xref: /dpdk/doc/guides/prog_guide/bpf_lib.rst (revision 41dd9a6bc2d9c6e20e139ad713cc9d172572dd43)
1..  SPDX-License-Identifier: BSD-3-Clause
2    Copyright(c) 2018 Intel Corporation.
3
4Berkeley Packet Filter (BPF) Library
5====================================
6
7The DPDK provides an BPF library that gives the ability
8to load and execute Enhanced Berkeley Packet Filter (eBPF) bytecode within
9user-space dpdk application.
10
11It supports basic set of features from eBPF spec.
12Please refer to the
13`eBPF spec <https://www.kernel.org/doc/Documentation/networking/filter.txt>`_
14for more information.
15Also it introduces basic framework to load/unload BPF-based filters
16on eth devices (right now only via SW RX/TX callbacks).
17
18The library API provides the following basic operations:
19
20*  Create a new BPF execution context and load user provided eBPF code into it.
21
22*   Destroy an BPF execution context and its runtime structures and free the associated memory.
23
24*   Execute eBPF bytecode associated with provided input parameter.
25
26*   Provide information about natively compiled code for given BPF context.
27
28*   Load BPF program from the ELF file and install callback to execute it on given ethdev port/queue.
29
30Packet data load instructions
31-----------------------------
32
33DPDK supports two non-generic instructions: ``(BPF_ABS | size | BPF_LD)``
34and ``(BPF_IND | size | BPF_LD)`` which are used to access packet data.
35These instructions can only be used when execution context is a pointer to
36``struct rte_mbuf`` and have seven implicit operands.
37Register ``R6`` is an implicit input that must contain pointer to ``rte_mbuf``.
38Register ``R0`` is an implicit output which contains the data fetched from the
39packet. Registers ``R1-R5`` are scratch registers
40and must not be used to store the data across these instructions.
41These instructions have implicit program exit condition as well. When
42eBPF program is trying to access the data beyond the packet boundary,
43the interpreter will abort the execution of the program. JIT compilers
44therefore must preserve this property. ``src_reg`` and ``imm32`` fields are
45explicit inputs to these instructions.
46For example, ``(BPF_IND | BPF_W | BPF_LD)`` means:
47
48.. code-block:: c
49
50    uint32_t tmp;
51    R0 = rte_pktmbuf_read((const struct rte_mbuf *)R6,  src_reg + imm32, sizeof(tmp), &tmp);
52    if (R0 == NULL)
53        return FAILED;
54    R0 = ntohl(*(uint32_t *)R0);
55
56and ``R1-R5`` were scratched.
57
58
59Not currently supported eBPF features
60-------------------------------------
61
62 - JIT support only available for X86_64 and arm64 platforms
63 - cBPF
64 - tail-pointer call
65 - eBPF MAP
66 - external function calls for 32-bit platforms
67