github.com/yaling888/clash@v1.53.0/component/ebpf/bpf/bpf_helper_defs.h (about) 1 /* This is auto-generated file. See bpf_doc.py for details. */ 2 3 /* Forward declarations of BPF structs */ 4 struct bpf_fib_lookup; 5 struct bpf_sk_lookup; 6 struct bpf_perf_event_data; 7 struct bpf_perf_event_value; 8 struct bpf_pidns_info; 9 struct bpf_redir_neigh; 10 struct bpf_sock; 11 struct bpf_sock_addr; 12 struct bpf_sock_ops; 13 struct bpf_sock_tuple; 14 struct bpf_spin_lock; 15 struct bpf_sysctl; 16 struct bpf_tcp_sock; 17 struct bpf_tunnel_key; 18 struct bpf_xfrm_state; 19 struct linux_binprm; 20 struct pt_regs; 21 struct sk_reuseport_md; 22 struct sockaddr; 23 struct tcphdr; 24 struct seq_file; 25 struct tcp6_sock; 26 struct tcp_sock; 27 struct tcp_timewait_sock; 28 struct tcp_request_sock; 29 struct udp6_sock; 30 struct unix_sock; 31 struct task_struct; 32 struct __sk_buff; 33 struct sk_msg_md; 34 struct xdp_md; 35 struct path; 36 struct btf_ptr; 37 struct inode; 38 struct socket; 39 struct file; 40 struct bpf_timer; 41 42 /* 43 * bpf_map_lookup_elem 44 * 45 * Perform a lookup in *map* for an entry associated to *key*. 46 * 47 * Returns 48 * Map value associated to *key*, or **NULL** if no entry was 49 * found. 50 */ 51 static void *(*bpf_map_lookup_elem)(void *map, const void *key) = (void *) 1; 52 53 /* 54 * bpf_map_update_elem 55 * 56 * Add or update the value of the entry associated to *key* in 57 * *map* with *value*. *flags* is one of: 58 * 59 * **BPF_NOEXIST** 60 * The entry for *key* must not exist in the map. 61 * **BPF_EXIST** 62 * The entry for *key* must already exist in the map. 63 * **BPF_ANY** 64 * No condition on the existence of the entry for *key*. 65 * 66 * Flag value **BPF_NOEXIST** cannot be used for maps of types 67 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 68 * elements always exist), the helper would return an error. 69 * 70 * Returns 71 * 0 on success, or a negative error in case of failure. 72 */ 73 static long (*bpf_map_update_elem)(void *map, const void *key, const void *value, __u64 flags) = (void *) 2; 74 75 /* 76 * bpf_map_delete_elem 77 * 78 * Delete entry with *key* from *map*. 79 * 80 * Returns 81 * 0 on success, or a negative error in case of failure. 82 */ 83 static long (*bpf_map_delete_elem)(void *map, const void *key) = (void *) 3; 84 85 /* 86 * bpf_probe_read 87 * 88 * For tracing programs, safely attempt to read *size* bytes from 89 * kernel space address *unsafe_ptr* and store the data in *dst*. 90 * 91 * Generally, use **bpf_probe_read_user**\ () or 92 * **bpf_probe_read_kernel**\ () instead. 93 * 94 * Returns 95 * 0 on success, or a negative error in case of failure. 96 */ 97 static long (*bpf_probe_read)(void *dst, __u32 size, const void *unsafe_ptr) = (void *) 4; 98 99 /* 100 * bpf_ktime_get_ns 101 * 102 * Return the time elapsed since system boot, in nanoseconds. 103 * Does not include time the system was suspended. 104 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 105 * 106 * Returns 107 * Current *ktime*. 108 */ 109 static __u64 (*bpf_ktime_get_ns)(void) = (void *) 5; 110 111 /* 112 * bpf_trace_printk 113 * 114 * This helper is a "printk()-like" facility for debugging. It 115 * prints a message defined by format *fmt* (of size *fmt_size*) 116 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 117 * available. It can take up to three additional **u64** 118 * arguments (as an eBPF helpers, the total number of arguments is 119 * limited to five). 120 * 121 * Each time the helper is called, it appends a line to the trace. 122 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 123 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 124 * The format of the trace is customizable, and the exact output 125 * one will get depends on the options set in 126 * *\/sys/kernel/debug/tracing/trace_options* (see also the 127 * *README* file under the same directory). However, it usually 128 * defaults to something like: 129 * 130 * :: 131 * 132 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 133 * 134 * In the above: 135 * 136 * * ``telnet`` is the name of the current task. 137 * * ``470`` is the PID of the current task. 138 * * ``001`` is the CPU number on which the task is 139 * running. 140 * * In ``.N..``, each character refers to a set of 141 * options (whether irqs are enabled, scheduling 142 * options, whether hard/softirqs are running, level of 143 * preempt_disabled respectively). **N** means that 144 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 145 * are set. 146 * * ``419421.045894`` is a timestamp. 147 * * ``0x00000001`` is a fake value used by BPF for the 148 * instruction pointer register. 149 * * ``<formatted msg>`` is the message formatted with 150 * *fmt*. 151 * 152 * The conversion specifiers supported by *fmt* are similar, but 153 * more limited than for printk(). They are **%d**, **%i**, 154 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 155 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 156 * of field, padding with zeroes, etc.) is available, and the 157 * helper will return **-EINVAL** (but print nothing) if it 158 * encounters an unknown specifier. 159 * 160 * Also, note that **bpf_trace_printk**\ () is slow, and should 161 * only be used for debugging purposes. For this reason, a notice 162 * block (spanning several lines) is printed to kernel logs and 163 * states that the helper should not be used "for production use" 164 * the first time this helper is used (or more precisely, when 165 * **trace_printk**\ () buffers are allocated). For passing values 166 * to user space, perf events should be preferred. 167 * 168 * Returns 169 * The number of bytes written to the buffer, or a negative error 170 * in case of failure. 171 */ 172 static long (*bpf_trace_printk)(const char *fmt, __u32 fmt_size, ...) = (void *) 6; 173 174 /* 175 * bpf_get_prandom_u32 176 * 177 * Get a pseudo-random number. 178 * 179 * From a security point of view, this helper uses its own 180 * pseudo-random internal state, and cannot be used to infer the 181 * seed of other random functions in the kernel. However, it is 182 * essential to note that the generator used by the helper is not 183 * cryptographically secure. 184 * 185 * Returns 186 * A random 32-bit unsigned value. 187 */ 188 static __u32 (*bpf_get_prandom_u32)(void) = (void *) 7; 189 190 /* 191 * bpf_get_smp_processor_id 192 * 193 * Get the SMP (symmetric multiprocessing) processor id. Note that 194 * all programs run with migration disabled, which means that the 195 * SMP processor id is stable during all the execution of the 196 * program. 197 * 198 * Returns 199 * The SMP id of the processor running the program. 200 */ 201 static __u32 (*bpf_get_smp_processor_id)(void) = (void *) 8; 202 203 /* 204 * bpf_skb_store_bytes 205 * 206 * Store *len* bytes from address *from* into the packet 207 * associated to *skb*, at *offset*. *flags* are a combination of 208 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 209 * checksum for the packet after storing the bytes) and 210 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 211 * **->swhash** and *skb*\ **->l4hash** to 0). 212 * 213 * A call to this helper is susceptible to change the underlying 214 * packet buffer. Therefore, at load time, all checks on pointers 215 * previously done by the verifier are invalidated and must be 216 * performed again, if the helper is used in combination with 217 * direct packet access. 218 * 219 * Returns 220 * 0 on success, or a negative error in case of failure. 221 */ 222 static long (*bpf_skb_store_bytes)(struct __sk_buff *skb, __u32 offset, const void *from, __u32 len, __u64 flags) = (void *) 9; 223 224 /* 225 * bpf_l3_csum_replace 226 * 227 * Recompute the layer 3 (e.g. IP) checksum for the packet 228 * associated to *skb*. Computation is incremental, so the helper 229 * must know the former value of the header field that was 230 * modified (*from*), the new value of this field (*to*), and the 231 * number of bytes (2 or 4) for this field, stored in *size*. 232 * Alternatively, it is possible to store the difference between 233 * the previous and the new values of the header field in *to*, by 234 * setting *from* and *size* to 0. For both methods, *offset* 235 * indicates the location of the IP checksum within the packet. 236 * 237 * This helper works in combination with **bpf_csum_diff**\ (), 238 * which does not update the checksum in-place, but offers more 239 * flexibility and can handle sizes larger than 2 or 4 for the 240 * checksum to update. 241 * 242 * A call to this helper is susceptible to change the underlying 243 * packet buffer. Therefore, at load time, all checks on pointers 244 * previously done by the verifier are invalidated and must be 245 * performed again, if the helper is used in combination with 246 * direct packet access. 247 * 248 * Returns 249 * 0 on success, or a negative error in case of failure. 250 */ 251 static long (*bpf_l3_csum_replace)(struct __sk_buff *skb, __u32 offset, __u64 from, __u64 to, __u64 size) = (void *) 10; 252 253 /* 254 * bpf_l4_csum_replace 255 * 256 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 257 * packet associated to *skb*. Computation is incremental, so the 258 * helper must know the former value of the header field that was 259 * modified (*from*), the new value of this field (*to*), and the 260 * number of bytes (2 or 4) for this field, stored on the lowest 261 * four bits of *flags*. Alternatively, it is possible to store 262 * the difference between the previous and the new values of the 263 * header field in *to*, by setting *from* and the four lowest 264 * bits of *flags* to 0. For both methods, *offset* indicates the 265 * location of the IP checksum within the packet. In addition to 266 * the size of the field, *flags* can be added (bitwise OR) actual 267 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 268 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 269 * for updates resulting in a null checksum the value is set to 270 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 271 * the checksum is to be computed against a pseudo-header. 272 * 273 * This helper works in combination with **bpf_csum_diff**\ (), 274 * which does not update the checksum in-place, but offers more 275 * flexibility and can handle sizes larger than 2 or 4 for the 276 * checksum to update. 277 * 278 * A call to this helper is susceptible to change the underlying 279 * packet buffer. Therefore, at load time, all checks on pointers 280 * previously done by the verifier are invalidated and must be 281 * performed again, if the helper is used in combination with 282 * direct packet access. 283 * 284 * Returns 285 * 0 on success, or a negative error in case of failure. 286 */ 287 static long (*bpf_l4_csum_replace)(struct __sk_buff *skb, __u32 offset, __u64 from, __u64 to, __u64 flags) = (void *) 11; 288 289 /* 290 * bpf_tail_call 291 * 292 * This special helper is used to trigger a "tail call", or in 293 * other words, to jump into another eBPF program. The same stack 294 * frame is used (but values on stack and in registers for the 295 * caller are not accessible to the callee). This mechanism allows 296 * for program chaining, either for raising the maximum number of 297 * available eBPF instructions, or to execute given programs in 298 * conditional blocks. For security reasons, there is an upper 299 * limit to the number of successive tail calls that can be 300 * performed. 301 * 302 * Upon call of this helper, the program attempts to jump into a 303 * program referenced at index *index* in *prog_array_map*, a 304 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 305 * *ctx*, a pointer to the context. 306 * 307 * If the call succeeds, the kernel immediately runs the first 308 * instruction of the new program. This is not a function call, 309 * and it never returns to the previous program. If the call 310 * fails, then the helper has no effect, and the caller continues 311 * to run its subsequent instructions. A call can fail if the 312 * destination program for the jump does not exist (i.e. *index* 313 * is superior to the number of entries in *prog_array_map*), or 314 * if the maximum number of tail calls has been reached for this 315 * chain of programs. This limit is defined in the kernel by the 316 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 317 * which is currently set to 33. 318 * 319 * Returns 320 * 0 on success, or a negative error in case of failure. 321 */ 322 static long (*bpf_tail_call)(void *ctx, void *prog_array_map, __u32 index) = (void *) 12; 323 324 /* 325 * bpf_clone_redirect 326 * 327 * Clone and redirect the packet associated to *skb* to another 328 * net device of index *ifindex*. Both ingress and egress 329 * interfaces can be used for redirection. The **BPF_F_INGRESS** 330 * value in *flags* is used to make the distinction (ingress path 331 * is selected if the flag is present, egress path otherwise). 332 * This is the only flag supported for now. 333 * 334 * In comparison with **bpf_redirect**\ () helper, 335 * **bpf_clone_redirect**\ () has the associated cost of 336 * duplicating the packet buffer, but this can be executed out of 337 * the eBPF program. Conversely, **bpf_redirect**\ () is more 338 * efficient, but it is handled through an action code where the 339 * redirection happens only after the eBPF program has returned. 340 * 341 * A call to this helper is susceptible to change the underlying 342 * packet buffer. Therefore, at load time, all checks on pointers 343 * previously done by the verifier are invalidated and must be 344 * performed again, if the helper is used in combination with 345 * direct packet access. 346 * 347 * Returns 348 * 0 on success, or a negative error in case of failure. 349 */ 350 static long (*bpf_clone_redirect)(struct __sk_buff *skb, __u32 ifindex, __u64 flags) = (void *) 13; 351 352 /* 353 * bpf_get_current_pid_tgid 354 * 355 * 356 * Returns 357 * A 64-bit integer containing the current tgid and pid, and 358 * created as such: 359 * *current_task*\ **->tgid << 32 \|** 360 * *current_task*\ **->pid**. 361 */ 362 static __u64 (*bpf_get_current_pid_tgid)(void) = (void *) 14; 363 364 /* 365 * bpf_get_current_uid_gid 366 * 367 * 368 * Returns 369 * A 64-bit integer containing the current GID and UID, and 370 * created as such: *current_gid* **<< 32 \|** *current_uid*. 371 */ 372 static __u64 (*bpf_get_current_uid_gid)(void) = (void *) 15; 373 374 /* 375 * bpf_get_current_comm 376 * 377 * Copy the **comm** attribute of the current task into *buf* of 378 * *size_of_buf*. The **comm** attribute contains the name of 379 * the executable (excluding the path) for the current task. The 380 * *size_of_buf* must be strictly positive. On success, the 381 * helper makes sure that the *buf* is NUL-terminated. On failure, 382 * it is filled with zeroes. 383 * 384 * Returns 385 * 0 on success, or a negative error in case of failure. 386 */ 387 static long (*bpf_get_current_comm)(void *buf, __u32 size_of_buf) = (void *) 16; 388 389 /* 390 * bpf_get_cgroup_classid 391 * 392 * Retrieve the classid for the current task, i.e. for the net_cls 393 * cgroup to which *skb* belongs. 394 * 395 * This helper can be used on TC egress path, but not on ingress. 396 * 397 * The net_cls cgroup provides an interface to tag network packets 398 * based on a user-provided identifier for all traffic coming from 399 * the tasks belonging to the related cgroup. See also the related 400 * kernel documentation, available from the Linux sources in file 401 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 402 * 403 * The Linux kernel has two versions for cgroups: there are 404 * cgroups v1 and cgroups v2. Both are available to users, who can 405 * use a mixture of them, but note that the net_cls cgroup is for 406 * cgroup v1 only. This makes it incompatible with BPF programs 407 * run on cgroups, which is a cgroup-v2-only feature (a socket can 408 * only hold data for one version of cgroups at a time). 409 * 410 * This helper is only available is the kernel was compiled with 411 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 412 * "**y**" or to "**m**". 413 * 414 * Returns 415 * The classid, or 0 for the default unconfigured classid. 416 */ 417 static __u32 (*bpf_get_cgroup_classid)(struct __sk_buff *skb) = (void *) 17; 418 419 /* 420 * bpf_skb_vlan_push 421 * 422 * Push a *vlan_tci* (VLAN tag control information) of protocol 423 * *vlan_proto* to the packet associated to *skb*, then update 424 * the checksum. Note that if *vlan_proto* is different from 425 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 426 * be **ETH_P_8021Q**. 427 * 428 * A call to this helper is susceptible to change the underlying 429 * packet buffer. Therefore, at load time, all checks on pointers 430 * previously done by the verifier are invalidated and must be 431 * performed again, if the helper is used in combination with 432 * direct packet access. 433 * 434 * Returns 435 * 0 on success, or a negative error in case of failure. 436 */ 437 static long (*bpf_skb_vlan_push)(struct __sk_buff *skb, __be16 vlan_proto, __u16 vlan_tci) = (void *) 18; 438 439 /* 440 * bpf_skb_vlan_pop 441 * 442 * Pop a VLAN header from the packet associated to *skb*. 443 * 444 * A call to this helper is susceptible to change the underlying 445 * packet buffer. Therefore, at load time, all checks on pointers 446 * previously done by the verifier are invalidated and must be 447 * performed again, if the helper is used in combination with 448 * direct packet access. 449 * 450 * Returns 451 * 0 on success, or a negative error in case of failure. 452 */ 453 static long (*bpf_skb_vlan_pop)(struct __sk_buff *skb) = (void *) 19; 454 455 /* 456 * bpf_skb_get_tunnel_key 457 * 458 * Get tunnel metadata. This helper takes a pointer *key* to an 459 * empty **struct bpf_tunnel_key** of **size**, that will be 460 * filled with tunnel metadata for the packet associated to *skb*. 461 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 462 * indicates that the tunnel is based on IPv6 protocol instead of 463 * IPv4. 464 * 465 * The **struct bpf_tunnel_key** is an object that generalizes the 466 * principal parameters used by various tunneling protocols into a 467 * single struct. This way, it can be used to easily make a 468 * decision based on the contents of the encapsulation header, 469 * "summarized" in this struct. In particular, it holds the IP 470 * address of the remote end (IPv4 or IPv6, depending on the case) 471 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 472 * this struct exposes the *key*\ **->tunnel_id**, which is 473 * generally mapped to a VNI (Virtual Network Identifier), making 474 * it programmable together with the **bpf_skb_set_tunnel_key**\ 475 * () helper. 476 * 477 * Let's imagine that the following code is part of a program 478 * attached to the TC ingress interface, on one end of a GRE 479 * tunnel, and is supposed to filter out all messages coming from 480 * remote ends with IPv4 address other than 10.0.0.1: 481 * 482 * :: 483 * 484 * int ret; 485 * struct bpf_tunnel_key key = {}; 486 * 487 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 488 * if (ret < 0) 489 * return TC_ACT_SHOT; // drop packet 490 * 491 * if (key.remote_ipv4 != 0x0a000001) 492 * return TC_ACT_SHOT; // drop packet 493 * 494 * return TC_ACT_OK; // accept packet 495 * 496 * This interface can also be used with all encapsulation devices 497 * that can operate in "collect metadata" mode: instead of having 498 * one network device per specific configuration, the "collect 499 * metadata" mode only requires a single device where the 500 * configuration can be extracted from this helper. 501 * 502 * This can be used together with various tunnels such as VXLan, 503 * Geneve, GRE or IP in IP (IPIP). 504 * 505 * Returns 506 * 0 on success, or a negative error in case of failure. 507 */ 508 static long (*bpf_skb_get_tunnel_key)(struct __sk_buff *skb, struct bpf_tunnel_key *key, __u32 size, __u64 flags) = (void *) 20; 509 510 /* 511 * bpf_skb_set_tunnel_key 512 * 513 * Populate tunnel metadata for packet associated to *skb.* The 514 * tunnel metadata is set to the contents of *key*, of *size*. The 515 * *flags* can be set to a combination of the following values: 516 * 517 * **BPF_F_TUNINFO_IPV6** 518 * Indicate that the tunnel is based on IPv6 protocol 519 * instead of IPv4. 