| 1 | .. SPDX-License-Identifier: GPL-2.0-only |
| 2 | .. Copyright (C) 2022 Red Hat, Inc. |
| 3 | |
| 4 | ========================= |
| 5 | BPF_MAP_TYPE_BLOOM_FILTER |
| 6 | ========================= |
| 7 | |
| 8 | .. note:: |
| 9 | - ``BPF_MAP_TYPE_BLOOM_FILTER`` was introduced in kernel version 5.16 |
| 10 | |
| 11 | ``BPF_MAP_TYPE_BLOOM_FILTER`` provides a BPF bloom filter map. Bloom |
| 12 | filters are a space-efficient probabilistic data structure used to |
| 13 | quickly test whether an element exists in a set. In a bloom filter, |
| 14 | false positives are possible whereas false negatives are not. |
| 15 | |
| 16 | The bloom filter map does not have keys, only values. When the bloom |
| 17 | filter map is created, it must be created with a ``key_size`` of 0. The |
| 18 | bloom filter map supports two operations: |
| 19 | |
| 20 | - push: adding an element to the map |
| 21 | - peek: determining whether an element is present in the map |
| 22 | |
| 23 | BPF programs must use ``bpf_map_push_elem`` to add an element to the |
| 24 | bloom filter map and ``bpf_map_peek_elem`` to query the map. These |
| 25 | operations are exposed to userspace applications using the existing |
| 26 | ``bpf`` syscall in the following way: |
| 27 | |
| 28 | - ``BPF_MAP_UPDATE_ELEM`` -> push |
| 29 | - ``BPF_MAP_LOOKUP_ELEM`` -> peek |
| 30 | |
| 31 | The ``max_entries`` size that is specified at map creation time is used |
| 32 | to approximate a reasonable bitmap size for the bloom filter, and is not |
| 33 | otherwise strictly enforced. If the user wishes to insert more entries |
| 34 | into the bloom filter than ``max_entries``, this may lead to a higher |
| 35 | false positive rate. |
| 36 | |
| 37 | The number of hashes to use for the bloom filter is configurable using |
| 38 | the lower 4 bits of ``map_extra`` in ``union bpf_attr`` at map creation |
| 39 | time. If no number is specified, the default used will be 5 hash |
| 40 | functions. In general, using more hashes decreases both the false |
| 41 | positive rate and the speed of a lookup. |
| 42 | |
| 43 | It is not possible to delete elements from a bloom filter map. A bloom |
| 44 | filter map may be used as an inner map. The user is responsible for |
| 45 | synchronising concurrent updates and lookups to ensure no false negative |
| 46 | lookups occur. |
| 47 | |
| 48 | Usage |
| 49 | ===== |
| 50 | |
| 51 | Kernel BPF |
| 52 | ---------- |
| 53 | |
| 54 | bpf_map_push_elem() |
| 55 | ~~~~~~~~~~~~~~~~~~~ |
| 56 | |
| 57 | .. code-block:: c |
| 58 | |
| 59 | long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) |
| 60 | |
| 61 | A ``value`` can be added to a bloom filter using the |
| 62 | ``bpf_map_push_elem()`` helper. The ``flags`` parameter must be set to |
| 63 | ``BPF_ANY`` when adding an entry to the bloom filter. This helper |
| 64 | returns ``0`` on success, or negative error in case of failure. |
| 65 | |
| 66 | bpf_map_peek_elem() |
| 67 | ~~~~~~~~~~~~~~~~~~~ |
| 68 | |
| 69 | .. code-block:: c |
| 70 | |
| 71 | long bpf_map_peek_elem(struct bpf_map *map, void *value) |
| 72 | |
| 73 | The ``bpf_map_peek_elem()`` helper is used to determine whether |
| 74 | ``value`` is present in the bloom filter map. This helper returns ``0`` |
