` and ``flow_keys``.
22 ``flow_keys`` is ``struct bpf_flow_keys`` and contains flow dissector input
26 * ``nhoff`` - initial offset of the networking header
27 * ``thoff`` - initial offset of the transport header, initialized to nhoff
28 * ``n_proto`` - L3 protocol type, parsed out of L2 header
29 * ``flags`` - optional flags
31 Flow dissector BPF program should fill out the rest of the ``struct
32 bpf_flow_keys`` fields. Input arguments ``nhoff/thoff/n_proto`` should be
33 also adjusted accordingly.
35 The return code of the BPF program is either BPF_OK to indicate successful
36 dissection, or BPF_DROP to indicate parsing error.
41 In the VLAN-less case, this is what the initial state of the BPF flow
42 dissector looks like::
44 +------+------+------------+-----------+
45 | DMAC | SMAC | ETHER_TYPE | L3_HEADER |
46 +------+------+------------+-----------+
49 +-- flow dissector starts here
54 skb->data + flow_keys->nhoff point to the first byte of L3_HEADER
55 flow_keys->thoff = nhoff
56 flow_keys->n_proto = ETHER_TYPE
58 In case of VLAN, flow dissector can be called with the two different states.
62 +------+------+------+-----+-----------+-----------+
63 | DMAC | SMAC | TPID | TCI |ETHER_TYPE | L3_HEADER |
64 +------+------+------+-----+-----------+-----------+
67 +-- flow dissector starts here
71 skb->data + flow_keys->nhoff point the to first byte of TCI
72 flow_keys->thoff = nhoff
73 flow_keys->n_proto = TPID
75 Please note that TPID can be 802.1AD and, hence, BPF program would
76 have to parse VLAN information twice for double tagged packets.
81 +------+------+------+-----+-----------+-----------+
82 | DMAC | SMAC | TPID | TCI |ETHER_TYPE | L3_HEADER |
83 +------+------+------+-----+-----------+-----------+
86 +-- flow dissector starts here
90 skb->data + flow_keys->nhoff point the to first byte of L3_HEADER
91 flow_keys->thoff = nhoff
92 flow_keys->n_proto = ETHER_TYPE
94 In this case VLAN information has been processed before the flow dissector
95 and BPF flow dissector is not required to handle it.
98 The takeaway here is as follows: BPF flow dissector program can be called with
99 the optional VLAN header and should gracefully handle both cases: when single
100 or double VLAN is present and when it is not present. The same program
101 can be called for both cases and would have to be written carefully to
108 ``flow_keys->flags`` might contain optional input flags that work as follows:
110 * ``BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG`` - tells BPF flow dissector to
111 continue parsing first fragment; the default expected behavior is that
112 flow dissector returns as soon as it finds out that the packet is fragmented;
113 used by ``eth_get_headlen`` to estimate length of all headers for GRO.
114 * ``BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL`` - tells BPF flow dissector to
115 stop parsing as soon as it reaches IPv6 flow label; used by
116 ``___skb_get_hash`` and ``__skb_get_hash_symmetric`` to get flow hash.
117 * ``BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP`` - tells BPF flow dissector to stop
118 parsing as soon as it reaches encapsulated headers; used by routing
122 Reference Implementation
123 ========================
125 See ``tools/testing/selftests/bpf/progs/bpf_flow.c`` for the reference
126 implementation and ``tools/testing/selftests/bpf/flow_dissector_load.[hc]``
127 for the loader. bpftool can be used to load BPF flow dissector program as well.
129 The reference implementation is organized as follows:
130 * ``jmp_table`` map that contains sub-programs for each supported L3 protocol
131 * ``_dissect`` routine - entry point; it does input ``n_proto`` parsing and
132 does ``bpf_tail_call`` to the appropriate L3 handler
134 Since BPF at this point doesn't support looping (or any jumping back),
135 jmp_table is used instead to handle multiple levels of encapsulation (and
141 BPF flow dissector doesn't support exporting all the metadata that in-kernel
142 C-based implementation can export. Notable example is single VLAN (802.1Q)
143 and double VLAN (802.1AD) tags. Please refer to the ``struct bpf_flow_keys``
144 for a set of information that's currently can be exported from the BPF context.