520 * **BPF_F_ZERO_CSUM_TX** 521 * For IPv4 packets, add a flag to tunnel metadata 522 * indicating that checksum computation should be skipped 523 * and checksum set to zeroes. 524 * **BPF_F_DONT_FRAGMENT** 525 * Add a flag to tunnel metadata indicating that the 526 * packet should not be fragmented. 527 * **BPF_F_SEQ_NUMBER** 528 * Add a flag to tunnel metadata indicating that a 529 * sequence number should be added to tunnel header before 530 * sending the packet. This flag was added for GRE 531 * encapsulation, but might be used with other protocols 532 * as well in the future. 533 * 534 * Here is a typical usage on the transmit path: 535 * 536 * :: 537 * 538 * struct bpf_tunnel_key key; 539 * populate key ... 540 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 541 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 542 * 543 * See also the description of the **bpf_skb_get_tunnel_key**\ () 544 * helper for additional information. 545 * 546 * Returns 547 * 0 on success, or a negative error in case of failure. 548 */ 549 static long (*bpf_skb_set_tunnel_key)(struct __sk_buff *skb, struct bpf_tunnel_key *key, __u32 size, __u64 flags) = (void *) 21; 550 551 /* 552 * bpf_perf_event_read 553 * 554 * Read the value of a perf event counter. This helper relies on a 555 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 556 * the perf event counter is selected when *map* is updated with 557 * perf event file descriptors. The *map* is an array whose size 558 * is the number of available CPUs, and each cell contains a value 559 * relative to one CPU. The value to retrieve is indicated by 560 * *flags*, that contains the index of the CPU to look up, masked 561 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 562 * **BPF_F_CURRENT_CPU** to indicate that the value for the 563 * current CPU should be retrieved. 564 * 565 * Note that before Linux 4.13, only hardware perf event can be 566 * retrieved. 567 * 568 * Also, be aware that the newer helper 569 * **bpf_perf_event_read_value**\ () is recommended over 570 * **bpf_perf_event_read**\ () in general. The latter has some ABI 571 * quirks where error and counter value are used as a return code 572 * (which is wrong to do since ranges may overlap). This issue is 573 * fixed with **bpf_perf_event_read_value**\ (), which at the same 574 * time provides more features over the **bpf_perf_event_read**\ 575 * () interface. Please refer to the description of 576 * **bpf_perf_event_read_value**\ () for details. 577 * 578 * Returns 579 * The value of the perf event counter read from the map, or a 580 * negative error code in case of failure. 581 */ 582 static __u64 (*bpf_perf_event_read)(void *map, __u64 flags) = (void *) 22; 583 584 /* 585 * bpf_redirect 586 * 587 * Redirect the packet to another net device of index *ifindex*. 588 * This helper is somewhat similar to **bpf_clone_redirect**\ 589 * (), except that the packet is not cloned, which provides 590 * increased performance. 591 * 592 * Except for XDP, both ingress and egress interfaces can be used 593 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 594 * to make the distinction (ingress path is selected if the flag 595 * is present, egress path otherwise). Currently, XDP only 596 * supports redirection to the egress interface, and accepts no 597 * flag at all. 598 * 599 * The same effect can also be attained with the more generic 600 * **bpf_redirect_map**\ (), which uses a BPF map to store the 601 * redirect target instead of providing it directly to the helper. 602 * 603 * Returns 604 * For XDP, the helper returns **XDP_REDIRECT** on success or 605 * **XDP_ABORTED** on error. For other program types, the values 606 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 607 * error. 608 */ 609 static long (*bpf_redirect)(__u32 ifindex, __u64 flags) = (void *) 23; 610 611 /* 612 * bpf_get_route_realm 613 * 614 * Retrieve the realm or the route, that is to say the 615 * **tclassid** field of the destination for the *skb*. The 616 * identifier retrieved is a user-provided tag, similar to the 617 * one used with the net_cls cgroup (see description for 618 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 619 * held by a route (a destination entry), not by a task. 620 * 621 * Retrieving this identifier works with the clsact TC egress hook 622 * (see also **tc-bpf(8)**), or alternatively on conventional 623 * classful egress qdiscs, but not on TC ingress path. In case of 624 * clsact TC egress hook, this has the advantage that, internally, 625 * the destination entry has not been dropped yet in the transmit 626 * path. Therefore, the destination entry does not need to be 627 * artificially held via **netif_keep_dst**\ () for a classful 628 * qdisc until the *skb* is freed. 629 * 630 * This helper is available only if the kernel was compiled with 631 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 632 * 633 * Returns 634 * The realm of the route for the packet associated to *skb*, or 0 635 * if none was found. 636 */ 637 static __u32 (*bpf_get_route_realm)(struct __sk_buff *skb) = (void *) 24; 638 639 /* 640 * bpf_perf_event_output 641 * 642 * Write raw *data* blob into a special BPF perf event held by 643 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 644 * event must have the following attributes: **PERF_SAMPLE_RAW** 645 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 646 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 647 * 648 * The *flags* are used to indicate the index in *map* for which 649 * the value must be put, masked with **BPF_F_INDEX_MASK**. 650 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 651 * to indicate that the index of the current CPU core should be 652 * used. 653 * 654 * The value to write, of *size*, is passed through eBPF stack and 655 * pointed by *data*. 656 * 657 * The context of the program *ctx* needs also be passed to the 658 * helper. 659 * 660 * On user space, a program willing to read the values needs to 661 * call **perf_event_open**\ () on the perf event (either for 662 * one or for all CPUs) and to store the file descriptor into the 663 * *map*. This must be done before the eBPF program can send data 664 * into it. An example is available in file 665 * *samples/bpf/trace_output_user.c* in the Linux kernel source 666 * tree (the eBPF program counterpart is in 667 * *samples/bpf/trace_output_kern.c*). 668 * 669 * **bpf_perf_event_output**\ () achieves better performance 670 * than **bpf_trace_printk**\ () for sharing data with user 671 * space, and is much better suitable for streaming data from eBPF 672 * programs. 673 * 674 * Note that this helper is not restricted to tracing use cases 675 * and can be used with programs attached to TC or XDP as well, 676 * where it allows for passing data to user space listeners. Data 677 * can be: 678 * 679 * * Only custom structs, 680 * * Only the packet payload, or 681 * * A combination of both. 682 * 683 * Returns 684 * 0 on success, or a negative error in case of failure. 685 */ 686 static long (*bpf_perf_event_output)(void *ctx, void *map, __u64 flags, void *data, __u64 size) = (void *) 25; 687 688 /* 689 * bpf_skb_load_bytes 690 * 691 * This helper was provided as an easy way to load data from a 692 * packet. It can be used to load *len* bytes from *offset* from 693 * the packet associated to *skb*, into the buffer pointed by 694 * *to*. 695 * 696 * Since Linux 4.7, usage of this helper has mostly been replaced 697 * by "direct packet access", enabling packet data to be 698 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 699 * pointing respectively to the first byte of packet data and to 700 * the byte after the last byte of packet data. However, it 701 * remains useful if one wishes to read large quantities of data 702 * at once from a packet into the eBPF stack. 703 * 704 * Returns 705 * 0 on success, or a negative error in case of failure. 706 */ 707 static long (*bpf_skb_load_bytes)(const void *skb, __u32 offset, void *to, __u32 len) = (void *) 26; 708 709 /* 710 * bpf_get_stackid 711 * 712 * Walk a user or a kernel stack and return its id. To achieve 713 * this, the helper needs *ctx*, which is a pointer to the context 714 * on which the tracing program is executed, and a pointer to a 715 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 716 * 717 * The last argument, *flags*, holds the number of stack frames to 718 * skip (from 0 to 255), masked with 719 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 720 * a combination of the following flags: 721 * 722 * **BPF_F_USER_STACK** 723 * Collect a user space stack instead of a kernel stack. 724 * **BPF_F_FAST_STACK_CMP** 725 * Compare stacks by hash only. 726 * **BPF_F_REUSE_STACKID** 727 * If two different stacks hash into the same *stackid*, 728 * discard the old one. 729 * 730 * The stack id retrieved is a 32 bit long integer handle which 731 * can be further combined with other data (including other stack 732 * ids) and used as a key into maps. This can be useful for 733 * generating a variety of graphs (such as flame graphs or off-cpu 734 * graphs). 735 * 736 * For walking a stack, this helper is an improvement over 737 * **bpf_probe_read**\ (), which can be used with unrolled loops 738 * but is not efficient and consumes a lot of eBPF instructions. 739 * Instead, **bpf_get_stackid**\ () can collect up to 740 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 741 * this limit can be controlled with the **sysctl** program, and 742 * that it should be manually increased in order to profile long 743 * user stacks (such as stacks for Java programs). To do so, use: 744 * 745 * :: 746 * 747 * # sysctl kernel.perf_event_max_stack=<new value> 748 * 749 * Returns 750 * The positive or null stack id on success, or a negative error 751 * in case of failure. 752 */ 753 static long (*bpf_get_stackid)(void *ctx, void *map, __u64 flags) = (void *) 27; 754 755 /* 756 * bpf_csum_diff 757 * 758 * Compute a checksum difference, from the raw buffer pointed by 759 * *from*, of length *from_size* (that must be a multiple of 4), 760 * towards the raw buffer pointed by *to*, of size *to_size* 761 * (same remark). An optional *seed* can be added to the value 762 * (this can be cascaded, the seed may come from a previous call 763 * to the helper). 764 * 765 * This is flexible enough to be used in several ways: 766 * 767 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 768 * checksum, it can be used when pushing new data. 769 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 770 * checksum, it can be used when removing data from a packet. 771 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 772 * can be used to compute a diff. Note that *from_size* and 773 * *to_size* do not need to be equal. 774 * 775 * This helper can be used in combination with 776 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 777 * which one can feed in the difference computed with 778 * **bpf_csum_diff**\ (). 779 * 780 * Returns 781 * The checksum result, or a negative error code in case of 782 * failure. 783 */ 784 static __s64 (*bpf_csum_diff)(__be32 *from, __u32 from_size, __be32 *to, __u32 to_size, __wsum seed) = (void *) 28; 785 786 /* 787 * bpf_skb_get_tunnel_opt 788 * 789 * Retrieve tunnel options metadata for the packet associated to 790 * *skb*, and store the raw tunnel option data to the buffer *opt* 791 * of *size*. 792 * 793 * This helper can be used with encapsulation devices that can 794 * operate in "collect metadata" mode (please refer to the related 795 * note in the description of **bpf_skb_get_tunnel_key**\ () for 796 * more details). A particular example where this can be used is 797 * in combination with the Geneve encapsulation protocol, where it 798 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 799 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 800 * the eBPF program. This allows for full customization of these 801 * headers. 802 * 803 * Returns 804 * The size of the option data retrieved. 805 */ 806 static long (*bpf_skb_get_tunnel_opt)(struct __sk_buff *skb, void *opt, __u32 size) = (void *) 29; 807 808 /* 809 * bpf_skb_set_tunnel_opt 810 * 811 * Set tunnel options metadata for the packet associated to *skb* 812 * to the option data contained in the raw buffer *opt* of *size*. 813 * 814 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 815 * helper for additional information. 816 * 817 * Returns 818 * 0 on success, or a negative error in case of failure. 819 */ 820 static long (*bpf_skb_set_tunnel_opt)(struct __sk_buff *skb, void *opt, __u32 size) = (void *) 30; 821 822 /* 823 * bpf_skb_change_proto 824 * 825 * Change the protocol of the *skb* to *proto*. Currently 826 * supported are transition from IPv4 to IPv6, and from IPv6 to 827 * IPv4. The helper takes care of the groundwork for the 828 * transition, including resizing the socket buffer. The eBPF 829 * program is expected to fill the new headers, if any, via 830 * **skb_store_bytes**\ () and to recompute the checksums with 831 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 832 * (). The main case for this helper is to perform NAT64 833 * operations out of an eBPF program. 834 * 835 * Internally, the GSO type is marked as dodgy so that headers are 836 * checked and segments are recalculated by the GSO/GRO engine. 837 * The size for GSO target is adapted as well. 838 * 839 * All values for *flags* are reserved for future usage, and must 840 * be left at zero. 841 * 842 * A call to this helper is susceptible to change the underlying 843 * packet buffer. Therefore, at load time, all checks on pointers 844 * previously done by the verifier are invalidated and must be 845 * performed again, if the helper is used in combination with 846 * direct packet access. 847 * 848 * Returns 849 * 0 on success, or a negative error in case of failure. 850 */ 851 static long (*bpf_skb_change_proto)(struct __sk_buff *skb, __be16 proto, __u64 flags) = (void *) 31; 852 853 /* 854 * bpf_skb_change_type 855 * 856 * Change the packet type for the packet associated to *skb*. This 857 * comes down to setting *skb*\ **->pkt_type** to *type*, except 858 * the eBPF program does not have a write access to *skb*\ 859 * **->pkt_type** beside this helper. Using a helper here allows 860 * for graceful handling of errors. 861 * 862 * The major use case is to change incoming *skb*s to 863 * **PACKET_HOST** in a programmatic way instead of having to 864 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 865 * example. 866 * 867 * Note that *type* only allows certain values. At this time, they 868 * are: 869 * 870 * **PACKET_HOST** 871 * Packet is for us. 872 * **PACKET_BROADCAST** 873 * Send packet to all. 874 * **PACKET_MULTICAST** 875 * Send packet to group. 876 * **PACKET_OTHERHOST** 877 * Send packet to someone else. 878 * 879 * Returns 880 * 0 on success, or a negative error in case of failure. 881 */ 882 static long (*bpf_skb_change_type)(struct __sk_buff *skb, __u32 type) = (void *) 32; 883 884 /* 885 * bpf_skb_under_cgroup 886 * 887 * Check whether *skb* is a descendant of the cgroup2 held by 888 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 889 * 890 * Returns 891 * The return value depends on the result of the test, and can be: 892 * 893 * * 0, if the *skb* failed the cgroup2 descendant test. 894 * * 1, if the *skb* succeeded the cgroup2 descendant test. 895 * * A negative error code, if an error occurred. 896 */ 897 static long (*bpf_skb_under_cgroup)(struct __sk_buff *skb, void *map, __u32 index) = (void *) 33; 898 899 /* 900 * bpf_get_hash_recalc 901 * 902 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 903 * not set, in particular if the hash was cleared due to mangling, 904 * recompute this hash. Later accesses to the hash can be done 905 * directly with *skb*\ **->hash**. 906 * 907 * Calling **bpf_set_hash_invalid**\ (), changing a packet 908 * prototype with **bpf_skb_change_proto**\ (), or calling 909 * **bpf_skb_store_bytes**\ () with the 910 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 911 * the hash and to trigger a new computation for the next call to 912 * **bpf_get_hash_recalc**\ (). 913 * 914 * Returns 915 * The 32-bit hash. 916 */ 917 static __u32 (*bpf_get_hash_recalc)(struct __sk_buff *skb) = (void *) 34; 918 919 /* 920 * bpf_get_current_task 921 * 922 * 923 * Returns 924 * A pointer to the current task struct. 925 */ 926 static __u64 (*bpf_get_current_task)(void) = (void *) 35; 927 928 /* 929 * bpf_probe_write_user 930 * 931 * Attempt in a safe way to write *len* bytes from the buffer 932 * *src* to *dst* in memory. It only works for threads that are in 933 * user context, and *dst* must be a valid user space address. 934 * 935 * This helper should not be used to implement any kind of 936 * security mechanism because of TOC-TOU attacks, but rather to 937 * debug, divert, and manipulate execution of semi-cooperative 938 * processes. 939 * 940 * Keep in mind that this feature is meant for experiments, and it 941 * has a risk of crashing the system and running programs. 942 * Therefore, when an eBPF program using this helper is attached, 943 * a warning including PID and process name is printed to kernel 944 * logs. 945 * 946 * Returns 947 * 0 on success, or a negative error in case of failure. 948 */ 949 static long (*bpf_probe_write_user)(void *dst, const void *src, __u32 len) = (void *) 36; 950 951 /* 952 * bpf_current_task_under_cgroup 953 * 954 * Check whether the probe is being run is the context of a given 955 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 956 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 957 * 958 * Returns 959 * The return value depends on the result of the test, and can be: 960 * 961 * * 0, if current task belongs to the cgroup2. 962 * * 1, if current task does not belong to the cgroup2. 963 * * A negative error code, if an error occurred. 964 */ 965 static long (*bpf_current_task_under_cgroup)(void *map, __u32 index) = (void *) 37; 966 967 /* 968 * bpf_skb_change_tail 969 * 970 * Resize (trim or grow) the packet associated to *skb* to the 971 * new *len*. The *flags* are reserved for future usage, and must 972 * be left at zero. 973 * 974 * The basic idea is that the helper performs the needed work to 975 * change the size of the packet, then the eBPF program rewrites 976 * the rest via helpers like **bpf_skb_store_bytes**\ (), 977 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 978 * and others. This helper is a slow path utility intended for 979 * replies with control messages. And because it is targeted for 980 * slow path, the helper itself can afford to be slow: it 981 * implicitly linearizes, unclones and drops offloads from the 982 * *skb*. 983 * 984 * A call to this helper is susceptible to change the underlying 985 * packet buffer. Therefore, at load time, all checks on pointers 986 * previously done by the verifier are invalidated and must be 987 * performed again, if the helper is used in combination with 988 * direct packet access. 989 * 990 * Returns 991 * 0 on success, or a negative error in case of failure. 992 */ 993 static long (*bpf_skb_change_tail)(struct __sk_buff *skb, __u32 len, __u64 flags) = (void *) 38; 994 995 /* 996 * bpf_skb_pull_data 997 * 998 * Pull in non-linear data in case the *skb* is non-linear and not 999 * all of *len* are part of the linear section. Make *len* bytes 1000 * from *skb* readable and writable. If a zero value is passed for 1001 * *len*, then the whole length of the *skb* is pulled. 1002 * 1003 * This helper is only needed for reading and writing with direct 1004 * packet access. 1005 * 1006 * For direct packet access, testing that offsets to access 1007 * are within packet boundaries (test on *skb*\ **->data_end**) is 1008 * susceptible to fail if offsets are invalid, or if the requested 1009 * data is in non-linear parts of the *skb*. On failure the 1010 * program can just bail out, or in the case of a non-linear 1011 * buffer, use a helper to make the data available. The 1012 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1013 * the data. Another one consists in using **bpf_skb_pull_data** 1014 * to pull in once the non-linear parts, then retesting and 1015 * eventually access the data. 1016 * 1017 * At the same time, this also makes sure the *skb* is uncloned, 1018 * which is a necessary condition for direct write. As this needs 1019 * to be an invariant for the write part only, the verifier 1020 * detects writes and adds a prologue that is calling 1021 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1022 * the very beginning in case it is indeed cloned. 