| 75 | if ``value`` is probably present in the map, or ``-ENOENT`` if ``value`` |
| 76 | is definitely not present in the map. |
| 77 | |
| 78 | Userspace |
| 79 | --------- |
| 80 | |
| 81 | bpf_map_update_elem() |
| 82 | ~~~~~~~~~~~~~~~~~~~~~ |
| 83 | |
| 84 | .. code-block:: c |
| 85 | |
| 86 | int bpf_map_update_elem (int fd, const void *key, const void *value, __u64 flags) |
| 87 | |
| 88 | A userspace program can add a ``value`` to a bloom filter using libbpf's |
| 89 | ``bpf_map_update_elem`` function. The ``key`` parameter must be set to |
| 90 | ``NULL`` and ``flags`` must be set to ``BPF_ANY``. Returns ``0`` on |
| 91 | success, or negative error in case of failure. |
| 92 | |
| 93 | bpf_map_lookup_elem() |
| 94 | ~~~~~~~~~~~~~~~~~~~~~ |
| 95 | |
| 96 | .. code-block:: c |
| 97 | |
| 98 | int bpf_map_lookup_elem (int fd, const void *key, void *value) |
| 99 | |
| 100 | A userspace program can determine the presence of ``value`` in a bloom |
| 101 | filter using libbpf's ``bpf_map_lookup_elem`` function. The ``key`` |
| 102 | parameter must be set to ``NULL``. Returns ``0`` if ``value`` is |
| 103 | probably present in the map, or ``-ENOENT`` if ``value`` is definitely |
| 104 | not present in the map. |
| 105 | |
| 106 | Examples |
| 107 | ======== |
| 108 | |
| 109 | Kernel BPF |
| 110 | ---------- |
| 111 | |
| 112 | This snippet shows how to declare a bloom filter in a BPF program: |
| 113 | |
| 114 | .. code-block:: c |
| 115 | |
| 116 | struct { |
| 117 | __uint(type, BPF_MAP_TYPE_BLOOM_FILTER); |
| 118 | __type(value, __u32); |
| 119 | __uint(max_entries, 1000); |
| 120 | __uint(map_extra, 3); |
| 121 | } bloom_filter SEC(".maps"); |
| 122 | |
| 123 | This snippet shows how to determine presence of a value in a bloom |
| 124 | filter in a BPF program: |
| 125 | |
| 126 | .. code-block:: c |
| 127 | |
| 128 | void *lookup(__u32 key) |
| 129 | { |
| 130 | if (bpf_map_peek_elem(&bloom_filter, &key) == 0) { |
| 131 | /* Verify not a false positive and fetch an associated |
| 132 | * value using a secondary lookup, e.g. in a hash table |
| 133 | */ |
| 134 | return bpf_map_lookup_elem(&hash_table, &key); |
| 135 | } |
| 136 | return 0; |
| 137 | } |
| 138 | |
| 139 | Userspace |
| 140 | --------- |
| 141 | |
| 142 | This snippet shows how to use libbpf to create a bloom filter map from |
| 143 | userspace: |
| 144 | |
| 145 | .. code-block:: c |
| 146 | |
| 147 | int create_bloom() |
| 148 | { |
| 149 | LIBBPF_OPTS(bpf_map_create_opts, opts, |
| 150 | .map_extra = 3); /* number of hashes */ |
| 151 | |
| 152 | return bpf_map_create(BPF_MAP_TYPE_BLOOM_FILTER, |
| 153 | "ipv6_bloom", /* name */ |
| 154 | 0, /* key size, must be zero */ |
| 155 | sizeof(ipv6_addr), /* value size */ |
| 156 | 10000, /* max entries */ |
| 157 | &opts); /* create options */ |
| 158 | } |
| 159 | |
| 160 | This snippet shows how to add an element to a bloom filter from |
| 161 | userspace: |
| 162 | |
| 163 | .. code-block:: c |
| 164 | |
| 165 | int add_element(struct bpf_map *bloom_map, __u32 value) |
| 166 | { |
| 167 | int bloom_fd = bpf_map__fd(bloom_map); |
| 168 | return bpf_map_update_elem(bloom_fd, NULL, &value, BPF_ANY); |
| 169 | } |
| 170 | |
| 171 | References |
| 172 | ========== |
| 173 | |
| 174 | https://lwn.net/ml/bpf/20210831225005.2762202-1-joannekoong@fb.com/ |