| 1 | /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com |
| 2 | * |
| 3 | * This program is free software; you can redistribute it and/or |
| 4 | * modify it under the terms of version 2 of the GNU General Public |
| 5 | * License as published by the Free Software Foundation. |
| 6 | * |
| 7 | * This program is distributed in the hope that it will be useful, but |
| 8 | * WITHOUT ANY WARRANTY; without even the implied warranty of |
| 9 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 10 | * General Public License for more details. |
| 11 | */ |
| 12 | #include <linux/bpf.h> |
| 13 | #include <linux/err.h> |
| 14 | #include <linux/vmalloc.h> |
| 15 | #include <linux/slab.h> |
| 16 | #include <linux/mm.h> |
| 17 | #include <linux/filter.h> |
| 18 | #include <linux/perf_event.h> |
| 19 | |
| 20 | /* Called from syscall */ |
| 21 | static struct bpf_map *array_map_alloc(union bpf_attr *attr) |
| 22 | { |
| 23 | struct bpf_array *array; |
| 24 | u32 elem_size, array_size; |
| 25 | |
| 26 | /* check sanity of attributes */ |
| 27 | if (attr->max_entries == 0 || attr->key_size != 4 || |
| 28 | attr->value_size == 0) |
| 29 | return ERR_PTR(-EINVAL); |
| 30 | |
| 31 | elem_size = round_up(attr->value_size, 8); |
| 32 | |
| 33 | /* check round_up into zero and u32 overflow */ |
| 34 | if (elem_size == 0 || |
| 35 | attr->max_entries > (U32_MAX - sizeof(*array)) / elem_size) |
| 36 | return ERR_PTR(-ENOMEM); |
| 37 | |
| 38 | array_size = sizeof(*array) + attr->max_entries * elem_size; |
| 39 | |
| 40 | /* allocate all map elements and zero-initialize them */ |
| 41 | array = kzalloc(array_size, GFP_USER | __GFP_NOWARN); |
| 42 | if (!array) { |
| 43 | array = vzalloc(array_size); |
| 44 | if (!array) |
| 45 | return ERR_PTR(-ENOMEM); |
| 46 | } |
| 47 | |
| 48 | /* copy mandatory map attributes */ |
| 49 | array->map.key_size = attr->key_size; |
| 50 | array->map.value_size = attr->value_size; |
| 51 | array->map.max_entries = attr->max_entries; |
| 52 | array->map.pages = round_up(array_size, PAGE_SIZE) >> PAGE_SHIFT; |
| 53 | array->elem_size = elem_size; |
| 54 | |
| 55 | return &array->map; |
| 56 | } |
| 57 | |
| 58 | /* Called from syscall or from eBPF program */ |
| 59 | static void *array_map_lookup_elem(struct bpf_map *map, void *key) |
| 60 | { |
| 61 | struct bpf_array *array = container_of(map, struct bpf_array, map); |
| 62 | u32 index = *(u32 *)key; |
| 63 | |
| 64 | if (index >= array->map.max_entries) |
| 65 | return NULL; |
| 66 | |
| 67 | return array->value + array->elem_size * index; |
| 68 | } |
| 69 | |
| 70 | /* Called from syscall */ |
| 71 | static int array_map_get_next_key(struct bpf_map *map, void *key, void *next_key) |
| 72 | { |
| 73 | struct bpf_array *array = container_of(map, struct bpf_array, map); |
| 74 | u32 index = *(u32 *)key; |
| 75 | u32 *next = (u32 *)next_key; |
| 76 | |
| 77 | if (index >= array->map.max_entries) { |
| 78 | *next = 0; |
| 79 | return 0; |
| 80 | } |
| 81 | |
| 82 | if (index == array->map.max_entries - 1) |
| 83 | return -ENOENT; |
| 84 | |
| 85 | *next = index + 1; |
| 86 | return 0; |
| 87 | } |
| 88 | |
| 89 | /* Called from syscall or from eBPF program */ |
| 90 | static int array_map_update_elem(struct bpf_map *map, void *key, void *value, |
| 91 | u64 map_flags) |
| 92 | { |
| 93 | struct bpf_array *array = container_of(map, struct bpf_array, map); |
| 94 | u32 index = *(u32 *)key; |
| 95 | |
| 96 | if (map_flags > BPF_EXIST) |
| 97 | /* unknown flags */ |
| 98 | return -EINVAL; |
| 99 | |
| 100 | if (index >= array->map.max_entries) |
| 101 | /* all elements were pre-allocated, cannot insert a new one */ |
| 102 | return -E2BIG; |
| 103 | |
| 104 | if (map_flags == BPF_NOEXIST) |
| 105 | /* all elements already exist */ |
| 106 | return -EEXIST; |
| 107 | |
| 108 | memcpy(array->value + array->elem_size * index, value, array->elem_size); |