1023 * 1024 * A call to this helper is susceptible to change the underlying 1025 * packet buffer. Therefore, at load time, all checks on pointers 1026 * previously done by the verifier are invalidated and must be 1027 * performed again, if the helper is used in combination with 1028 * direct packet access. 1029 * 1030 * Returns 1031 * 0 on success, or a negative error in case of failure. 1032 */ 1033 static long (*bpf_skb_pull_data)(struct __sk_buff *skb, __u32 len) = (void *) 39; 1034 1035 /* 1036 * bpf_csum_update 1037 * 1038 * Add the checksum *csum* into *skb*\ **->csum** in case the 1039 * driver has supplied a checksum for the entire packet into that 1040 * field. Return an error otherwise. This helper is intended to be 1041 * used in combination with **bpf_csum_diff**\ (), in particular 1042 * when the checksum needs to be updated after data has been 1043 * written into the packet through direct packet access. 1044 * 1045 * Returns 1046 * The checksum on success, or a negative error code in case of 1047 * failure. 1048 */ 1049 static __s64 (*bpf_csum_update)(struct __sk_buff *skb, __wsum csum) = (void *) 40; 1050 1051 /* 1052 * bpf_set_hash_invalid 1053 * 1054 * Invalidate the current *skb*\ **->hash**. It can be used after 1055 * mangling on headers through direct packet access, in order to 1056 * indicate that the hash is outdated and to trigger a 1057 * recalculation the next time the kernel tries to access this 1058 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1059 * 1060 */ 1061 static void (*bpf_set_hash_invalid)(struct __sk_buff *skb) = (void *) 41; 1062 1063 /* 1064 * bpf_get_numa_node_id 1065 * 1066 * Return the id of the current NUMA node. The primary use case 1067 * for this helper is the selection of sockets for the local NUMA 1068 * node, when the program is attached to sockets using the 1069 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1070 * but the helper is also available to other eBPF program types, 1071 * similarly to **bpf_get_smp_processor_id**\ (). 1072 * 1073 * Returns 1074 * The id of current NUMA node. 1075 */ 1076 static long (*bpf_get_numa_node_id)(void) = (void *) 42; 1077 1078 /* 1079 * bpf_skb_change_head 1080 * 1081 * Grows headroom of packet associated to *skb* and adjusts the 1082 * offset of the MAC header accordingly, adding *len* bytes of 1083 * space. It automatically extends and reallocates memory as 1084 * required. 1085 * 1086 * This helper can be used on a layer 3 *skb* to push a MAC header 1087 * for redirection into a layer 2 device. 1088 * 1089 * All values for *flags* are reserved for future usage, and must 1090 * be left at zero. 1091 * 1092 * A call to this helper is susceptible to change the underlying 1093 * packet buffer. Therefore, at load time, all checks on pointers 1094 * previously done by the verifier are invalidated and must be 1095 * performed again, if the helper is used in combination with 1096 * direct packet access. 1097 * 1098 * Returns 1099 * 0 on success, or a negative error in case of failure. 1100 */ 1101 static long (*bpf_skb_change_head)(struct __sk_buff *skb, __u32 len, __u64 flags) = (void *) 43; 1102 1103 /* 1104 * bpf_xdp_adjust_head 1105 * 1106 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1107 * it is possible to use a negative value for *delta*. This helper 1108 * can be used to prepare the packet for pushing or popping 1109 * headers. 1110 * 1111 * A call to this helper is susceptible to change the underlying 1112 * packet buffer. Therefore, at load time, all checks on pointers 1113 * previously done by the verifier are invalidated and must be 1114 * performed again, if the helper is used in combination with 1115 * direct packet access. 1116 * 1117 * Returns 1118 * 0 on success, or a negative error in case of failure. 1119 */ 1120 static long (*bpf_xdp_adjust_head)(struct xdp_md *xdp_md, int delta) = (void *) 44; 1121 1122 /* 1123 * bpf_probe_read_str 1124 * 1125 * Copy a NUL terminated string from an unsafe kernel address 1126 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 1127 * more details. 1128 * 1129 * Generally, use **bpf_probe_read_user_str**\ () or 1130 * **bpf_probe_read_kernel_str**\ () instead. 1131 * 1132 * Returns 1133 * On success, the strictly positive length of the string, 1134 * including the trailing NUL character. On error, a negative 1135 * value. 1136 */ 1137 static long (*bpf_probe_read_str)(void *dst, __u32 size, const void *unsafe_ptr) = (void *) 45; 1138 1139 /* 1140 * bpf_get_socket_cookie 1141 * 1142 * If the **struct sk_buff** pointed by *skb* has a known socket, 1143 * retrieve the cookie (generated by the kernel) of this socket. 1144 * If no cookie has been set yet, generate a new cookie. Once 1145 * generated, the socket cookie remains stable for the life of the 1146 * socket. This helper can be useful for monitoring per socket 1147 * networking traffic statistics as it provides a global socket 1148 * identifier that can be assumed unique. 1149 * 1150 * Returns 1151 * A 8-byte long unique number on success, or 0 if the socket 1152 * field is missing inside *skb*. 1153 */ 1154 static __u64 (*bpf_get_socket_cookie)(void *ctx) = (void *) 46; 1155 1156 /* 1157 * bpf_get_socket_uid 1158 * 1159 * 1160 * Returns 1161 * The owner UID of the socket associated to *skb*. If the socket 1162 * is **NULL**, or if it is not a full socket (i.e. if it is a 1163 * time-wait or a request socket instead), **overflowuid** value 1164 * is returned (note that **overflowuid** might also be the actual 1165 * UID value for the socket). 1166 */ 1167 static __u32 (*bpf_get_socket_uid)(struct __sk_buff *skb) = (void *) 47; 1168 1169 /* 1170 * bpf_set_hash 1171 * 1172 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1173 * to value *hash*. 1174 * 1175 * Returns 1176 * 0 1177 */ 1178 static long (*bpf_set_hash)(struct __sk_buff *skb, __u32 hash) = (void *) 48; 1179 1180 /* 1181 * bpf_setsockopt 1182 * 1183 * Emulate a call to **setsockopt()** on the socket associated to 1184 * *bpf_socket*, which must be a full socket. The *level* at 1185 * which the option resides and the name *optname* of the option 1186 * must be specified, see **setsockopt(2)** for more information. 1187 * The option value of length *optlen* is pointed by *optval*. 1188 * 1189 * *bpf_socket* should be one of the following: 1190 * 1191 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1192 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 1193 * and **BPF_CGROUP_INET6_CONNECT**. 1194 * 1195 * This helper actually implements a subset of **setsockopt()**. 1196 * It supports the following *level*\ s: 1197 * 1198 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1199 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1200 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 1201 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 1202 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1203 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1204 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 1205 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 1206 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**. 1207 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1208 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1209 * 1210 * Returns 1211 * 0 on success, or a negative error in case of failure. 1212 */ 1213 static long (*bpf_setsockopt)(void *bpf_socket, int level, int optname, void *optval, int optlen) = (void *) 49; 1214 1215 /* 1216 * bpf_skb_adjust_room 1217 * 1218 * Grow or shrink the room for data in the packet associated to 1219 * *skb* by *len_diff*, and according to the selected *mode*. 1220 * 1221 * By default, the helper will reset any offloaded checksum 1222 * indicator of the skb to CHECKSUM_NONE. This can be avoided 1223 * by the following flag: 1224 * 1225 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 1226 * checksum data of the skb to CHECKSUM_NONE. 1227 * 1228 * There are two supported modes at this time: 1229 * 1230 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1231 * (room space is added or removed below the layer 2 header). 1232 * 1233 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1234 * (room space is added or removed below the layer 3 header). 1235 * 1236 * The following flags are supported at this time: 1237 * 1238 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1239 * Adjusting mss in this way is not allowed for datagrams. 1240 * 1241 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 1242 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1243 * Any new space is reserved to hold a tunnel header. 1244 * Configure skb offsets and other fields accordingly. 1245 * 1246 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 1247 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1248 * Use with ENCAP_L3 flags to further specify the tunnel type. 1249 * 1250 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 1251 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1252 * type; *len* is the length of the inner MAC header. 1253 * 1254 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 1255 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 1256 * L2 type as Ethernet. 1257 * 1258 * A call to this helper is susceptible to change the underlying 1259 * packet buffer. Therefore, at load time, all checks on pointers 1260 * previously done by the verifier are invalidated and must be 1261 * performed again, if the helper is used in combination with 1262 * direct packet access. 1263 * 1264 * Returns 1265 * 0 on success, or a negative error in case of failure. 1266 */ 1267 static long (*bpf_skb_adjust_room)(struct __sk_buff *skb, __s32 len_diff, __u32 mode, __u64 flags) = (void *) 50; 1268 1269 /* 1270 * bpf_redirect_map 1271 * 1272 * Redirect the packet to the endpoint referenced by *map* at 1273 * index *key*. Depending on its type, this *map* can contain 1274 * references to net devices (for forwarding packets through other 1275 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1276 * but this is only implemented for native XDP (with driver 1277 * support) as of this writing). 1278 * 1279 * The lower two bits of *flags* are used as the return code if 1280 * the map lookup fails. This is so that the return value can be 1281 * one of the XDP program return codes up to **XDP_TX**, as chosen 1282 * by the caller. The higher bits of *flags* can be set to 1283 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 1284 * 1285 * With BPF_F_BROADCAST the packet will be broadcasted to all the 1286 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 1287 * interface will be excluded when do broadcasting. 1288 * 1289 * See also **bpf_redirect**\ (), which only supports redirecting 1290 * to an ifindex, but doesn't require a map to do so. 1291 * 1292 * Returns 1293 * **XDP_REDIRECT** on success, or the value of the two lower bits 1294 * of the *flags* argument on error. 1295 */ 1296 static long (*bpf_redirect_map)(void *map, __u32 key, __u64 flags) = (void *) 51; 1297 1298 /* 1299 * bpf_sk_redirect_map 1300 * 1301 * Redirect the packet to the socket referenced by *map* (of type 1302 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1303 * egress interfaces can be used for redirection. The 1304 * **BPF_F_INGRESS** value in *flags* is used to make the 1305 * distinction (ingress path is selected if the flag is present, 1306 * egress path otherwise). This is the only flag supported for now. 1307 * 1308 * Returns 1309 * **SK_PASS** on success, or **SK_DROP** on error. 1310 */ 1311 static long (*bpf_sk_redirect_map)(struct __sk_buff *skb, void *map, __u32 key, __u64 flags) = (void *) 52; 1312 1313 /* 1314 * bpf_sock_map_update 1315 * 1316 * Add an entry to, or update a *map* referencing sockets. The 1317 * *skops* is used as a new value for the entry associated to 1318 * *key*. *flags* is one of: 1319 * 1320 * **BPF_NOEXIST** 1321 * The entry for *key* must not exist in the map. 1322 * **BPF_EXIST** 1323 * The entry for *key* must already exist in the map. 1324 * **BPF_ANY** 1325 * No condition on the existence of the entry for *key*. 1326 * 1327 * If the *map* has eBPF programs (parser and verdict), those will 1328 * be inherited by the socket being added. If the socket is 1329 * already attached to eBPF programs, this results in an error. 1330 * 1331 * Returns 1332 * 0 on success, or a negative error in case of failure. 1333 */ 1334 static long (*bpf_sock_map_update)(struct bpf_sock_ops *skops, void *map, void *key, __u64 flags) = (void *) 53; 1335 1336 /* 1337 * bpf_xdp_adjust_meta 1338 * 1339 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1340 * *delta* (which can be positive or negative). Note that this 1341 * operation modifies the address stored in *xdp_md*\ **->data**, 1342 * so the latter must be loaded only after the helper has been 1343 * called. 1344 * 1345 * The use of *xdp_md*\ **->data_meta** is optional and programs 1346 * are not required to use it. The rationale is that when the 1347 * packet is processed with XDP (e.g. as DoS filter), it is 1348 * possible to push further meta data along with it before passing 1349 * to the stack, and to give the guarantee that an ingress eBPF 1350 * program attached as a TC classifier on the same device can pick 1351 * this up for further post-processing. Since TC works with socket 1352 * buffers, it remains possible to set from XDP the **mark** or 1353 * **priority** pointers, or other pointers for the socket buffer. 1354 * Having this scratch space generic and programmable allows for 1355 * more flexibility as the user is free to store whatever meta 1356 * data they need. 1357 * 1358 * A call to this helper is susceptible to change the underlying 1359 * packet buffer. Therefore, at load time, all checks on pointers 1360 * previously done by the verifier are invalidated and must be 1361 * performed again, if the helper is used in combination with 1362 * direct packet access. 1363 * 1364 * Returns 1365 * 0 on success, or a negative error in case of failure. 1366 */ 1367 static long (*bpf_xdp_adjust_meta)(struct xdp_md *xdp_md, int delta) = (void *) 54; 1368 1369 /* 1370 * bpf_perf_event_read_value 1371 * 1372 * Read the value of a perf event counter, and store it into *buf* 1373 * of size *buf_size*. This helper relies on a *map* of type 1374 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1375 * counter is selected when *map* is updated with perf event file 1376 * descriptors. The *map* is an array whose size is the number of 1377 * available CPUs, and each cell contains a value relative to one 1378 * CPU. The value to retrieve is indicated by *flags*, that 1379 * contains the index of the CPU to look up, masked with 1380 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1381 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1382 * current CPU should be retrieved. 1383 * 1384 * This helper behaves in a way close to 1385 * **bpf_perf_event_read**\ () helper, save that instead of 1386 * just returning the value observed, it fills the *buf* 1387 * structure. This allows for additional data to be retrieved: in 1388 * particular, the enabled and running times (in *buf*\ 1389 * **->enabled** and *buf*\ **->running**, respectively) are 1390 * copied. In general, **bpf_perf_event_read_value**\ () is 1391 * recommended over **bpf_perf_event_read**\ (), which has some 1392 * ABI issues and provides fewer functionalities. 1393 * 1394 * These values are interesting, because hardware PMU (Performance 1395 * Monitoring Unit) counters are limited resources. When there are 1396 * more PMU based perf events opened than available counters, 1397 * kernel will multiplex these events so each event gets certain 1398 * percentage (but not all) of the PMU time. In case that 1399 * multiplexing happens, the number of samples or counter value 1400 * will not reflect the case compared to when no multiplexing 1401 * occurs. This makes comparison between different runs difficult. 1402 * Typically, the counter value should be normalized before 1403 * comparing to other experiments. The usual normalization is done 1404 * as follows. 1405 * 1406 * :: 1407 * 1408 * normalized_counter = counter * t_enabled / t_running 1409 * 1410 * Where t_enabled is the time enabled for event and t_running is 1411 * the time running for event since last normalization. The 1412 * enabled and running times are accumulated since the perf event 1413 * open. To achieve scaling factor between two invocations of an 1414 * eBPF program, users can use CPU id as the key (which is 1415 * typical for perf array usage model) to remember the previous 1416 * value and do the calculation inside the eBPF program. 1417 * 1418 * Returns 1419 * 0 on success, or a negative error in case of failure. 1420 */ 1421 static long (*bpf_perf_event_read_value)(void *map, __u64 flags, struct bpf_perf_event_value *buf, __u32 buf_size) = (void *) 55; 1422 1423 /* 1424 * bpf_perf_prog_read_value 1425 * 1426 * For en eBPF program attached to a perf event, retrieve the 1427 * value of the event counter associated to *ctx* and store it in 1428 * the structure pointed by *buf* and of size *buf_size*. Enabled 1429 * and running times are also stored in the structure (see 1430 * description of helper **bpf_perf_event_read_value**\ () for 1431 * more details). 1432 * 1433 * Returns 1434 * 0 on success, or a negative error in case of failure. 1435 */ 1436 static long (*bpf_perf_prog_read_value)(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, __u32 buf_size) = (void *) 56; 1437 1438 /* 1439 * bpf_getsockopt 1440 * 1441 * Emulate a call to **getsockopt()** on the socket associated to 1442 * *bpf_socket*, which must be a full socket. The *level* at 1443 * which the option resides and the name *optname* of the option 1444 * must be specified, see **getsockopt(2)** for more information. 1445 * The retrieved value is stored in the structure pointed by 1446 * *opval* and of length *optlen*. 1447 * 1448 * *bpf_socket* should be one of the following: 1449 * 1450 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1451 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 1452 * and **BPF_CGROUP_INET6_CONNECT**. 1453 * 1454 * This helper actually implements a subset of **getsockopt()**. 1455 * It supports the following *level*\ s: 1456 * 1457 * * **IPPROTO_TCP**, which supports *optname* 1458 * **TCP_CONGESTION**. 1459 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1460 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 1461 * 1462 * Returns 1463 * 0 on success, or a negative error in case of failure. 1464 */ 1465 static long (*bpf_getsockopt)(void *bpf_socket, int level, int optname, void *optval, int optlen) = (void *) 57; 1466 1467 /* 1468 * bpf_override_return 1469 * 1470 * Used for error injection, this helper uses kprobes to override 1471 * the return value of the probed function, and to set it to *rc*. 1472 * The first argument is the context *regs* on which the kprobe 1473 * works. 1474 * 1475 * This helper works by setting the PC (program counter) 1476 * to an override function which is run in place of the original 1477 * probed function. This means the probed function is not run at 1478 * all. The replacement function just returns with the required 1479 * value. 1480 * 1481 * This helper has security implications, and thus is subject to 1482 * restrictions. It is only available if the kernel was compiled 1483 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1484 * option, and in this case it only works on functions tagged with 1485 * **ALLOW_ERROR_INJECTION** in the kernel code. 1486 * 1487 * Also, the helper is only available for the architectures having 1488 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1489 * x86 architecture is the only one to support this feature. 1490 * 1491 * Returns 1492 * 0 1493 */ 1494 static long (*bpf_override_return)(struct pt_regs *regs, __u64 rc) = (void *) 58; 1495 1496 /* 1497 * bpf_sock_ops_cb_flags_set 1498 * 1499 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1500 * for the full TCP socket associated to *bpf_sock_ops* to 1501 * *argval*. 1502 * 1503 * The primary use of this field is to determine if there should 1504 * be calls to eBPF programs of type 1505 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1506 * code. A program of the same type can change its value, per 1507 * connection and as necessary, when the connection is 1508 * established. This field is directly accessible for reading, but 1509 * this helper must be used for updates in order to return an 1510 * error if an eBPF program tries to set a callback that is not 1511 * supported in the current kernel. 1512 * 1513 * *argval* is a flag array which can combine these flags: 1514 * 1515 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1516 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1517 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1518 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 1519 * 1520 * Therefore, this function can be used to clear a callback flag by 1521 * setting the appropriate bit to zero. e.g. to disable the RTO 1522 * callback: 1523 * 1524 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1525 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1526 * 1527 * Here are some examples of where one could call such eBPF 1528 * program: 1529 * 1530 * * When RTO fires. 1531 * * When a packet is retransmitted. 1532 * * When the connection terminates. 1533 * * When a packet is sent. 1534 * * When a packet is received. 1535 * 1536 * Returns 1537 * Code **-EINVAL** if the socket is not a full TCP socket; 1538 * otherwise, a positive number containing the bits that could not 1539 * be set is returned (which comes down to 0 if all bits were set 1540 * as required). 1541 */ 1542 static long (*bpf_sock_ops_cb_flags_set)(struct bpf_sock_ops *bpf_sock, int argval) = (void *) 59; 1543 1544 /* 1545 * bpf_msg_redirect_map 1546 * 1547 * This helper is used in programs implementing policies at the 1548 * socket level. If the message *msg* is allowed to pass (i.e. if 1549 * the verdict eBPF program returns **SK_PASS**), redirect it to 1550 * the socket referenced by *map* (of type 1551 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1552 * egress interfaces can be used for redirection. The 1553 * **BPF_F_INGRESS** value in *flags* is used to make the 1554 * distinction (ingress path is selected if the flag is present, 1555 * egress path otherwise). This is the only flag supported for now. 1556 * 1557 * Returns 1558 * **SK_PASS** on success, or **SK_DROP** on error. 1559 */ 1560 static long (*bpf_msg_redirect_map)(struct sk_msg_md *msg, void *map, __u32 key, __u64 flags) = (void *) 60; 1561 1562 /* 1563 * bpf_msg_apply_bytes 1564 * 1565 * For socket policies, apply the verdict of the eBPF program to 1566 * the next *bytes* (number of bytes) of message *msg*. 1567 * 1568 * For example, this helper can be used in the following cases: 1569 * 1570 * * A single **sendmsg**\ () or **sendfile**\ () system call 1571 * contains multiple logical messages that the eBPF program is 1572 * supposed to read and for which it should apply a verdict. 1573 * * An eBPF program only cares to read the first *bytes* of a 1574 * *msg*. If the message has a large payload, then setting up 1575 * and calling the eBPF program repeatedly for all bytes, even 1576 * though the verdict is already known, would create unnecessary 1577 * overhead. 1578 * 1579 * When called from within an eBPF program, the helper sets a 1580 * counter internal to the BPF infrastructure, that is used to 1581 * apply the last verdict to the next *bytes*. If *bytes* is 1582 * smaller than the current data being processed from a 1583 * **sendmsg**\ () or **sendfile**\ () system call, the first 1584 * *bytes* will be sent and the eBPF program will be re-run with 1585 * the pointer for start of data pointing to byte number *bytes* 1586 * **+ 1**. If *bytes* is larger than the current data being 1587 * processed, then the eBPF verdict will be applied to multiple 1588 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1589 * consumed. 1590 * 1591 * Note that if a socket closes with the internal counter holding 1592 * a non-zero value, this is not a problem because data is not 1593 * being buffered for *bytes* and is sent as it is received. 1594 * 1595 * Returns 1596 * 0 1597 */ 1598 static long (*bpf_msg_apply_bytes)(struct sk_msg_md *msg, __u32 bytes) = (void *) 61; 1599 1600 /* 1601 * bpf_msg_cork_bytes 1602 * 1603 * For socket policies, prevent the execution of the verdict eBPF 1604 * program for message *msg* until *bytes* (byte number) have been 1605 * accumulated. 1606 * 1607 * This can be used when one needs a specific number of bytes 1608 * before a verdict can be assigned, even if the data spans 1609 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1610 * case would be a user calling **sendmsg**\ () repeatedly with 1611 * 1-byte long message segments. Obviously, this is bad for 1612 * performance, but it is still valid. If the eBPF program needs 1613 * *bytes* bytes to validate a header, this helper can be used to 1614 * prevent the eBPF program to be called again until *bytes* have 1615 * been accumulated. 1616 * 1617 * Returns 1618 * 0 1619 */ 1620 static long (*bpf_msg_cork_bytes)(struct sk_msg_md *msg, __u32 bytes) = (void *) 62; 1621 1622 /* 1623 * bpf_msg_pull_data 1624 * 1625 * For socket policies, pull in non-linear data from user space 1626 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1627 * **->data_end** to *start* and *end* bytes offsets into *msg*, 1628 * respectively. 1629 * 1630 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1631 * *msg* it can only parse data that the (**data**, **data_end**) 1632 * pointers have already consumed. For **sendmsg**\ () hooks this 1633 * is likely the first scatterlist element. But for calls relying 1634 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1635 * be the range (**0**, **0**) because the data is shared with 1636 * user space and by default the objective is to avoid allowing 1637 * user space to modify data while (or after) eBPF verdict is 1638 * being decided. This helper can be used to pull in data and to 1639 * set the start and end pointer to given values. Data will be 1640 * copied if necessary (i.e. if data was not linear and if start 1641 * and end pointers do not point to the same chunk). 1642 * 1643 * A call to this helper is susceptible to change the underlying 1644 * packet buffer. Therefore, at load time, all checks on pointers 1645 * previously done by the verifier are invalidated and must be 1646 * performed again, if the helper is used in combination with 1647 * direct packet access. 1648 * 1649 * All values for *flags* are reserved for future usage, and must 1650 * be left at zero. 1651 * 1652 * Returns 1653 * 0 on success, or a negative error in case of failure. 1654 */ 1655 static long (*bpf_msg_pull_data)(struct sk_msg_md *msg, __u32 start, __u32 end, __u64 flags) = (void *) 63; 1656 1657 /* 1658 * bpf_bind 1659 * 1660 * Bind the socket associated to *ctx* to the address pointed by 1661 * *addr*, of length *addr_len*. This allows for making outgoing 1662 * connection from the desired IP address, which can be useful for 1663 * example when all processes inside a cgroup should use one 1664 * single IP address on a host that has multiple IP configured. 1665 * 1666 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 1667 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 1668 * **AF_INET6**). It's advised to pass zero port (**sin_port** 1669 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 1670 * behavior and lets the kernel efficiently pick up an unused 1671 * port as long as 4-tuple is unique. Passing non-zero port might 1672 * lead to degraded performance. 1673 * 1674 * Returns 1675 * 0 on success, or a negative error in case of failure. 1676 */ 1677 static long (*bpf_bind)(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) = (void *) 64; 1678 1679 /* 1680 * bpf_xdp_adjust_tail 1681 * 1682 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 1683 * possible to both shrink and grow the packet tail. 1684 * Shrink done via *delta* being a negative integer. 1685 * 1686 * A call to this helper is susceptible to change the underlying 1687 * packet buffer. Therefore, at load time, all checks on pointers 1688 * previously done by the verifier are invalidated and must be 1689 * performed again, if the helper is used in combination with 1690 * direct packet access. 1691 * 1692 * Returns 1693 * 0 on success, or a negative error in case of failure. 1694 */ 1695 static long (*bpf_xdp_adjust_tail)(struct xdp_md *xdp_md, int delta) = (void *) 65; 1696 1697 /* 1698 * bpf_skb_get_xfrm_state 1699 * 1700 * Retrieve the XFRM state (IP transform framework, see also 1701 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 1702 * 1703 * The retrieved value is stored in the **struct bpf_xfrm_state** 1704 * pointed by *xfrm_state* and of length *size*. 1705 * 1706 * All values for *flags* are reserved for future usage, and must 1707 * be left at zero. 1708 * 1709 * This helper is available only if the kernel was compiled with 1710 * **CONFIG_XFRM** configuration option. 1711 * 1712 * Returns 1713 * 0 on success, or a negative error in case of failure. 1714 */ 1715 static long (*bpf_skb_get_xfrm_state)(struct __sk_buff *skb, __u32 index, struct bpf_xfrm_state *xfrm_state, __u32 size, __u64 flags) = (void *) 66; 1716 1717 /* 1718 * bpf_get_stack 1719 * 1720 * Return a user or a kernel stack in bpf program provided buffer. 1721 * To achieve this, the helper needs *ctx*, which is a pointer 1722 * to the context on which the tracing program is executed. 1723 * To store the stacktrace, the bpf program provides *buf* with 1724 * a nonnegative *size*. 1725 * 1726 * The last argument, *flags*, holds the number of stack frames to 1727 * skip (from 0 to 255), masked with 1728 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1729 * the following flags: 1730 * 1731 * **BPF_F_USER_STACK** 1732 * Collect a user space stack instead of a kernel stack. 1733 * **BPF_F_USER_BUILD_ID** 1734 * Collect buildid+offset instead of ips for user stack, 1735 * only valid if **BPF_F_USER_STACK** is also specified. 1736 * 1737 * **bpf_get_stack**\ () can collect up to 1738 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 1739 * to sufficient large buffer size. Note that 1740 * this limit can be controlled with the **sysctl** program, and 1741 * that it should be manually increased in order to profile long 1742 * user stacks (such as stacks for Java programs). To do so, use: 1743 * 1744 * :: 1745 * 1746 * # sysctl kernel.perf_event_max_stack=<new value> 1747 * 1748 * Returns 1749 * A non-negative value equal to or less than *size* on success, 1750 * or a negative error in case of failure. 1751 */ 1752 static long (*bpf_get_stack)(void *ctx, void *buf, __u32 size, __u64 flags) = (void *) 67; 1753 1754 /* 1755 * bpf_skb_load_bytes_relative 1756 * 1757 * This helper is similar to **bpf_skb_load_bytes**\ () in that 1758 * it provides an easy way to load *len* bytes from *offset* 1759 * from the packet associated to *skb*, into the buffer pointed 1760 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 1761 * a fifth argument *start_header* exists in order to select a 1762 * base offset to start from. *start_header* can be one of: 1763 * 1764 * **BPF_HDR_START_MAC** 1765 * Base offset to load data from is *skb*'s mac header. 1766 * **BPF_HDR_START_NET** 1767 * Base offset to load data from is *skb*'s network header. 1768 * 1769 * In general, "direct packet access" is the preferred method to 1770 * access packet data, however, this helper is in particular useful 1771 * in socket filters where *skb*\ **->data** does not always point 1772 * to the start of the mac header and where "direct packet access" 1773 * is not available. 1774 * 1775 * Returns 1776 * 0 on success, or a negative error in case of failure. 1777 */ 1778 static long (*bpf_skb_load_bytes_relative)(const void *skb, __u32 offset, void *to, __u32 len, __u32 start_header) = (void *) 68; 1779 1780 /* 1781 * bpf_fib_lookup 1782 * 1783 * Do FIB lookup in kernel tables using parameters in *params*. 1784 * If lookup is successful and result shows packet is to be 1785 * forwarded, the neighbor tables are searched for the nexthop. 1786 * If successful (ie., FIB lookup shows forwarding and nexthop 1787 * is resolved), the nexthop address is returned in ipv4_dst 1788 * or ipv6_dst based on family, smac is set to mac address of 1789 * egress device, dmac is set to nexthop mac address, rt_metric 1790 * is set to metric from route (IPv4/IPv6 only), and ifindex 1791 * is set to the device index of the nexthop from the FIB lookup. 1792 * 1793 * *plen* argument is the size of the passed in struct. 1794 * *flags* argument can be a combination of one or more of the 1795 * following values: 1796 * 1797 * **BPF_FIB_LOOKUP_DIRECT** 1798 * Do a direct table lookup vs full lookup using FIB 1799 * rules. 1800 * **BPF_FIB_LOOKUP_OUTPUT** 1801 * Perform lookup from an egress perspective (default is 1802 * ingress). 1803 * 1804 * *ctx* is either **struct xdp_md** for XDP programs or 1805 * **struct sk_buff** tc cls_act programs. 1806 * 1807 * Returns 1808 * * < 0 if any input argument is invalid 1809 * * 0 on success (packet is forwarded, nexthop neighbor exists) 1810 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 1811 * packet is not forwarded or needs assist from full stack 1812 * 1813 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 1814 * was exceeded and output params->mtu_result contains the MTU. 1815 */ 1816 static long (*bpf_fib_lookup)(void *ctx, struct bpf_fib_lookup *params, int plen, __u32 flags) = (void *) 69; 1817 1818 /* 1819 * bpf_sock_hash_update 1820 * 1821 * Add an entry to, or update a sockhash *map* referencing sockets. 1822 * The *skops* is used as a new value for the entry associated to 1823 * *key*. *flags* is one of: 1824 * 1825 * **BPF_NOEXIST** 1826 * The entry for *key* must not exist in the map. 1827 * **BPF_EXIST** 1828 * The entry for *key* must already exist in the map. 1829 * **BPF_ANY** 1830 * No condition on the existence of the entry for *key*. 1831 * 1832 * If the *map* has eBPF programs (parser and verdict), those will 1833 * be inherited by the socket being added. If the socket is 1834 * already attached to eBPF programs, this results in an error. 1835 * 1836 * Returns 1837 * 0 on success, or a negative error in case of failure. 1838 */ 1839 static long (*bpf_sock_hash_update)(struct bpf_sock_ops *skops, void *map, void *key, __u64 flags) = (void *) 70; 1840 1841 /* 1842 * bpf_msg_redirect_hash 1843 * 1844 * This helper is used in programs implementing policies at the 1845 * socket level. If the message *msg* is allowed to pass (i.e. if 1846 * the verdict eBPF program returns **SK_PASS**), redirect it to 1847 * the socket referenced by *map* (of type 1848 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 1849 * egress interfaces can be used for redirection. The 1850 * **BPF_F_INGRESS** value in *flags* is used to make the 1851 * distinction (ingress path is selected if the flag is present, 1852 * egress path otherwise). This is the only flag supported for now. 1853 * 1854 * Returns 1855 * **SK_PASS** on success, or **SK_DROP** on error. 1856 */ 1857 static long (*bpf_msg_redirect_hash)(struct sk_msg_md *msg, void *map, void *key, __u64 flags) = (void *) 71; 1858 1859 /* 1860 * bpf_sk_redirect_hash 1861 * 1862 * This helper is used in programs implementing policies at the 1863 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 1864 * if the verdict eBPF program returns **SK_PASS**), redirect it 1865 * to the socket referenced by *map* (of type 1866 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 1867 * egress interfaces can be used for redirection. The 1868 * **BPF_F_INGRESS** value in *flags* is used to make the 1869 * distinction (ingress path is selected if the flag is present, 1870 * egress otherwise). This is the only flag supported for now. 1871 * 1872 * Returns 1873 * **SK_PASS** on success, or **SK_DROP** on error. 1874 */ 1875 static long (*bpf_sk_redirect_hash)(struct __sk_buff *skb, void *map, void *key, __u64 flags) = (void *) 72; 1876 1877 /* 1878 * bpf_lwt_push_encap 1879 * 1880 * Encapsulate the packet associated to *skb* within a Layer 3 1881 * protocol header. This header is provided in the buffer at 1882 * address *hdr*, with *len* its size in bytes. *type* indicates 1883 * the protocol of the header and can be one of: 1884 * 1885 * **BPF_LWT_ENCAP_SEG6** 1886 * IPv6 encapsulation with Segment Routing Header 1887 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 1888 * the IPv6 header is computed by the kernel. 1889 * **BPF_LWT_ENCAP_SEG6_INLINE** 1890 * Only works if *skb* contains an IPv6 packet. Insert a 1891 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 1892 * the IPv6 header. 1893 * **BPF_LWT_ENCAP_IP** 1894 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 1895 * must be IPv4 or IPv6, followed by zero or more 1896 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 1897 * total bytes in all prepended headers. Please note that 1898 * if **skb_is_gso**\ (*skb*) is true, no more than two 1899 * headers can be prepended, and the inner header, if 1900 * present, should be either GRE or UDP/GUE. 1901 * 1902 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 1903 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 1904 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 1905 * **BPF_PROG_TYPE_LWT_XMIT**. 1906 * 1907 * A call to this helper is susceptible to change the underlying 1908 * packet buffer. Therefore, at load time, all checks on pointers 1909 * previously done by the verifier are invalidated and must be 1910 * performed again, if the helper is used in combination with 1911 * direct packet access. 1912 * 1913 * Returns 1914 * 0 on success, or a negative error in case of failure. 1915 */ 1916 static long (*bpf_lwt_push_encap)(struct __sk_buff *skb, __u32 type, void *hdr, __u32 len) = (void *) 73; 1917 1918 /* 1919 * bpf_lwt_seg6_store_bytes 1920 * 1921 * Store *len* bytes from address *from* into the packet 1922 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 1923 * inside the outermost IPv6 Segment Routing Header can be 1924 * modified through this helper. 1925 * 1926 * A call to this helper is susceptible to change the underlying 1927 * packet buffer. Therefore, at load time, all checks on pointers 1928 * previously done by the verifier are invalidated and must be 1929 * performed again, if the helper is used in combination with 1930 * direct packet access. 1931 * 1932 * Returns 1933 * 0 on success, or a negative error in case of failure. 1934 */ 1935 static long (*bpf_lwt_seg6_store_bytes)(struct __sk_buff *skb, __u32 offset, const void *from, __u32 len) = (void *) 74; 1936 1937 /* 1938 * bpf_lwt_seg6_adjust_srh 1939 * 1940 * Adjust the size allocated to TLVs in the outermost IPv6 1941 * Segment Routing Header contained in the packet associated to 1942 * *skb*, at position *offset* by *delta* bytes. Only offsets 1943 * after the segments are accepted. *delta* can be as well 1944 * positive (growing) as negative (shrinking). 1945 * 1946 * A call to this helper is susceptible to change the underlying 1947 * packet buffer. Therefore, at load time, all checks on pointers 1948 * previously done by the verifier are invalidated and must be 1949 * performed again, if the helper is used in combination with 1950 * direct packet access. 1951 * 1952 * Returns 1953 * 0 on success, or a negative error in case of failure. 1954 */ 1955 static long (*bpf_lwt_seg6_adjust_srh)(struct __sk_buff *skb, __u32 offset, __s32 delta) = (void *) 75; 1956 1957 /* 1958 * bpf_lwt_seg6_action 1959 * 1960 * Apply an IPv6 Segment Routing action of type *action* to the 1961 * packet associated to *skb*. Each action takes a parameter 1962 * contained at address *param*, and of length *param_len* bytes. 1963 * *action* can be one of: 1964 * 1965 * **SEG6_LOCAL_ACTION_END_X** 1966 * End.X action: Endpoint with Layer-3 cross-connect. 1967 * Type of *param*: **struct in6_addr**. 1968 * **SEG6_LOCAL_ACTION_END_T** 1969 * End.T action: Endpoint with specific IPv6 table lookup. 1970 * Type of *param*: **int**. 1971 * **SEG6_LOCAL_ACTION_END_B6** 1972 * End.B6 action: Endpoint bound to an SRv6 policy. 1973 * Type of *param*: **struct ipv6_sr_hdr**. 1974 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 1975 * End.B6.Encap action: Endpoint bound to an SRv6 1976 * encapsulation policy. 1977 * Type of *param*: **struct ipv6_sr_hdr**. 