| 109 | return 0; |
| 110 | } |
| 111 | |
| 112 | /* Called from syscall or from eBPF program */ |
| 113 | static int array_map_delete_elem(struct bpf_map *map, void *key) |
| 114 | { |
| 115 | return -EINVAL; |
| 116 | } |
| 117 | |
| 118 | /* Called when map->refcnt goes to zero, either from workqueue or from syscall */ |
| 119 | static void array_map_free(struct bpf_map *map) |
| 120 | { |
| 121 | struct bpf_array *array = container_of(map, struct bpf_array, map); |
| 122 | |
| 123 | /* at this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0, |
| 124 | * so the programs (can be more than one that used this map) were |
| 125 | * disconnected from events. Wait for outstanding programs to complete |
| 126 | * and free the array |
| 127 | */ |
| 128 | synchronize_rcu(); |
| 129 | |
| 130 | kvfree(array); |
| 131 | } |
| 132 | |
| 133 | static const struct bpf_map_ops array_ops = { |
| 134 | .map_alloc = array_map_alloc, |
| 135 | .map_free = array_map_free, |
| 136 | .map_get_next_key = array_map_get_next_key, |
| 137 | .map_lookup_elem = array_map_lookup_elem, |
| 138 | .map_update_elem = array_map_update_elem, |
| 139 | .map_delete_elem = array_map_delete_elem, |
| 140 | }; |
| 141 | |
| 142 | static struct bpf_map_type_list array_type __read_mostly = { |
| 143 | .ops = &array_ops, |
| 144 | .type = BPF_MAP_TYPE_ARRAY, |
| 145 | }; |
| 146 | |
| 147 | static int __init register_array_map(void) |
| 148 | { |
| 149 | bpf_register_map_type(&array_type); |
| 150 | return 0; |
| 151 | } |
| 152 | late_initcall(register_array_map); |
| 153 | |
| 154 | static struct bpf_map *fd_array_map_alloc(union bpf_attr *attr) |
| 155 | { |
| 156 | /* only file descriptors can be stored in this type of map */ |
| 157 | if (attr->value_size != sizeof(u32)) |
| 158 | return ERR_PTR(-EINVAL); |
| 159 | return array_map_alloc(attr); |
| 160 | } |
| 161 | |
| 162 | static void fd_array_map_free(struct bpf_map *map) |
| 163 | { |
| 164 | struct bpf_array *array = container_of(map, struct bpf_array, map); |
| 165 | int i; |
| 166 | |
| 167 | synchronize_rcu(); |
| 168 | |
| 169 | /* make sure it's empty */ |
| 170 | for (i = 0; i < array->map.max_entries; i++) |
| 171 | BUG_ON(array->ptrs[i] != NULL); |
| 172 | kvfree(array); |
| 173 | } |
| 174 | |
| 175 | static void *fd_array_map_lookup_elem(struct bpf_map *map, void *key) |
| 176 | { |
| 177 | return NULL; |
| 178 | } |
| 179 | |
| 180 | /* only called from syscall */ |
| 181 | static int fd_array_map_update_elem(struct bpf_map *map, void *key, |
| 182 | void *value, u64 map_flags) |
| 183 | { |
| 184 | struct bpf_array *array = container_of(map, struct bpf_array, map); |
| 185 | void *new_ptr, *old_ptr; |
| 186 | u32 index = *(u32 *)key, ufd; |
| 187 | |
| 188 | if (map_flags != BPF_ANY) |
| 189 | return -EINVAL; |
| 190 | |
| 191 | if (index >= array->map.max_entries) |
| 192 | return -E2BIG; |
| 193 | |
| 194 | ufd = *(u32 *)value; |
| 195 | new_ptr = map->ops->map_fd_get_ptr(map, ufd); |
| 196 | if (IS_ERR(new_ptr)) |
| 197 | return PTR_ERR(new_ptr); |
| 198 | |
| 199 | old_ptr = xchg(array->ptrs + index, new_ptr); |
| 200 | if (old_ptr) |
| 201 | map->ops->map_fd_put_ptr(old_ptr); |
| 202 | |
| 203 | return 0; |
| 204 | } |
| 205 | |
| 206 | static int fd_array_map_delete_elem(struct bpf_map *map, void *key) |
| 207 | { |
| 208 | struct bpf_array *array = container_of(map, struct bpf_array, map); |
| 209 | void *old_ptr; |
| 210 | u32 index = *(u32 *)key; |
| 211 | |
| 212 | if (index >= array->map.max_entries) |
| 213 | return -E2BIG; |
| 214 | |
| 215 | old_ptr = xchg(array->ptrs + index, NULL); |
| 216 | if (old_ptr) { |
| 217 | map->ops->map_fd_put_ptr(old_ptr); |
| 218 | return 0; |
| 219 | } else { |
| 220 | return -ENOENT; |
| 221 | } |
| 222 | } |
| 223 | |
| 224 | static void *prog_fd_array_get_ptr(struct bpf_map *map, int fd) |