1978 * 1979 * A call to this helper is susceptible to change the underlying 1980 * packet buffer. Therefore, at load time, all checks on pointers 1981 * previously done by the verifier are invalidated and must be 1982 * performed again, if the helper is used in combination with 1983 * direct packet access. 1984 * 1985 * Returns 1986 * 0 on success, or a negative error in case of failure. 1987 */ 1988 static long (*bpf_lwt_seg6_action)(struct __sk_buff *skb, __u32 action, void *param, __u32 param_len) = (void *) 76; 1989 1990 /* 1991 * bpf_rc_repeat 1992 * 1993 * This helper is used in programs implementing IR decoding, to 1994 * report a successfully decoded repeat key message. This delays 1995 * the generation of a key up event for previously generated 1996 * key down event. 1997 * 1998 * Some IR protocols like NEC have a special IR message for 1999 * repeating last button, for when a button is held down. 2000 * 2001 * The *ctx* should point to the lirc sample as passed into 2002 * the program. 2003 * 2004 * This helper is only available is the kernel was compiled with 2005 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2006 * "**y**". 2007 * 2008 * Returns 2009 * 0 2010 */ 2011 static long (*bpf_rc_repeat)(void *ctx) = (void *) 77; 2012 2013 /* 2014 * bpf_rc_keydown 2015 * 2016 * This helper is used in programs implementing IR decoding, to 2017 * report a successfully decoded key press with *scancode*, 2018 * *toggle* value in the given *protocol*. The scancode will be 2019 * translated to a keycode using the rc keymap, and reported as 2020 * an input key down event. After a period a key up event is 2021 * generated. This period can be extended by calling either 2022 * **bpf_rc_keydown**\ () again with the same values, or calling 2023 * **bpf_rc_repeat**\ (). 2024 * 2025 * Some protocols include a toggle bit, in case the button was 2026 * released and pressed again between consecutive scancodes. 2027 * 2028 * The *ctx* should point to the lirc sample as passed into 2029 * the program. 2030 * 2031 * The *protocol* is the decoded protocol number (see 2032 * **enum rc_proto** for some predefined values). 2033 * 2034 * This helper is only available is the kernel was compiled with 2035 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2036 * "**y**". 2037 * 2038 * Returns 2039 * 0 2040 */ 2041 static long (*bpf_rc_keydown)(void *ctx, __u32 protocol, __u64 scancode, __u32 toggle) = (void *) 78; 2042 2043 /* 2044 * bpf_skb_cgroup_id 2045 * 2046 * Return the cgroup v2 id of the socket associated with the *skb*. 2047 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2048 * helper for cgroup v1 by providing a tag resp. identifier that 2049 * can be matched on or used for map lookups e.g. to implement 2050 * policy. The cgroup v2 id of a given path in the hierarchy is 2051 * exposed in user space through the f_handle API in order to get 2052 * to the same 64-bit id. 2053 * 2054 * This helper can be used on TC egress path, but not on ingress, 2055 * and is available only if the kernel was compiled with the 2056 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2057 * 2058 * Returns 2059 * The id is returned or 0 in case the id could not be retrieved. 2060 */ 2061 static __u64 (*bpf_skb_cgroup_id)(struct __sk_buff *skb) = (void *) 79; 2062 2063 /* 2064 * bpf_get_current_cgroup_id 2065 * 2066 * 2067 * Returns 2068 * A 64-bit integer containing the current cgroup id based 2069 * on the cgroup within which the current task is running. 2070 */ 2071 static __u64 (*bpf_get_current_cgroup_id)(void) = (void *) 80; 2072 2073 /* 2074 * bpf_get_local_storage 2075 * 2076 * Get the pointer to the local storage area. 2077 * The type and the size of the local storage is defined 2078 * by the *map* argument. 2079 * The *flags* meaning is specific for each map type, 2080 * and has to be 0 for cgroup local storage. 2081 * 2082 * Depending on the BPF program type, a local storage area 2083 * can be shared between multiple instances of the BPF program, 2084 * running simultaneously. 2085 * 2086 * A user should care about the synchronization by himself. 2087 * For example, by using the **BPF_ATOMIC** instructions to alter 2088 * the shared data. 2089 * 2090 * Returns 2091 * A pointer to the local storage area. 2092 */ 2093 static void *(*bpf_get_local_storage)(void *map, __u64 flags) = (void *) 81; 2094 2095 /* 2096 * bpf_sk_select_reuseport 2097 * 2098 * Select a **SO_REUSEPORT** socket from a 2099 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 2100 * It checks the selected socket is matching the incoming 2101 * request in the socket buffer. 2102 * 2103 * Returns 2104 * 0 on success, or a negative error in case of failure. 2105 */ 2106 static long (*bpf_sk_select_reuseport)(struct sk_reuseport_md *reuse, void *map, void *key, __u64 flags) = (void *) 82; 2107 2108 /* 2109 * bpf_skb_ancestor_cgroup_id 2110 * 2111 * Return id of cgroup v2 that is ancestor of cgroup associated 2112 * with the *skb* at the *ancestor_level*. The root cgroup is at 2113 * *ancestor_level* zero and each step down the hierarchy 2114 * increments the level. If *ancestor_level* == level of cgroup 2115 * associated with *skb*, then return value will be same as that 2116 * of **bpf_skb_cgroup_id**\ (). 2117 * 2118 * The helper is useful to implement policies based on cgroups 2119 * that are upper in hierarchy than immediate cgroup associated 2120 * with *skb*. 2121 * 2122 * The format of returned id and helper limitations are same as in 2123 * **bpf_skb_cgroup_id**\ (). 2124 * 2125 * Returns 2126 * The id is returned or 0 in case the id could not be retrieved. 2127 */ 2128 static __u64 (*bpf_skb_ancestor_cgroup_id)(struct __sk_buff *skb, int ancestor_level) = (void *) 83; 2129 2130 /* 2131 * bpf_sk_lookup_tcp 2132 * 2133 * Look for TCP socket matching *tuple*, optionally in a child 2134 * network namespace *netns*. The return value must be checked, 2135 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2136 * 2137 * The *ctx* should point to the context of the program, such as 2138 * the skb or socket (depending on the hook in use). This is used 2139 * to determine the base network namespace for the lookup. 2140 * 2141 * *tuple_size* must be one of: 2142 * 2143 * **sizeof**\ (*tuple*\ **->ipv4**) 2144 * Look for an IPv4 socket. 2145 * **sizeof**\ (*tuple*\ **->ipv6**) 2146 * Look for an IPv6 socket. 2147 * 2148 * If the *netns* is a negative signed 32-bit integer, then the 2149 * socket lookup table in the netns associated with the *ctx* 2150 * will be used. For the TC hooks, this is the netns of the device 2151 * in the skb. For socket hooks, this is the netns of the socket. 2152 * If *netns* is any other signed 32-bit value greater than or 2153 * equal to zero then it specifies the ID of the netns relative to 2154 * the netns associated with the *ctx*. *netns* values beyond the 2155 * range of 32-bit integers are reserved for future use. 2156 * 2157 * All values for *flags* are reserved for future usage, and must 2158 * be left at zero. 2159 * 2160 * This helper is available only if the kernel was compiled with 2161 * **CONFIG_NET** configuration option. 2162 * 2163 * Returns 2164 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2165 * For sockets with reuseport option, the **struct bpf_sock** 2166 * result is from *reuse*\ **->socks**\ [] using the hash of the 2167 * tuple. 2168 */ 2169 static struct bpf_sock *(*bpf_sk_lookup_tcp)(void *ctx, struct bpf_sock_tuple *tuple, __u32 tuple_size, __u64 netns, __u64 flags) = (void *) 84; 2170 2171 /* 2172 * bpf_sk_lookup_udp 2173 * 2174 * Look for UDP socket matching *tuple*, optionally in a child 2175 * network namespace *netns*. The return value must be checked, 2176 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2177 * 2178 * The *ctx* should point to the context of the program, such as 2179 * the skb or socket (depending on the hook in use). This is used 2180 * to determine the base network namespace for the lookup. 2181 * 2182 * *tuple_size* must be one of: 2183 * 2184 * **sizeof**\ (*tuple*\ **->ipv4**) 2185 * Look for an IPv4 socket. 2186 * **sizeof**\ (*tuple*\ **->ipv6**) 2187 * Look for an IPv6 socket. 2188 * 2189 * If the *netns* is a negative signed 32-bit integer, then the 2190 * socket lookup table in the netns associated with the *ctx* 2191 * will be used. For the TC hooks, this is the netns of the device 2192 * in the skb. For socket hooks, this is the netns of the socket. 2193 * If *netns* is any other signed 32-bit value greater than or 2194 * equal to zero then it specifies the ID of the netns relative to 2195 * the netns associated with the *ctx*. *netns* values beyond the 2196 * range of 32-bit integers are reserved for future use. 2197 * 2198 * All values for *flags* are reserved for future usage, and must 2199 * be left at zero. 2200 * 2201 * This helper is available only if the kernel was compiled with 2202 * **CONFIG_NET** configuration option. 2203 * 2204 * Returns 2205 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2206 * For sockets with reuseport option, the **struct bpf_sock** 2207 * result is from *reuse*\ **->socks**\ [] using the hash of the 2208 * tuple. 2209 */ 2210 static struct bpf_sock *(*bpf_sk_lookup_udp)(void *ctx, struct bpf_sock_tuple *tuple, __u32 tuple_size, __u64 netns, __u64 flags) = (void *) 85; 2211 2212 /* 2213 * bpf_sk_release 2214 * 2215 * Release the reference held by *sock*. *sock* must be a 2216 * non-**NULL** pointer that was returned from 2217 * **bpf_sk_lookup_xxx**\ (). 2218 * 2219 * Returns 2220 * 0 on success, or a negative error in case of failure. 2221 */ 2222 static long (*bpf_sk_release)(void *sock) = (void *) 86; 2223 2224 /* 2225 * bpf_map_push_elem 2226 * 2227 * Push an element *value* in *map*. *flags* is one of: 2228 * 2229 * **BPF_EXIST** 2230 * If the queue/stack is full, the oldest element is 2231 * removed to make room for this. 2232 * 2233 * Returns 2234 * 0 on success, or a negative error in case of failure. 2235 */ 2236 static long (*bpf_map_push_elem)(void *map, const void *value, __u64 flags) = (void *) 87; 2237 2238 /* 2239 * bpf_map_pop_elem 2240 * 2241 * Pop an element from *map*. 2242 * 2243 * Returns 2244 * 0 on success, or a negative error in case of failure. 2245 */ 2246 static long (*bpf_map_pop_elem)(void *map, void *value) = (void *) 88; 2247 2248 /* 2249 * bpf_map_peek_elem 2250 * 2251 * Get an element from *map* without removing it. 2252 * 2253 * Returns 2254 * 0 on success, or a negative error in case of failure. 2255 */ 2256 static long (*bpf_map_peek_elem)(void *map, void *value) = (void *) 89; 2257 2258 /* 2259 * bpf_msg_push_data 2260 * 2261 * For socket policies, insert *len* bytes into *msg* at offset 2262 * *start*. 2263 * 2264 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2265 * *msg* it may want to insert metadata or options into the *msg*. 2266 * This can later be read and used by any of the lower layer BPF 2267 * hooks. 2268 * 2269 * This helper may fail if under memory pressure (a malloc 2270 * fails) in these cases BPF programs will get an appropriate 2271 * error and BPF programs will need to handle them. 2272 * 2273 * Returns 2274 * 0 on success, or a negative error in case of failure. 2275 */ 2276 static long (*bpf_msg_push_data)(struct sk_msg_md *msg, __u32 start, __u32 len, __u64 flags) = (void *) 90; 2277 2278 /* 2279 * bpf_msg_pop_data 2280 * 2281 * Will remove *len* bytes from a *msg* starting at byte *start*. 2282 * This may result in **ENOMEM** errors under certain situations if 2283 * an allocation and copy are required due to a full ring buffer. 2284 * However, the helper will try to avoid doing the allocation 2285 * if possible. Other errors can occur if input parameters are 2286 * invalid either due to *start* byte not being valid part of *msg* 2287 * payload and/or *pop* value being to large. 2288 * 2289 * Returns 2290 * 0 on success, or a negative error in case of failure. 2291 */ 2292 static long (*bpf_msg_pop_data)(struct sk_msg_md *msg, __u32 start, __u32 len, __u64 flags) = (void *) 91; 2293 2294 /* 2295 * bpf_rc_pointer_rel 2296 * 2297 * This helper is used in programs implementing IR decoding, to 2298 * report a successfully decoded pointer movement. 2299 * 2300 * The *ctx* should point to the lirc sample as passed into 2301 * the program. 2302 * 2303 * This helper is only available is the kernel was compiled with 2304 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2305 * "**y**". 2306 * 2307 * Returns 2308 * 0 2309 */ 2310 static long (*bpf_rc_pointer_rel)(void *ctx, __s32 rel_x, __s32 rel_y) = (void *) 92; 2311 2312 /* 2313 * bpf_spin_lock 2314 * 2315 * Acquire a spinlock represented by the pointer *lock*, which is 2316 * stored as part of a value of a map. Taking the lock allows to 2317 * safely update the rest of the fields in that value. The 2318 * spinlock can (and must) later be released with a call to 2319 * **bpf_spin_unlock**\ (\ *lock*\ ). 2320 * 2321 * Spinlocks in BPF programs come with a number of restrictions 2322 * and constraints: 2323 * 2324 * * **bpf_spin_lock** objects are only allowed inside maps of 2325 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2326 * list could be extended in the future). 2327 * * BTF description of the map is mandatory. 2328 * * The BPF program can take ONE lock at a time, since taking two 2329 * or more could cause dead locks. 2330 * * Only one **struct bpf_spin_lock** is allowed per map element. 2331 * * When the lock is taken, calls (either BPF to BPF or helpers) 2332 * are not allowed. 2333 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2334 * allowed inside a spinlock-ed region. 2335 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2336 * the lock, on all execution paths, before it returns. 2337 * * The BPF program can access **struct bpf_spin_lock** only via 2338 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2339 * helpers. Loading or storing data into the **struct 2340 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2341 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2342 * of the map value must be a struct and have **struct 2343 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2344 * Nested lock inside another struct is not allowed. 2345 * * The **struct bpf_spin_lock** *lock* field in a map value must 2346 * be aligned on a multiple of 4 bytes in that value. 2347 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2348 * the **bpf_spin_lock** field to user space. 2349 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2350 * a BPF program, do not update the **bpf_spin_lock** field. 2351 * * **bpf_spin_lock** cannot be on the stack or inside a 2352 * networking packet (it can only be inside of a map values). 2353 * * **bpf_spin_lock** is available to root only. 2354 * * Tracing programs and socket filter programs cannot use 2355 * **bpf_spin_lock**\ () due to insufficient preemption checks 2356 * (but this may change in the future). 2357 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2358 * 2359 * Returns 2360 * 0 2361 */ 2362 static long (*bpf_spin_lock)(struct bpf_spin_lock *lock) = (void *) 93; 2363 2364 /* 2365 * bpf_spin_unlock 2366 * 2367 * Release the *lock* previously locked by a call to 2368 * **bpf_spin_lock**\ (\ *lock*\ ). 2369 * 2370 * Returns 2371 * 0 2372 */ 2373 static long (*bpf_spin_unlock)(struct bpf_spin_lock *lock) = (void *) 94; 2374 2375 /* 2376 * bpf_sk_fullsock 2377 * 2378 * This helper gets a **struct bpf_sock** pointer such 2379 * that all the fields in this **bpf_sock** can be accessed. 2380 * 2381 * Returns 2382 * A **struct bpf_sock** pointer on success, or **NULL** in 2383 * case of failure. 2384 */ 2385 static struct bpf_sock *(*bpf_sk_fullsock)(struct bpf_sock *sk) = (void *) 95; 2386 2387 /* 2388 * bpf_tcp_sock 2389 * 2390 * This helper gets a **struct bpf_tcp_sock** pointer from a 2391 * **struct bpf_sock** pointer. 2392 * 2393 * Returns 2394 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2395 * case of failure. 2396 */ 2397 static struct bpf_tcp_sock *(*bpf_tcp_sock)(struct bpf_sock *sk) = (void *) 96; 2398 2399 /* 2400 * bpf_skb_ecn_set_ce 2401 * 2402 * Set ECN (Explicit Congestion Notification) field of IP header 2403 * to **CE** (Congestion Encountered) if current value is **ECT** 2404 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2405 * and IPv4. 2406 * 2407 * Returns 2408 * 1 if the **CE** flag is set (either by the current helper call 2409 * or because it was already present), 0 if it is not set. 2410 */ 2411 static long (*bpf_skb_ecn_set_ce)(struct __sk_buff *skb) = (void *) 97; 2412 2413 /* 2414 * bpf_get_listener_sock 2415 * 2416 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2417 * **bpf_sk_release**\ () is unnecessary and not allowed. 2418 * 2419 * Returns 2420 * A **struct bpf_sock** pointer on success, or **NULL** in 2421 * case of failure. 2422 */ 2423 static struct bpf_sock *(*bpf_get_listener_sock)(struct bpf_sock *sk) = (void *) 98; 2424 2425 /* 2426 * bpf_skc_lookup_tcp 2427 * 2428 * Look for TCP socket matching *tuple*, optionally in a child 2429 * network namespace *netns*. The return value must be checked, 2430 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2431 * 2432 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 2433 * that it also returns timewait or request sockets. Use 2434 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 2435 * full structure. 2436 * 2437 * This helper is available only if the kernel was compiled with 2438 * **CONFIG_NET** configuration option. 2439 * 2440 * Returns 2441 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2442 * For sockets with reuseport option, the **struct bpf_sock** 2443 * result is from *reuse*\ **->socks**\ [] using the hash of the 2444 * tuple. 2445 */ 2446 static struct bpf_sock *(*bpf_skc_lookup_tcp)(void *ctx, struct bpf_sock_tuple *tuple, __u32 tuple_size, __u64 netns, __u64 flags) = (void *) 99; 2447 2448 /* 2449 * bpf_tcp_check_syncookie 2450 * 2451 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 2452 * the listening socket in *sk*. 2453 * 2454 * *iph* points to the start of the IPv4 or IPv6 header, while 2455 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2456 * **sizeof**\ (**struct ip6hdr**). 2457 * 2458 * *th* points to the start of the TCP header, while *th_len* 2459 * contains **sizeof**\ (**struct tcphdr**). 2460 * 2461 * Returns 2462 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 2463 * error otherwise. 2464 */ 2465 static long (*bpf_tcp_check_syncookie)(void *sk, void *iph, __u32 iph_len, struct tcphdr *th, __u32 th_len) = (void *) 100; 2466 2467 /* 2468 * bpf_sysctl_get_name 2469 * 2470 * Get name of sysctl in /proc/sys/ and copy it into provided by 2471 * program buffer *buf* of size *buf_len*. 2472 * 2473 * The buffer is always NUL terminated, unless it's zero-sized. 2474 * 2475 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2476 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2477 * only (e.g. "tcp_mem"). 2478 * 2479 * Returns 2480 * Number of character copied (not including the trailing NUL). 2481 * 2482 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2483 * truncated name in this case). 2484 */ 2485 static long (*bpf_sysctl_get_name)(struct bpf_sysctl *ctx, char *buf, unsigned long buf_len, __u64 flags) = (void *) 101; 2486 2487 /* 2488 * bpf_sysctl_get_current_value 2489 * 2490 * Get current value of sysctl as it is presented in /proc/sys 2491 * (incl. newline, etc), and copy it as a string into provided 2492 * by program buffer *buf* of size *buf_len*. 2493 * 2494 * The whole value is copied, no matter what file position user 2495 * space issued e.g. sys_read at. 2496 * 2497 * The buffer is always NUL terminated, unless it's zero-sized. 2498 * 2499 * Returns 2500 * Number of character copied (not including the trailing NUL). 2501 * 2502 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2503 * truncated name in this case). 2504 * 2505 * **-EINVAL** if current value was unavailable, e.g. because 2506 * sysctl is uninitialized and read returns -EIO for it. 2507 */ 2508 static long (*bpf_sysctl_get_current_value)(struct bpf_sysctl *ctx, char *buf, unsigned long buf_len) = (void *) 102; 2509 2510 /* 2511 * bpf_sysctl_get_new_value 2512 * 2513 * Get new value being written by user space to sysctl (before 2514 * the actual write happens) and copy it as a string into 2515 * provided by program buffer *buf* of size *buf_len*. 2516 * 2517 * User space may write new value at file position > 0. 2518 * 2519 * The buffer is always NUL terminated, unless it's zero-sized. 2520 * 2521 * Returns 2522 * Number of character copied (not including the trailing NUL). 2523 * 2524 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2525 * truncated name in this case). 2526 * 2527 * **-EINVAL** if sysctl is being read. 2528 */ 2529 static long (*bpf_sysctl_get_new_value)(struct bpf_sysctl *ctx, char *buf, unsigned long buf_len) = (void *) 103; 2530 2531 /* 2532 * bpf_sysctl_set_new_value 2533 * 2534 * Override new value being written by user space to sysctl with 2535 * value provided by program in buffer *buf* of size *buf_len*. 2536 * 2537 * *buf* should contain a string in same form as provided by user 2538 * space on sysctl write. 2539 * 2540 * User space may write new value at file position > 0. To override 2541 * the whole sysctl value file position should be set to zero. 2542 * 2543 * Returns 2544 * 0 on success. 2545 * 2546 * **-E2BIG** if the *buf_len* is too big. 2547 * 2548 * **-EINVAL** if sysctl is being read. 2549 */ 2550 static long (*bpf_sysctl_set_new_value)(struct bpf_sysctl *ctx, const char *buf, unsigned long buf_len) = (void *) 104; 2551 2552 /* 2553 * bpf_strtol 2554 * 2555 * Convert the initial part of the string from buffer *buf* of 2556 * size *buf_len* to a long integer according to the given base 2557 * and save the result in *res*. 2558 * 2559 * The string may begin with an arbitrary amount of white space 2560 * (as determined by **isspace**\ (3)) followed by a single 2561 * optional '**-**' sign. 2562 * 2563 * Five least significant bits of *flags* encode base, other bits 2564 * are currently unused. 2565 * 2566 * Base must be either 8, 10, 16 or 0 to detect it automatically 2567 * similar to user space **strtol**\ (3). 2568 * 2569 * Returns 2570 * Number of characters consumed on success. Must be positive but 2571 * no more than *buf_len*. 2572 * 2573 * **-EINVAL** if no valid digits were found or unsupported base 2574 * was provided. 2575 * 2576 * **-ERANGE** if resulting value was out of range. 2577 */ 2578 static long (*bpf_strtol)(const char *buf, unsigned long buf_len, __u64 flags, long *res) = (void *) 105; 2579 2580 /* 2581 * bpf_strtoul 2582 * 2583 * Convert the initial part of the string from buffer *buf* of 2584 * size *buf_len* to an unsigned long integer according to the 2585 * given base and save the result in *res*. 2586 * 2587 * The string may begin with an arbitrary amount of white space 2588 * (as determined by **isspace**\ (3)). 2589 * 2590 * Five least significant bits of *flags* encode base, other bits 2591 * are currently unused. 2592 * 2593 * Base must be either 8, 10, 16 or 0 to detect it automatically 2594 * similar to user space **strtoul**\ (3). 2595 * 2596 * Returns 2597 * Number of characters consumed on success. Must be positive but 2598 * no more than *buf_len*. 2599 * 2600 * **-EINVAL** if no valid digits were found or unsupported base 2601 * was provided. 2602 * 2603 * **-ERANGE** if resulting value was out of range. 2604 */ 2605 static long (*bpf_strtoul)(const char *buf, unsigned long buf_len, __u64 flags, unsigned long *res) = (void *) 106; 2606 2607 /* 2608 * bpf_sk_storage_get 2609 * 2610 * Get a bpf-local-storage from a *sk*. 2611 * 2612 * Logically, it could be thought of getting the value from 2613 * a *map* with *sk* as the **key**. From this 2614 * perspective, the usage is not much different from 2615 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 2616 * helper enforces the key must be a full socket and the map must 2617 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 2618 * 2619 * Underneath, the value is stored locally at *sk* instead of 2620 * the *map*. The *map* is used as the bpf-local-storage 2621 * "type". The bpf-local-storage "type" (i.e. the *map*) is 2622 * searched against all bpf-local-storages residing at *sk*. 2623 * 2624 * *sk* is a kernel **struct sock** pointer for LSM program. 2625 * *sk* is a **struct bpf_sock** pointer for other program types. 2626 * 2627 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 2628 * used such that a new bpf-local-storage will be 2629 * created if one does not exist. *value* can be used 2630 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 2631 * the initial value of a bpf-local-storage. If *value* is 2632 * **NULL**, the new bpf-local-storage will be zero initialized. 2633 * 2634 * Returns 2635 * A bpf-local-storage pointer is returned on success. 2636 * 2637 * **NULL** if not found or there was an error in adding 2638 * a new bpf-local-storage. 2639 */ 2640 static void *(*bpf_sk_storage_get)(void *map, void *sk, void *value, __u64 flags) = (void *) 107; 2641 2642 /* 2643 * bpf_sk_storage_delete 2644 * 2645 * Delete a bpf-local-storage from a *sk*. 2646 * 2647 * Returns 2648 * 0 on success. 2649 * 2650 * **-ENOENT** if the bpf-local-storage cannot be found. 2651 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 2652 */ 2653 static long (*bpf_sk_storage_delete)(void *map, void *sk) = (void *) 108; 2654 2655 /* 2656 * bpf_send_signal 2657 * 2658 * Send signal *sig* to the process of the current task. 2659 * The signal may be delivered to any of this process's threads. 2660 * 2661 * Returns 2662 * 0 on success or successfully queued. 2663 * 2664 * **-EBUSY** if work queue under nmi is full. 2665 * 2666 * **-EINVAL** if *sig* is invalid. 2667 * 2668 * **-EPERM** if no permission to send the *sig*. 2669 * 2670 * **-EAGAIN** if bpf program can try again. 2671 */ 2672 static long (*bpf_send_signal)(__u32 sig) = (void *) 109; 2673 2674 /* 2675 * bpf_tcp_gen_syncookie 2676 * 2677 * Try to issue a SYN cookie for the packet with corresponding 2678 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 2679 * 2680 * *iph* points to the start of the IPv4 or IPv6 header, while 2681 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2682 * **sizeof**\ (**struct ip6hdr**). 2683 * 2684 * *th* points to the start of the TCP header, while *th_len* 2685 * contains the length of the TCP header. 2686 * 2687 * Returns 2688 * On success, lower 32 bits hold the generated SYN cookie in 2689 * followed by 16 bits which hold the MSS value for that cookie, 2690 * and the top 16 bits are unused. 2691 * 2692 * On failure, the returned value is one of the following: 2693 * 2694 * **-EINVAL** SYN cookie cannot be issued due to error 2695 * 2696 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 2697 * 2698 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 2699 * 2700 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 2701 */ 2702 static __s64 (*bpf_tcp_gen_syncookie)(void *sk, void *iph, __u32 iph_len, struct tcphdr *th, __u32 th_len) = (void *) 110; 2703 2704 /* 2705 * bpf_skb_output 2706 * 2707 * Write raw *data* blob into a special BPF perf event held by 2708 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2709 * event must have the following attributes: **PERF_SAMPLE_RAW** 2710 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2711 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2712 * 2713 * The *flags* are used to indicate the index in *map* for which 2714 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2715 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2716 * to indicate that the index of the current CPU core should be 2717 * used. 2718 * 2719 * The value to write, of *size*, is passed through eBPF stack and 2720 * pointed by *data*. 2721 * 2722 * *ctx* is a pointer to in-kernel struct sk_buff. 2723 * 2724 * This helper is similar to **bpf_perf_event_output**\ () but 2725 * restricted to raw_tracepoint bpf programs. 2726 * 2727 * Returns 2728 * 0 on success, or a negative error in case of failure. 2729 */ 2730 static long (*bpf_skb_output)(void *ctx, void *map, __u64 flags, void *data, __u64 size) = (void *) 111; 2731 2732 /* 2733 * bpf_probe_read_user 2734 * 2735 * Safely attempt to read *size* bytes from user space address 2736 * *unsafe_ptr* and store the data in *dst*. 2737 * 2738 * Returns 2739 * 0 on success, or a negative error in case of failure. 2740 */ 2741 static long (*bpf_probe_read_user)(void *dst, __u32 size, const void *unsafe_ptr) = (void *) 112; 2742 2743 /* 2744 * bpf_probe_read_kernel 2745 * 2746 * Safely attempt to read *size* bytes from kernel space address 2747 * *unsafe_ptr* and store the data in *dst*. 2748 * 2749 * Returns 2750 * 0 on success, or a negative error in case of failure. 2751 */ 2752 static long (*bpf_probe_read_kernel)(void *dst, __u32 size, const void *unsafe_ptr) = (void *) 113; 2753 2754 /* 2755 * bpf_probe_read_user_str 2756 * 2757 * Copy a NUL terminated string from an unsafe user address 2758 * *unsafe_ptr* to *dst*. The *size* should include the 2759 * terminating NUL byte. In case the string length is smaller than 2760 * *size*, the target is not padded with further NUL bytes. If the 2761 * string length is larger than *size*, just *size*-1 bytes are 2762 * copied and the last byte is set to NUL. 2763 * 2764 * On success, returns the number of bytes that were written, 2765 * including the terminal NUL. This makes this helper useful in 2766 * tracing programs for reading strings, and more importantly to 2767 * get its length at runtime. See the following snippet: 2768 * 2769 * :: 2770 * 2771 * SEC("kprobe/sys_open") 2772 * void bpf_sys_open(struct pt_regs *ctx) 2773 * { 2774 * char buf[PATHLEN]; // PATHLEN is defined to 256 2775 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 2776 * ctx->di); 2777 * 2778 * // Consume buf, for example push it to 2779 * // userspace via bpf_perf_event_output(); we 2780 * // can use res (the string length) as event 2781 * // size, after checking its boundaries. 2782 * } 2783 * 2784 * In comparison, using **bpf_probe_read_user**\ () helper here 2785 * instead to read the string would require to estimate the length 2786 * at compile time, and would often result in copying more memory 2787 * than necessary. 2788 * 2789 * Another useful use case is when parsing individual process 2790 * arguments or individual environment variables navigating 2791 * *current*\ **->mm->arg_start** and *current*\ 2792 * **->mm->env_start**: using this helper and the return value, 2793 * one can quickly iterate at the right offset of the memory area. 2794 * 2795 * Returns 2796 * On success, the strictly positive length of the output string, 2797 * including the trailing NUL character. On error, a negative 2798 * value. 2799 */ 2800 static long (*bpf_probe_read_user_str)(void *dst, __u32 size, const void *unsafe_ptr) = (void *) 114; 2801 2802 /* 2803 * bpf_probe_read_kernel_str 2804 * 2805 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 2806 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 2807 * 2808 * Returns 2809 * On success, the strictly positive length of the string, including 2810 * the trailing NUL character. On error, a negative value. 2811 */ 2812 static long (*bpf_probe_read_kernel_str)(void *dst, __u32 size, const void *unsafe_ptr) = (void *) 115; 2813 2814 /* 2815 * bpf_tcp_send_ack 2816 * 2817 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 2818 * *rcv_nxt* is the ack_seq to be sent out. 2819 * 2820 * Returns 2821 * 0 on success, or a negative error in case of failure. 2822 */ 2823 static long (*bpf_tcp_send_ack)(void *tp, __u32 rcv_nxt) = (void *) 116; 2824 2825 /* 2826 * bpf_send_signal_thread 2827 * 2828 * Send signal *sig* to the thread corresponding to the current task. 2829 * 2830 * Returns 2831 * 0 on success or successfully queued. 2832 * 2833 * **-EBUSY** if work queue under nmi is full. 2834 * 2835 * **-EINVAL** if *sig* is invalid. 2836 * 2837 * **-EPERM** if no permission to send the *sig*. 2838 * 2839 * **-EAGAIN** if bpf program can try again. 2840 */ 2841 static long (*bpf_send_signal_thread)(__u32 sig) = (void *) 117; 2842 2843 /* 2844 * bpf_jiffies64 2845 * 2846 * Obtain the 64bit jiffies 2847 * 2848 * Returns 2849 * The 64 bit jiffies 2850 */ 2851 static __u64 (*bpf_jiffies64)(void) = (void *) 118; 2852 2853 /* 2854 * bpf_read_branch_records 2855 * 2856 * For an eBPF program attached to a perf event, retrieve the 2857 * branch records (**struct perf_branch_entry**) associated to *ctx* 2858 * and store it in the buffer pointed by *buf* up to size 2859 * *size* bytes. 2860 * 2861 * Returns 2862 * On success, number of bytes written to *buf*. On error, a 2863 * negative value. 2864 * 2865 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 2866 * instead return the number of bytes required to store all the 2867 * branch entries. If this flag is set, *buf* may be NULL. 2868 * 2869 * **-EINVAL** if arguments invalid or **size** not a multiple 2870 * of **sizeof**\ (**struct perf_branch_entry**\ ). 2871 * 2872 * **-ENOENT** if architecture does not support branch records. 2873 */ 2874 static long (*bpf_read_branch_records)(struct bpf_perf_event_data *ctx, void *buf, __u32 size, __u64 flags) = (void *) 119; 2875 2876 /* 2877 * bpf_get_ns_current_pid_tgid 2878 * 2879 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 2880 * *namespace* will be returned in *nsdata*. 2881 * 2882 * Returns 2883 * 0 on success, or one of the following in case of failure: 2884 * 2885 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 2886 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 2887 * 2888 * **-ENOENT** if pidns does not exists for the current task. 2889 */ 2890 static long (*bpf_get_ns_current_pid_tgid)(__u64 dev, __u64 ino, struct bpf_pidns_info *nsdata, __u32 size) = (void *) 120; 2891 2892 /* 2893 * bpf_xdp_output 2894 * 2895 * Write raw *data* blob into a special BPF perf event held by 2896 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2897 * event must have the following attributes: **PERF_SAMPLE_RAW** 2898 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2899 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2900 * 2901 * The *flags* are used to indicate the index in *map* for which 2902 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2903 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2904 * to indicate that the index of the current CPU core should be 2905 * used. 2906 * 2907 * The value to write, of *size*, is passed through eBPF stack and 2908 * pointed by *data*. 2909 * 2910 * *ctx* is a pointer to in-kernel struct xdp_buff. 2911 * 2912 * This helper is similar to **bpf_perf_eventoutput**\ () but 2913 * restricted to raw_tracepoint bpf programs. 2914 * 2915 * Returns 2916 * 0 on success, or a negative error in case of failure. 2917 */ 2918 static long (*bpf_xdp_output)(void *ctx, void *map, __u64 flags, void *data, __u64 size) = (void *) 121; 2919 2920 /* 2921 * bpf_get_netns_cookie 2922 * 2923 * Retrieve the cookie (generated by the kernel) of the network 2924 * namespace the input *ctx* is associated with. The network 2925 * namespace cookie remains stable for its lifetime and provides 2926 * a global identifier that can be assumed unique. If *ctx* is 2927 * NULL, then the helper returns the cookie for the initial 2928 * network namespace. The cookie itself is very similar to that 2929 * of **bpf_get_socket_cookie**\ () helper, but for network 2930 * namespaces instead of sockets. 2931 * 2932 * Returns 2933 * A 8-byte long opaque number. 2934 */ 2935 static __u64 (*bpf_get_netns_cookie)(void *ctx) = (void *) 122; 2936 2937 /* 2938 * bpf_get_current_ancestor_cgroup_id 2939 * 2940 * Return id of cgroup v2 that is ancestor of the cgroup associated 2941 * with the current task at the *ancestor_level*. The root cgroup 2942 * is at *ancestor_level* zero and each step down the hierarchy 2943 * increments the level. If *ancestor_level* == level of cgroup 2944 * associated with the current task, then return value will be the 2945 * same as that of **bpf_get_current_cgroup_id**\ (). 2946 * 2947 * The helper is useful to implement policies based on cgroups 2948 * that are upper in hierarchy than immediate cgroup associated 2949 * with the current task. 2950 * 2951 * The format of returned id and helper limitations are same as in 2952 * **bpf_get_current_cgroup_id**\ (). 2953 * 2954 * Returns 2955 * The id is returned or 0 in case the id could not be retrieved. 2956 */ 2957 static __u64 (*bpf_get_current_ancestor_cgroup_id)(int ancestor_level) = (void *) 123; 2958 2959 /* 2960 * bpf_sk_assign 2961 * 2962 * Helper is overloaded depending on BPF program type. This 2963 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 2964 * **BPF_PROG_TYPE_SCHED_ACT** programs. 2965 * 2966 * Assign the *sk* to the *skb*. When combined with appropriate 2967 * routing configuration to receive the packet towards the socket, 2968 * will cause *skb* to be delivered to the specified socket. 2969 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 2970 * **bpf_clone_redirect**\ () or other methods outside of BPF may 2971 * interfere with successful delivery to the socket. 2972 * 2973 * This operation is only valid from TC ingress path. 2974 * 2975 * The *flags* argument must be zero. 2976 * 2977 * Returns 2978 * 0 on success, or a negative error in case of failure: 2979 * 2980 * **-EINVAL** if specified *flags* are not supported. 2981 * 2982 * **-ENOENT** if the socket is unavailable for assignment. 2983 * 2984 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 2985 * 2986 * **-EOPNOTSUPP** if the operation is not supported, for example 2987 * a call from outside of TC ingress. 2988 * 2989 * **-ESOCKTNOSUPPORT** if the socket type is not supported 2990 * (reuseport). 2991 */ 2992 static long (*bpf_sk_assign)(void *ctx, void *sk, __u64 flags) = (void *) 124; 2993 2994 /* 2995 * bpf_ktime_get_boot_ns 2996 * 2997 * Return the time elapsed since system boot, in nanoseconds. 2998 * Does include the time the system was suspended. 2999 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 3000 * 3001 * Returns 3002 * Current *ktime*. 3003 */ 3004 static __u64 (*bpf_ktime_get_boot_ns)(void) = (void *) 125; 3005 3006 /* 3007 * bpf_seq_printf 3008 * 3009 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 3010 * out the format string. 3011 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 3012 * the format string itself. The *data* and *data_len* are format string 3013 * arguments. The *data* are a **u64** array and corresponding format string 3014 * values are stored in the array. For strings and pointers where pointees 3015 * are accessed, only the pointer values are stored in the *data* array. 3016 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 3017 * 3018 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 3019 * Reading kernel memory may fail due to either invalid address or 3020 * valid address but requiring a major memory fault. If reading kernel memory 3021 * fails, the string for **%s** will be an empty string, and the ip 3022 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 3023 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 3024 * 3025 * Returns 3026 * 0 on success, or a negative error in case of failure: 3027 * 3028 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 3029 * by returning 1 from bpf program. 3030 * 3031 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 3032 * 3033 * **-E2BIG** if *fmt* contains too many format specifiers. 3034 * 3035 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3036 */ 3037 static long (*bpf_seq_printf)(struct seq_file *m, const char *fmt, __u32 fmt_size, const void *data, __u32 data_len) = (void *) 126; 3038 3039 /* 3040 * bpf_seq_write 3041 * 3042 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 3043 * The *m* represents the seq_file. The *data* and *len* represent the 3044 * data to write in bytes. 3045 * 3046 * Returns 3047 * 0 on success, or a negative error in case of failure: 3048 * 3049 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3050 */ 3051 static long (*bpf_seq_write)(struct seq_file *m, const void *data, __u32 len) = (void *) 127; 3052 3053 /* 3054 * bpf_sk_cgroup_id 3055 * 3056 * Return the cgroup v2 id of the socket *sk*. 3057 * 3058 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 3059 * returned from **bpf_sk_lookup_xxx**\ (), 3060 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 3061 * same as in **bpf_skb_cgroup_id**\ (). 3062 * 3063 * This helper is available only if the kernel was compiled with 3064 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 3065 * 3066 * Returns 3067 * The id is returned or 0 in case the id could not be retrieved. 