| 225 | { |
| 226 | struct bpf_array *array = container_of(map, struct bpf_array, map); |
| 227 | struct bpf_prog *prog = bpf_prog_get(fd); |
| 228 | if (IS_ERR(prog)) |
| 229 | return prog; |
| 230 | |
| 231 | if (!bpf_prog_array_compatible(array, prog)) { |
| 232 | bpf_prog_put(prog); |
| 233 | return ERR_PTR(-EINVAL); |
| 234 | } |
| 235 | return prog; |
| 236 | } |
| 237 | |
| 238 | static void prog_fd_array_put_ptr(void *ptr) |
| 239 | { |
| 240 | struct bpf_prog *prog = ptr; |
| 241 | |
| 242 | bpf_prog_put_rcu(prog); |
| 243 | } |
| 244 | |
| 245 | /* decrement refcnt of all bpf_progs that are stored in this map */ |
| 246 | void bpf_fd_array_map_clear(struct bpf_map *map) |
| 247 | { |
| 248 | struct bpf_array *array = container_of(map, struct bpf_array, map); |
| 249 | int i; |
| 250 | |
| 251 | for (i = 0; i < array->map.max_entries; i++) |
| 252 | fd_array_map_delete_elem(map, &i); |
| 253 | } |
| 254 | |
| 255 | static const struct bpf_map_ops prog_array_ops = { |
| 256 | .map_alloc = fd_array_map_alloc, |
| 257 | .map_free = fd_array_map_free, |
| 258 | .map_get_next_key = array_map_get_next_key, |
| 259 | .map_lookup_elem = fd_array_map_lookup_elem, |
| 260 | .map_update_elem = fd_array_map_update_elem, |
| 261 | .map_delete_elem = fd_array_map_delete_elem, |
| 262 | .map_fd_get_ptr = prog_fd_array_get_ptr, |
| 263 | .map_fd_put_ptr = prog_fd_array_put_ptr, |
| 264 | }; |
| 265 | |
| 266 | static struct bpf_map_type_list prog_array_type __read_mostly = { |
| 267 | .ops = &prog_array_ops, |
| 268 | .type = BPF_MAP_TYPE_PROG_ARRAY, |
| 269 | }; |
| 270 | |
| 271 | static int __init register_prog_array_map(void) |
| 272 | { |
| 273 | bpf_register_map_type(&prog_array_type); |
| 274 | return 0; |
| 275 | } |
| 276 | late_initcall(register_prog_array_map); |
| 277 | |
| 278 | static void perf_event_array_map_free(struct bpf_map *map) |
| 279 | { |
| 280 | bpf_fd_array_map_clear(map); |
| 281 | fd_array_map_free(map); |
| 282 | } |
| 283 | |
| 284 | static void *perf_event_fd_array_get_ptr(struct bpf_map *map, int fd) |
| 285 | { |
| 286 | struct perf_event *event; |
| 287 | const struct perf_event_attr *attr; |
| 288 | |
| 289 | event = perf_event_get(fd); |
| 290 | if (IS_ERR(event)) |
| 291 | return event; |
| 292 | |
| 293 | attr = perf_event_attrs(event); |
| 294 | if (IS_ERR(attr)) |
| 295 | return (void *)attr; |
| 296 | |
| 297 | if (attr->type != PERF_TYPE_RAW && |
| 298 | !(attr->type == PERF_TYPE_SOFTWARE && |
| 299 | attr->config == PERF_COUNT_SW_BPF_OUTPUT) && |
| 300 | attr->type != PERF_TYPE_HARDWARE) { |
| 301 | perf_event_release_kernel(event); |
| 302 | return ERR_PTR(-EINVAL); |
| 303 | } |
| 304 | return event; |
| 305 | } |
| 306 | |
| 307 | static void perf_event_fd_array_put_ptr(void *ptr) |
| 308 | { |
| 309 | struct perf_event *event = ptr; |
| 310 | |
| 311 | perf_event_release_kernel(event); |
| 312 | } |
| 313 | |
| 314 | static const struct bpf_map_ops perf_event_array_ops = { |
| 315 | .map_alloc = fd_array_map_alloc, |
| 316 | .map_free = perf_event_array_map_free, |
| 317 | .map_get_next_key = array_map_get_next_key, |
| 318 | .map_lookup_elem = fd_array_map_lookup_elem, |
| 319 | .map_update_elem = fd_array_map_update_elem, |
| 320 | .map_delete_elem = fd_array_map_delete_elem, |
| 321 | .map_fd_get_ptr = perf_event_fd_array_get_ptr, |
| 322 | .map_fd_put_ptr = perf_event_fd_array_put_ptr, |
| 323 | }; |
| 324 | |
| 325 | static struct bpf_map_type_list perf_event_array_type __read_mostly = { |
| 326 | .ops = &perf_event_array_ops, |
| 327 | .type = BPF_MAP_TYPE_PERF_EVENT_ARRAY, |
| 328 | }; |
| 329 | |
| 330 | static int __init register_perf_event_array_map(void) |
| 331 | { |
| 332 | bpf_register_map_type(&perf_event_array_type); |
| 333 | return 0; |
| 334 | } |
| 335 | late_initcall(register_perf_event_array_map); |