3068 */ 3069 static __u64 (*bpf_sk_cgroup_id)(void *sk) = (void *) 128; 3070 3071 /* 3072 * bpf_sk_ancestor_cgroup_id 3073 * 3074 * Return id of cgroup v2 that is ancestor of cgroup associated 3075 * with the *sk* at the *ancestor_level*. The root cgroup is at 3076 * *ancestor_level* zero and each step down the hierarchy 3077 * increments the level. If *ancestor_level* == level of cgroup 3078 * associated with *sk*, then return value will be same as that 3079 * of **bpf_sk_cgroup_id**\ (). 3080 * 3081 * The helper is useful to implement policies based on cgroups 3082 * that are upper in hierarchy than immediate cgroup associated 3083 * with *sk*. 3084 * 3085 * The format of returned id and helper limitations are same as in 3086 * **bpf_sk_cgroup_id**\ (). 3087 * 3088 * Returns 3089 * The id is returned or 0 in case the id could not be retrieved. 3090 */ 3091 static __u64 (*bpf_sk_ancestor_cgroup_id)(void *sk, int ancestor_level) = (void *) 129; 3092 3093 /* 3094 * bpf_ringbuf_output 3095 * 3096 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 3097 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3098 * of new data availability is sent. 3099 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3100 * of new data availability is sent unconditionally. 3101 * If **0** is specified in *flags*, an adaptive notification 3102 * of new data availability is sent. 3103 * 3104 * An adaptive notification is a notification sent whenever the user-space 3105 * process has caught up and consumed all available payloads. In case the user-space 3106 * process is still processing a previous payload, then no notification is needed 3107 * as it will process the newly added payload automatically. 3108 * 3109 * Returns 3110 * 0 on success, or a negative error in case of failure. 3111 */ 3112 static long (*bpf_ringbuf_output)(void *ringbuf, void *data, __u64 size, __u64 flags) = (void *) 130; 3113 3114 /* 3115 * bpf_ringbuf_reserve 3116 * 3117 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 3118 * *flags* must be 0. 3119 * 3120 * Returns 3121 * Valid pointer with *size* bytes of memory available; NULL, 3122 * otherwise. 3123 */ 3124 static void *(*bpf_ringbuf_reserve)(void *ringbuf, __u64 size, __u64 flags) = (void *) 131; 3125 3126 /* 3127 * bpf_ringbuf_submit 3128 * 3129 * Submit reserved ring buffer sample, pointed to by *data*. 3130 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3131 * of new data availability is sent. 3132 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3133 * of new data availability is sent unconditionally. 3134 * If **0** is specified in *flags*, an adaptive notification 3135 * of new data availability is sent. 3136 * 3137 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 3138 * 3139 * Returns 3140 * Nothing. Always succeeds. 3141 */ 3142 static void (*bpf_ringbuf_submit)(void *data, __u64 flags) = (void *) 132; 3143 3144 /* 3145 * bpf_ringbuf_discard 3146 * 3147 * Discard reserved ring buffer sample, pointed to by *data*. 3148 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3149 * of new data availability is sent. 3150 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3151 * of new data availability is sent unconditionally. 3152 * If **0** is specified in *flags*, an adaptive notification 3153 * of new data availability is sent. 3154 * 3155 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 3156 * 3157 * Returns 3158 * Nothing. Always succeeds. 3159 */ 3160 static void (*bpf_ringbuf_discard)(void *data, __u64 flags) = (void *) 133; 3161 3162 /* 3163 * bpf_ringbuf_query 3164 * 3165 * Query various characteristics of provided ring buffer. What 3166 * exactly is queries is determined by *flags*: 3167 * 3168 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 3169 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 3170 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 3171 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 3172 * 3173 * Data returned is just a momentary snapshot of actual values 3174 * and could be inaccurate, so this facility should be used to 3175 * power heuristics and for reporting, not to make 100% correct 3176 * calculation. 3177 * 3178 * Returns 3179 * Requested value, or 0, if *flags* are not recognized. 3180 */ 3181 static __u64 (*bpf_ringbuf_query)(void *ringbuf, __u64 flags) = (void *) 134; 3182 3183 /* 3184 * bpf_csum_level 3185 * 3186 * Change the skbs checksum level by one layer up or down, or 3187 * reset it entirely to none in order to have the stack perform 3188 * checksum validation. The level is applicable to the following 3189 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 3190 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 3191 * through **bpf_skb_adjust_room**\ () helper with passing in 3192 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 3193 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 3194 * the UDP header is removed. Similarly, an encap of the latter 3195 * into the former could be accompanied by a helper call to 3196 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 3197 * skb is still intended to be processed in higher layers of the 3198 * stack instead of just egressing at tc. 3199 * 3200 * There are three supported level settings at this time: 3201 * 3202 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 3203 * with CHECKSUM_UNNECESSARY. 3204 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 3205 * with CHECKSUM_UNNECESSARY. 3206 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 3207 * sets CHECKSUM_NONE to force checksum validation by the stack. 3208 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 3209 * skb->csum_level. 3210 * 3211 * Returns 3212 * 0 on success, or a negative error in case of failure. In the 3213 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 3214 * is returned or the error code -EACCES in case the skb is not 3215 * subject to CHECKSUM_UNNECESSARY. 3216 */ 3217 static long (*bpf_csum_level)(struct __sk_buff *skb, __u64 level) = (void *) 135; 3218 3219 /* 3220 * bpf_skc_to_tcp6_sock 3221 * 3222 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 3223 * 3224 * Returns 3225 * *sk* if casting is valid, or **NULL** otherwise. 3226 */ 3227 static struct tcp6_sock *(*bpf_skc_to_tcp6_sock)(void *sk) = (void *) 136; 3228 3229 /* 3230 * bpf_skc_to_tcp_sock 3231 * 3232 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 3233 * 3234 * Returns 3235 * *sk* if casting is valid, or **NULL** otherwise. 3236 */ 3237 static struct tcp_sock *(*bpf_skc_to_tcp_sock)(void *sk) = (void *) 137; 3238 3239 /* 3240 * bpf_skc_to_tcp_timewait_sock 3241 * 3242 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 3243 * 3244 * Returns 3245 * *sk* if casting is valid, or **NULL** otherwise. 3246 */ 3247 static struct tcp_timewait_sock *(*bpf_skc_to_tcp_timewait_sock)(void *sk) = (void *) 138; 3248 3249 /* 3250 * bpf_skc_to_tcp_request_sock 3251 * 3252 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 3253 * 3254 * Returns 3255 * *sk* if casting is valid, or **NULL** otherwise. 3256 */ 3257 static struct tcp_request_sock *(*bpf_skc_to_tcp_request_sock)(void *sk) = (void *) 139; 3258 3259 /* 3260 * bpf_skc_to_udp6_sock 3261 * 3262 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 3263 * 3264 * Returns 3265 * *sk* if casting is valid, or **NULL** otherwise. 3266 */ 3267 static struct udp6_sock *(*bpf_skc_to_udp6_sock)(void *sk) = (void *) 140; 3268 3269 /* 3270 * bpf_get_task_stack 3271 * 3272 * Return a user or a kernel stack in bpf program provided buffer. 3273 * To achieve this, the helper needs *task*, which is a valid 3274 * pointer to **struct task_struct**. To store the stacktrace, the 3275 * bpf program provides *buf* with a nonnegative *size*. 3276 * 3277 * The last argument, *flags*, holds the number of stack frames to 3278 * skip (from 0 to 255), masked with 3279 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3280 * the following flags: 3281 * 3282 * **BPF_F_USER_STACK** 3283 * Collect a user space stack instead of a kernel stack. 3284 * **BPF_F_USER_BUILD_ID** 3285 * Collect buildid+offset instead of ips for user stack, 3286 * only valid if **BPF_F_USER_STACK** is also specified. 3287 * 3288 * **bpf_get_task_stack**\ () can collect up to 3289 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3290 * to sufficient large buffer size. Note that 3291 * this limit can be controlled with the **sysctl** program, and 3292 * that it should be manually increased in order to profile long 3293 * user stacks (such as stacks for Java programs). To do so, use: 3294 * 3295 * :: 3296 * 3297 * # sysctl kernel.perf_event_max_stack=<new value> 3298 * 3299 * Returns 3300 * A non-negative value equal to or less than *size* on success, 3301 * or a negative error in case of failure. 3302 */ 3303 static long (*bpf_get_task_stack)(struct task_struct *task, void *buf, __u32 size, __u64 flags) = (void *) 141; 3304 3305 /* 3306 * bpf_load_hdr_opt 3307 * 3308 * Load header option. Support reading a particular TCP header 3309 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 3310 * 3311 * If *flags* is 0, it will search the option from the 3312 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 3313 * has details on what skb_data contains under different 3314 * *skops*\ **->op**. 3315 * 3316 * The first byte of the *searchby_res* specifies the 3317 * kind that it wants to search. 3318 * 3319 * If the searching kind is an experimental kind 3320 * (i.e. 253 or 254 according to RFC6994). It also 3321 * needs to specify the "magic" which is either 3322 * 2 bytes or 4 bytes. It then also needs to 3323 * specify the size of the magic by using 3324 * the 2nd byte which is "kind-length" of a TCP 3325 * header option and the "kind-length" also 3326 * includes the first 2 bytes "kind" and "kind-length" 3327 * itself as a normal TCP header option also does. 3328 * 3329 * For example, to search experimental kind 254 with 3330 * 2 byte magic 0xeB9F, the searchby_res should be 3331 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 3332 * 3333 * To search for the standard window scale option (3), 3334 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 3335 * Note, kind-length must be 0 for regular option. 3336 * 3337 * Searching for No-Op (0) and End-of-Option-List (1) are 3338 * not supported. 3339 * 3340 * *len* must be at least 2 bytes which is the minimal size 3341 * of a header option. 3342 * 3343 * Supported flags: 3344 * 3345 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 3346 * saved_syn packet or the just-received syn packet. 3347 * 3348 * 3349 * Returns 3350 * > 0 when found, the header option is copied to *searchby_res*. 3351 * The return value is the total length copied. On failure, a 3352 * negative error code is returned: 3353 * 3354 * **-EINVAL** if a parameter is invalid. 3355 * 3356 * **-ENOMSG** if the option is not found. 3357 * 3358 * **-ENOENT** if no syn packet is available when 3359 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 3360 * 3361 * **-ENOSPC** if there is not enough space. Only *len* number of 3362 * bytes are copied. 3363 * 3364 * **-EFAULT** on failure to parse the header options in the 3365 * packet. 3366 * 3367 * **-EPERM** if the helper cannot be used under the current 3368 * *skops*\ **->op**. 3369 */ 3370 static long (*bpf_load_hdr_opt)(struct bpf_sock_ops *skops, void *searchby_res, __u32 len, __u64 flags) = (void *) 142; 3371 3372 /* 3373 * bpf_store_hdr_opt 3374 * 3375 * Store header option. The data will be copied 3376 * from buffer *from* with length *len* to the TCP header. 3377 * 3378 * The buffer *from* should have the whole option that 3379 * includes the kind, kind-length, and the actual 3380 * option data. The *len* must be at least kind-length 3381 * long. The kind-length does not have to be 4 byte 3382 * aligned. The kernel will take care of the padding 3383 * and setting the 4 bytes aligned value to th->doff. 3384 * 3385 * This helper will check for duplicated option 3386 * by searching the same option in the outgoing skb. 3387 * 3388 * This helper can only be called during 3389 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 3390 * 3391 * 3392 * Returns 3393 * 0 on success, or negative error in case of failure: 3394 * 3395 * **-EINVAL** If param is invalid. 3396 * 3397 * **-ENOSPC** if there is not enough space in the header. 3398 * Nothing has been written 3399 * 3400 * **-EEXIST** if the option already exists. 3401 * 3402 * **-EFAULT** on failrue to parse the existing header options. 3403 * 3404 * **-EPERM** if the helper cannot be used under the current 3405 * *skops*\ **->op**. 3406 */ 3407 static long (*bpf_store_hdr_opt)(struct bpf_sock_ops *skops, const void *from, __u32 len, __u64 flags) = (void *) 143; 3408 3409 /* 3410 * bpf_reserve_hdr_opt 3411 * 3412 * Reserve *len* bytes for the bpf header option. The 3413 * space will be used by **bpf_store_hdr_opt**\ () later in 3414 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 3415 * 3416 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 3417 * the total number of bytes will be reserved. 3418 * 3419 * This helper can only be called during 3420 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 3421 * 3422 * 3423 * Returns 3424 * 0 on success, or negative error in case of failure: 3425 * 3426 * **-EINVAL** if a parameter is invalid. 3427 * 3428 * **-ENOSPC** if there is not enough space in the header. 3429 * 3430 * **-EPERM** if the helper cannot be used under the current 3431 * *skops*\ **->op**. 3432 */ 3433 static long (*bpf_reserve_hdr_opt)(struct bpf_sock_ops *skops, __u32 len, __u64 flags) = (void *) 144; 3434 3435 /* 3436 * bpf_inode_storage_get 3437 * 3438 * Get a bpf_local_storage from an *inode*. 3439 * 3440 * Logically, it could be thought of as getting the value from 3441 * a *map* with *inode* as the **key**. From this 3442 * perspective, the usage is not much different from 3443 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 3444 * helper enforces the key must be an inode and the map must also 3445 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 3446 * 3447 * Underneath, the value is stored locally at *inode* instead of 3448 * the *map*. The *map* is used as the bpf-local-storage 3449 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3450 * searched against all bpf_local_storage residing at *inode*. 3451 * 3452 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 3453 * used such that a new bpf_local_storage will be 3454 * created if one does not exist. *value* can be used 3455 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 3456 * the initial value of a bpf_local_storage. If *value* is 3457 * **NULL**, the new bpf_local_storage will be zero initialized. 3458 * 3459 * Returns 3460 * A bpf_local_storage pointer is returned on success. 3461 * 3462 * **NULL** if not found or there was an error in adding 3463 * a new bpf_local_storage. 3464 */ 3465 static void *(*bpf_inode_storage_get)(void *map, void *inode, void *value, __u64 flags) = (void *) 145; 3466 3467 /* 3468 * bpf_inode_storage_delete 3469 * 3470 * Delete a bpf_local_storage from an *inode*. 3471 * 3472 * Returns 3473 * 0 on success. 3474 * 3475 * **-ENOENT** if the bpf_local_storage cannot be found. 3476 */ 3477 static int (*bpf_inode_storage_delete)(void *map, void *inode) = (void *) 146; 3478 3479 /* 3480 * bpf_d_path 3481 * 3482 * Return full path for given **struct path** object, which 3483 * needs to be the kernel BTF *path* object. The path is 3484 * returned in the provided buffer *buf* of size *sz* and 3485 * is zero terminated. 3486 * 3487 * 3488 * Returns 3489 * On success, the strictly positive length of the string, 3490 * including the trailing NUL character. On error, a negative 3491 * value. 3492 */ 3493 static long (*bpf_d_path)(struct path *path, char *buf, __u32 sz) = (void *) 147; 3494 3495 /* 3496 * bpf_copy_from_user 3497 * 3498 * Read *size* bytes from user space address *user_ptr* and store 3499 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 3500 * 3501 * Returns 3502 * 0 on success, or a negative error in case of failure. 3503 */ 3504 static long (*bpf_copy_from_user)(void *dst, __u32 size, const void *user_ptr) = (void *) 148; 3505 3506 /* 3507 * bpf_snprintf_btf 3508 * 3509 * Use BTF to store a string representation of *ptr*->ptr in *str*, 3510 * using *ptr*->type_id. This value should specify the type 3511 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 3512 * can be used to look up vmlinux BTF type ids. Traversing the 3513 * data structure using BTF, the type information and values are 3514 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 3515 * the pointer data is carried out to avoid kernel crashes during 3516 * operation. Smaller types can use string space on the stack; 3517 * larger programs can use map data to store the string 3518 * representation. 3519 * 3520 * The string can be subsequently shared with userspace via 3521 * bpf_perf_event_output() or ring buffer interfaces. 3522 * bpf_trace_printk() is to be avoided as it places too small 3523 * a limit on string size to be useful. 3524 * 3525 * *flags* is a combination of 3526 * 3527 * **BTF_F_COMPACT** 3528 * no formatting around type information 3529 * **BTF_F_NONAME** 3530 * no struct/union member names/types 3531 * **BTF_F_PTR_RAW** 3532 * show raw (unobfuscated) pointer values; 3533 * equivalent to printk specifier %px. 3534 * **BTF_F_ZERO** 3535 * show zero-valued struct/union members; they 3536 * are not displayed by default 3537 * 3538 * 3539 * Returns 3540 * The number of bytes that were written (or would have been 3541 * written if output had to be truncated due to string size), 3542 * or a negative error in cases of failure. 3543 */ 3544 static long (*bpf_snprintf_btf)(char *str, __u32 str_size, struct btf_ptr *ptr, __u32 btf_ptr_size, __u64 flags) = (void *) 149; 3545 3546 /* 3547 * bpf_seq_printf_btf 3548 * 3549 * Use BTF to write to seq_write a string representation of 3550 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 3551 * *flags* are identical to those used for bpf_snprintf_btf. 3552 * 3553 * Returns 3554 * 0 on success or a negative error in case of failure. 3555 */ 3556 static long (*bpf_seq_printf_btf)(struct seq_file *m, struct btf_ptr *ptr, __u32 ptr_size, __u64 flags) = (void *) 150; 3557 3558 /* 3559 * bpf_skb_cgroup_classid 3560 * 3561 * See **bpf_get_cgroup_classid**\ () for the main description. 3562 * This helper differs from **bpf_get_cgroup_classid**\ () in that 3563 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 3564 * associated socket instead of the current process. 3565 * 3566 * Returns 3567 * The id is returned or 0 in case the id could not be retrieved. 3568 */ 3569 static __u64 (*bpf_skb_cgroup_classid)(struct __sk_buff *skb) = (void *) 151; 3570 3571 /* 3572 * bpf_redirect_neigh 3573 * 3574 * Redirect the packet to another net device of index *ifindex* 3575 * and fill in L2 addresses from neighboring subsystem. This helper 3576 * is somewhat similar to **bpf_redirect**\ (), except that it 3577 * populates L2 addresses as well, meaning, internally, the helper 3578 * relies on the neighbor lookup for the L2 address of the nexthop. 3579 * 3580 * The helper will perform a FIB lookup based on the skb's 3581 * networking header to get the address of the next hop, unless 3582 * this is supplied by the caller in the *params* argument. The 3583 * *plen* argument indicates the len of *params* and should be set 3584 * to 0 if *params* is NULL. 3585 * 3586 * The *flags* argument is reserved and must be 0. The helper is 3587 * currently only supported for tc BPF program types, and enabled 3588 * for IPv4 and IPv6 protocols. 3589 * 3590 * Returns 3591 * The helper returns **TC_ACT_REDIRECT** on success or 3592 * **TC_ACT_SHOT** on error. 3593 */ 3594 static long (*bpf_redirect_neigh)(__u32 ifindex, struct bpf_redir_neigh *params, int plen, __u64 flags) = (void *) 152; 3595 3596 /* 3597 * bpf_per_cpu_ptr 3598 * 3599 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 3600 * pointer to the percpu kernel variable on *cpu*. A ksym is an 3601 * extern variable decorated with '__ksym'. For ksym, there is a 3602 * global var (either static or global) defined of the same name 3603 * in the kernel. The ksym is percpu if the global var is percpu. 3604 * The returned pointer points to the global percpu var on *cpu*. 3605 * 3606 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 3607 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 3608 * happens if *cpu* is larger than nr_cpu_ids. The caller of 3609 * bpf_per_cpu_ptr() must check the returned value. 3610 * 3611 * Returns 3612 * A pointer pointing to the kernel percpu variable on *cpu*, or 3613 * NULL, if *cpu* is invalid. 3614 */ 3615 static void *(*bpf_per_cpu_ptr)(const void *percpu_ptr, __u32 cpu) = (void *) 153; 3616 3617 /* 3618 * bpf_this_cpu_ptr 3619 * 3620 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 3621 * pointer to the percpu kernel variable on this cpu. See the 3622 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 3623 * 3624 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 3625 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 3626 * never return NULL. 3627 * 3628 * Returns 3629 * A pointer pointing to the kernel percpu variable on this cpu. 3630 */ 3631 static void *(*bpf_this_cpu_ptr)(const void *percpu_ptr) = (void *) 154; 3632 3633 /* 3634 * bpf_redirect_peer 3635 * 3636 * Redirect the packet to another net device of index *ifindex*. 3637 * This helper is somewhat similar to **bpf_redirect**\ (), except 3638 * that the redirection happens to the *ifindex*' peer device and 3639 * the netns switch takes place from ingress to ingress without 3640 * going through the CPU's backlog queue. 3641 * 3642 * The *flags* argument is reserved and must be 0. The helper is 3643 * currently only supported for tc BPF program types at the ingress 3644 * hook and for veth device types. The peer device must reside in a 3645 * different network namespace. 3646 * 3647 * Returns 3648 * The helper returns **TC_ACT_REDIRECT** on success or 3649 * **TC_ACT_SHOT** on error. 3650 */ 3651 static long (*bpf_redirect_peer)(__u32 ifindex, __u64 flags) = (void *) 155; 3652 3653 /* 3654 * bpf_task_storage_get 3655 * 3656 * Get a bpf_local_storage from the *task*. 3657 * 3658 * Logically, it could be thought of as getting the value from 3659 * a *map* with *task* as the **key**. From this 3660 * perspective, the usage is not much different from 3661 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 3662 * helper enforces the key must be an task_struct and the map must also 3663 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 3664 * 3665 * Underneath, the value is stored locally at *task* instead of 3666 * the *map*. The *map* is used as the bpf-local-storage 3667 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3668 * searched against all bpf_local_storage residing at *task*. 3669 * 3670 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 3671 * used such that a new bpf_local_storage will be 3672 * created if one does not exist. *value* can be used 3673 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 3674 * the initial value of a bpf_local_storage. If *value* is 3675 * **NULL**, the new bpf_local_storage will be zero initialized. 3676 * 3677 * Returns 3678 * A bpf_local_storage pointer is returned on success. 3679 * 3680 * **NULL** if not found or there was an error in adding 3681 * a new bpf_local_storage. 3682 */ 3683 static void *(*bpf_task_storage_get)(void *map, struct task_struct *task, void *value, __u64 flags) = (void *) 156; 3684 3685 /* 3686 * bpf_task_storage_delete 3687 * 3688 * Delete a bpf_local_storage from a *task*. 3689 * 3690 * Returns 3691 * 0 on success. 3692 * 3693 * **-ENOENT** if the bpf_local_storage cannot be found. 3694 */ 3695 static long (*bpf_task_storage_delete)(void *map, struct task_struct *task) = (void *) 157; 3696 3697 /* 3698 * bpf_get_current_task_btf 3699 * 3700 * Return a BTF pointer to the "current" task. 3701 * This pointer can also be used in helpers that accept an 3702 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 3703 * 3704 * Returns 3705 * Pointer to the current task. 3706 */ 3707 static struct task_struct *(*bpf_get_current_task_btf)(void) = (void *) 158; 3708 3709 /* 3710 * bpf_bprm_opts_set 3711 * 3712 * Set or clear certain options on *bprm*: 3713 * 3714 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 3715 * which sets the **AT_SECURE** auxv for glibc. The bit 3716 * is cleared if the flag is not specified. 3717 * 3718 * Returns 3719 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 3720 */ 3721 static long (*bpf_bprm_opts_set)(struct linux_binprm *bprm, __u64 flags) = (void *) 159; 3722 3723 /* 3724 * bpf_ktime_get_coarse_ns 3725 * 3726 * Return a coarse-grained version of the time elapsed since 3727 * system boot, in nanoseconds. Does not include time the system 3728 * was suspended. 3729 * 3730 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 3731 * 3732 * Returns 3733 * Current *ktime*. 3734 */ 3735 static __u64 (*bpf_ktime_get_coarse_ns)(void) = (void *) 160; 3736 3737 /* 3738 * bpf_ima_inode_hash 3739 * 3740 * Returns the stored IMA hash of the *inode* (if it's avaialable). 3741 * If the hash is larger than *size*, then only *size* 3742 * bytes will be copied to *dst* 3743 * 3744 * Returns 3745 * The **hash_algo** is returned on success, 3746 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 3747 * invalid arguments are passed. 3748 */ 3749 static long (*bpf_ima_inode_hash)(struct inode *inode, void *dst, __u32 size) = (void *) 161; 3750 3751 /* 3752 * bpf_sock_from_file 3753 * 3754 * If the given file represents a socket, returns the associated 3755 * socket. 3756 * 3757 * Returns 3758 * A pointer to a struct socket on success or NULL if the file is 3759 * not a socket. 3760 */ 3761 static struct socket *(*bpf_sock_from_file)(struct file *file) = (void *) 162; 3762 3763 /* 3764 * bpf_check_mtu 3765 * 3766 * Check packet size against exceeding MTU of net device (based 3767 * on *ifindex*). This helper will likely be used in combination 3768 * with helpers that adjust/change the packet size. 3769 * 3770 * The argument *len_diff* can be used for querying with a planned 3771 * size change. This allows to check MTU prior to changing packet 3772 * ctx. Providing an *len_diff* adjustment that is larger than the 3773 * actual packet size (resulting in negative packet size) will in 3774 * principle not exceed the MTU, why it is not considered a 3775 * failure. Other BPF-helpers are needed for performing the 3776 * planned size change, why the responsability for catch a negative 3777 * packet size belong in those helpers. 3778 * 3779 * Specifying *ifindex* zero means the MTU check is performed 3780 * against the current net device. This is practical if this isn't 3781 * used prior to redirect. 3782 * 3783 * On input *mtu_len* must be a valid pointer, else verifier will 3784 * reject BPF program. If the value *mtu_len* is initialized to 3785 * zero then the ctx packet size is use. When value *mtu_len* is 3786 * provided as input this specify the L3 length that the MTU check 3787 * is done against. Remember XDP and TC length operate at L2, but 3788 * this value is L3 as this correlate to MTU and IP-header tot_len 3789 * values which are L3 (similar behavior as bpf_fib_lookup). 3790 * 3791 * The Linux kernel route table can configure MTUs on a more 3792 * specific per route level, which is not provided by this helper. 3793 * For route level MTU checks use the **bpf_fib_lookup**\ () 3794 * helper. 3795 * 3796 * *ctx* is either **struct xdp_md** for XDP programs or 3797 * **struct sk_buff** for tc cls_act programs. 3798 * 3799 * The *flags* argument can be a combination of one or more of the 3800 * following values: 3801 * 3802 * **BPF_MTU_CHK_SEGS** 3803 * This flag will only works for *ctx* **struct sk_buff**. 3804 * If packet context contains extra packet segment buffers 3805 * (often knows as GSO skb), then MTU check is harder to 3806 * check at this point, because in transmit path it is 3807 * possible for the skb packet to get re-segmented 3808 * (depending on net device features). This could still be 3809 * a MTU violation, so this flag enables performing MTU 3810 * check against segments, with a different violation 3811 * return code to tell it apart. Check cannot use len_diff. 3812 * 3813 * On return *mtu_len* pointer contains the MTU value of the net 3814 * device. Remember the net device configured MTU is the L3 size, 3815 * which is returned here and XDP and TC length operate at L2. 3816 * Helper take this into account for you, but remember when using 3817 * MTU value in your BPF-code. 3818 * 3819 * 3820 * Returns 3821 * * 0 on success, and populate MTU value in *mtu_len* pointer. 3822 * 3823 * * < 0 if any input argument is invalid (*mtu_len* not updated) 3824 * 3825 * MTU violations return positive values, but also populate MTU 3826 * value in *mtu_len* pointer, as this can be needed for 3827 * implementing PMTU handing: 3828 * 3829 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 3830 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 3831 */ 3832 static long (*bpf_check_mtu)(void *ctx, __u32 ifindex, __u32 *mtu_len, __s32 len_diff, __u64 flags) = (void *) 163; 3833 3834 /* 3835 * bpf_for_each_map_elem 3836 * 3837 * For each element in **map**, call **callback_fn** function with 3838 * **map**, **callback_ctx** and other map-specific parameters. 3839 * The **callback_fn** should be a static function and 3840 * the **callback_ctx** should be a pointer to the stack. 3841 * The **flags** is used to control certain aspects of the helper. 3842 * Currently, the **flags** must be 0. 3843 * 3844 * The following are a list of supported map types and their 3845 * respective expected callback signatures: 3846 * 3847 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 3848 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 3849 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 3850 * 3851 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 3852 * 3853 * For per_cpu maps, the map_value is the value on the cpu where the 3854 * bpf_prog is running. 3855 * 3856 * If **callback_fn** return 0, the helper will continue to the next 3857 * element. If return value is 1, the helper will skip the rest of 3858 * elements and return. Other return values are not used now. 3859 * 3860 * 3861 * Returns 3862 * The number of traversed map elements for success, **-EINVAL** for 3863 * invalid **flags**. 3864 */ 3865 static long (*bpf_for_each_map_elem)(void *map, void *callback_fn, void *callback_ctx, __u64 flags) = (void *) 164; 3866 3867 /* 3868 * bpf_snprintf 3869 * 3870 * Outputs a string into the **str** buffer of size **str_size** 3871 * based on a format string stored in a read-only map pointed by 3872 * **fmt**. 3873 * 3874 * Each format specifier in **fmt** corresponds to one u64 element 3875 * in the **data** array. For strings and pointers where pointees 3876 * are accessed, only the pointer values are stored in the *data* 3877 * array. The *data_len* is the size of *data* in bytes - must be 3878 * a multiple of 8. 3879 * 3880 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 3881 * memory. Reading kernel memory may fail due to either invalid 3882 * address or valid address but requiring a major memory fault. If 3883 * reading kernel memory fails, the string for **%s** will be an 3884 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 3885 * Not returning error to bpf program is consistent with what 3886 * **bpf_trace_printk**\ () does for now. 3887 * 3888 * 3889 * Returns 3890 * The strictly positive length of the formatted string, including 3891 * the trailing zero character. If the return value is greater than 3892 * **str_size**, **str** contains a truncated string, guaranteed to 3893 * be zero-terminated except when **str_size** is 0. 3894 * 3895 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 3896 */ 3897 static long (*bpf_snprintf)(char *str, __u32 str_size, const char *fmt, __u64 *data, __u32 data_len) = (void *) 165; 3898 3899 /* 3900 * bpf_sys_bpf 3901 * 3902 * Execute bpf syscall with given arguments. 3903 * 3904 * Returns 3905 * A syscall result. 3906 */ 3907 static long (*bpf_sys_bpf)(__u32 cmd, void *attr, __u32 attr_size) = (void *) 166; 3908 3909 /* 3910 * bpf_btf_find_by_name_kind 3911 * 3912 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 3913 * 3914 * Returns 3915 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 3916 */ 3917 static long (*bpf_btf_find_by_name_kind)(char *name, int name_sz, __u32 kind, int flags) = (void *) 167; 3918 3919 /* 3920 * bpf_sys_close 3921 * 3922 * Execute close syscall for given FD. 3923 * 3924 * Returns 3925 * A syscall result. 3926 */ 3927 static long (*bpf_sys_close)(__u32 fd) = (void *) 168; 3928 3929 /* 3930 * bpf_timer_init 3931 * 3932 * Initialize the timer. 3933 * First 4 bits of *flags* specify clockid. 3934 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 3935 * All other bits of *flags* are reserved. 3936 * The verifier will reject the program if *timer* is not from 3937 * the same *map*. 3938 * 3939 * Returns 3940 * 0 on success. 3941 * **-EBUSY** if *timer* is already initialized. 3942 * **-EINVAL** if invalid *flags* are passed. 3943 * **-EPERM** if *timer* is in a map that doesn't have any user references. 3944 * The user space should either hold a file descriptor to a map with timers 3945 * or pin such map in bpffs. When map is unpinned or file descriptor is 3946 * closed all timers in the map will be cancelled and freed. 3947 */ 3948 static long (*bpf_timer_init)(struct bpf_timer *timer, void *map, __u64 flags) = (void *) 169; 3949 3950 /* 3951 * bpf_timer_set_callback 3952 * 3953 * Configure the timer to call *callback_fn* static function. 3954 * 3955 * Returns 3956 * 0 on success. 3957 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 3958 * **-EPERM** if *timer* is in a map that doesn't have any user references. 3959 * The user space should either hold a file descriptor to a map with timers 3960 * or pin such map in bpffs. When map is unpinned or file descriptor is 3961 * closed all timers in the map will be cancelled and freed. 3962 */ 3963 static long (*bpf_timer_set_callback)(struct bpf_timer *timer, void *callback_fn) = (void *) 170; 3964 3965 /* 3966 * bpf_timer_start 3967 * 3968 * Set timer expiration N nanoseconds from the current time. The 3969 * configured callback will be invoked in soft irq context on some cpu 3970 * and will not repeat unless another bpf_timer_start() is made. 3971 * In such case the next invocation can migrate to a different cpu. 3972 * Since struct bpf_timer is a field inside map element the map 3973 * owns the timer. The bpf_timer_set_callback() will increment refcnt 3974 * of BPF program to make sure that callback_fn code stays valid. 3975 * When user space reference to a map reaches zero all timers 3976 * in a map are cancelled and corresponding program's refcnts are 3977 * decremented. This is done to make sure that Ctrl-C of a user 3978 * process doesn't leave any timers running. If map is pinned in 3979 * bpffs the callback_fn can re-arm itself indefinitely. 3980 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 3981 * cancel and free the timer in the given map element. 3982 * The map can contain timers that invoke callback_fn-s from different 3983 * programs. The same callback_fn can serve different timers from 3984 * different maps if key/value layout matches across maps. 3985 * Every bpf_timer_set_callback() can have different callback_fn. 3986 * 3987 * 3988 * Returns 3989 * 0 on success. 3990 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 3991 * or invalid *flags* are passed. 3992 */ 3993 static long (*bpf_timer_start)(struct bpf_timer *timer, __u64 nsecs, __u64 flags) = (void *) 171; 3994 3995 /* 3996 * bpf_timer_cancel 3997 * 3998 * Cancel the timer and wait for callback_fn to finish if it was running. 3999 * 4000 * Returns 4001 * 0 if the timer was not active. 4002 * 1 if the timer was active. 4003 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4004 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 4005 * own timer which would have led to a deadlock otherwise. 4006 */ 4007 static long (*bpf_timer_cancel)(struct bpf_timer *timer) = (void *) 172; 4008 4009 /* 4010 * bpf_get_func_ip 4011 * 4012 * Get address of the traced function (for tracing and kprobe programs). 4013 * 4014 * Returns 4015 * Address of the traced function. 4016 */ 4017 static __u64 (*bpf_get_func_ip)(void *ctx) = (void *) 173; 4018 4019 /* 4020 * bpf_get_attach_cookie 4021 * 4022 * Get bpf_cookie value provided (optionally) during the program 4023 * attachment. It might be different for each individual 4024 * attachment, even if BPF program itself is the same. 4025 * Expects BPF program context *ctx* as a first argument. 4026 * 4027 * Supported for the following program types: 4028 * - kprobe/uprobe; 4029 * - tracepoint; 4030 * - perf_event. 4031 * 4032 * Returns 4033 * Value specified by user at BPF link creation/attachment time 4034 * or 0, if it was not specified. 4035 */ 4036 static __u64 (*bpf_get_attach_cookie)(void *ctx) = (void *) 174; 4037 4038 /* 4039 * bpf_task_pt_regs 4040 * 4041 * Get the struct pt_regs associated with **task**. 4042 * 4043 * Returns 4044 * A pointer to struct pt_regs. 4045 */ 4046 static long (*bpf_task_pt_regs)(struct task_struct *task) = (void *) 175; 4047 4048 /* 4049 * bpf_get_branch_snapshot 4050 * 4051 * Get branch trace from hardware engines like Intel LBR. The 4052 * hardware engine is stopped shortly after the helper is 4053 * called. Therefore, the user need to filter branch entries 4054 * based on the actual use case. To capture branch trace 4055 * before the trigger point of the BPF program, the helper 4056 * should be called at the beginning of the BPF program. 4057 * 4058 * The data is stored as struct perf_branch_entry into output 4059 * buffer *entries*. *size* is the size of *entries* in bytes. 4060 * *flags* is reserved for now and must be zero. 4061 * 4062 * 4063 * Returns 4064 * On success, number of bytes written to *buf*. On error, a 4065 * negative value. 4066 * 4067 * **-EINVAL** if *flags* is not zero. 4068 * 4069 * **-ENOENT** if architecture does not support branch records. 4070 */ 4071 static long (*bpf_get_branch_snapshot)(void *entries, __u32 size, __u64 flags) = (void *) 176; 4072 4073 /* 4074 * bpf_trace_vprintk 4075 * 4076 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 4077 * to format and can handle more format args as a result. 4078 * 4079 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 4080 * 4081 * Returns 4082 * The number of bytes written to the buffer, or a negative error 4083 * in case of failure. 4084 */ 4085 static long (*bpf_trace_vprintk)(const char *fmt, __u32 fmt_size, const void *data, __u32 data_len) = (void *) 177; 4086 4087 /* 4088 * bpf_skc_to_unix_sock 4089 * 4090 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 4091 * 4092 * Returns 4093 * *sk* if casting is valid, or **NULL** otherwise. 4094 */ 4095 static struct unix_sock *(*bpf_skc_to_unix_sock)(void *sk) = (void *) 178; 4096 4097 /* 4098 * bpf_kallsyms_lookup_name 4099 * 4100 * Get the address of a kernel symbol, returned in *res*. *res* is 4101 * set to 0 if the symbol is not found. 4102 * 4103 * Returns 4104 * On success, zero. On error, a negative value. 4105 * 4106 * **-EINVAL** if *flags* is not zero. 4107 * 4108 * **-EINVAL** if string *name* is not the same size as *name_sz*. 4109 * 4110 * **-ENOENT** if symbol is not found. 4111 * 4112 * **-EPERM** if caller does not have permission to obtain kernel address. 4113 */ 4114 static long (*bpf_kallsyms_lookup_name)(const char *name, int name_sz, int flags, __u64 *res) = (void *) 179; 4115 4116 /* 4117 * bpf_find_vma 4118 * 4119 * Find vma of *task* that contains *addr*, call *callback_fn* 4120 * function with *task*, *vma*, and *callback_ctx*. 4121 * The *callback_fn* should be a static function and 4122 * the *callback_ctx* should be a pointer to the stack. 4123 * The *flags* is used to control certain aspects of the helper. 4124 * Currently, the *flags* must be 0. 4125 * 4126 * The expected callback signature is 4127 * 4128 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 4129 * 4130 * 4131 * Returns 4132 * 0 on success. 4133 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 4134 * **-EBUSY** if failed to try lock mmap_lock. 4135 * **-EINVAL** for invalid **flags**. 4136 */ 4137 static long (*bpf_find_vma)(struct task_struct *task, __u64 addr, void *callback_fn, void *callback_ctx, __u64 flags) = (void *) 180; 4138 4139