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 cgroup; 33 struct __sk_buff; 34 struct sk_msg_md; 35 struct xdp_md; 36 struct path; 37 struct btf_ptr; 38 struct inode; 39 struct socket; 40 struct file; 41 struct bpf_timer; 42 struct mptcp_sock; 43 struct bpf_dynptr; 44 struct iphdr; 45 struct ipv6hdr; 46 47 /* 48 * bpf_map_lookup_elem 49 * 50 * Perform a lookup in *map* for an entry associated to *key*. 51 * 52 * Returns 53 * Map value associated to *key*, or **NULL** if no entry was 54 * found. 55 */ 56 static void *(* const bpf_map_lookup_elem)(void *map, const void *key) = (void *) 1; 57 58 /* 59 * bpf_map_update_elem 60 * 61 * Add or update the value of the entry associated to *key* in 62 * *map* with *value*. *flags* is one of: 63 * 64 * **BPF_NOEXIST** 65 * The entry for *key* must not exist in the map. 66 * **BPF_EXIST** 67 * The entry for *key* must already exist in the map. 68 * **BPF_ANY** 69 * No condition on the existence of the entry for *key*. 70 * 71 * Flag value **BPF_NOEXIST** cannot be used for maps of types 72 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 73 * elements always exist), the helper would return an error. 74 * 75 * Returns 76 * 0 on success, or a negative error in case of failure. 77 */ 78 static long (* const bpf_map_update_elem)(void *map, const void *key, const void *value, __u64 flags) = (void *) 2; 79 80 /* 81 * bpf_map_delete_elem 82 * 83 * Delete entry with *key* from *map*. 84 * 85 * Returns 86 * 0 on success, or a negative error in case of failure. 87 */ 88 static long (* const bpf_map_delete_elem)(void *map, const void *key) = (void *) 3; 89 90 /* 91 * bpf_probe_read 92 * 93 * For tracing programs, safely attempt to read *size* bytes from 94 * kernel space address *unsafe_ptr* and store the data in *dst*. 95 * 96 * Generally, use **bpf_probe_read_user**\ () or 97 * **bpf_probe_read_kernel**\ () instead. 98 * 99 * Returns 100 * 0 on success, or a negative error in case of failure. 101 */ 102 static long (* const bpf_probe_read)(void *dst, __u32 size, const void *unsafe_ptr) = (void *) 4; 103 104 /* 105 * bpf_ktime_get_ns 106 * 107 * Return the time elapsed since system boot, in nanoseconds. 108 * Does not include time the system was suspended. 109 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 110 * 111 * Returns 112 * Current *ktime*. 113 */ 114 static __u64 (* const bpf_ktime_get_ns)(void) = (void *) 5; 115 116 /* 117 * bpf_trace_printk 118 * 119 * This helper is a "printk()-like" facility for debugging. It 120 * prints a message defined by format *fmt* (of size *fmt_size*) 121 * to file *\/sys/kernel/tracing/trace* from TraceFS, if 122 * available. It can take up to three additional **u64** 123 * arguments (as an eBPF helpers, the total number of arguments is 124 * limited to five). 125 * 126 * Each time the helper is called, it appends a line to the trace. 127 * Lines are discarded while *\/sys/kernel/tracing/trace* is 128 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this. 129 * The format of the trace is customizable, and the exact output 130 * one will get depends on the options set in 131 * *\/sys/kernel/tracing/trace_options* (see also the 132 * *README* file under the same directory). However, it usually 133 * defaults to something like: 134 * 135 * :: 136 * 137 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 138 * 139 * In the above: 140 * 141 * * ``telnet`` is the name of the current task. 142 * * ``470`` is the PID of the current task. 143 * * ``001`` is the CPU number on which the task is 144 * running. 145 * * In ``.N..``, each character refers to a set of 146 * options (whether irqs are enabled, scheduling 147 * options, whether hard/softirqs are running, level of 148 * preempt_disabled respectively). **N** means that 149 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 150 * are set. 151 * * ``419421.045894`` is a timestamp. 152 * * ``0x00000001`` is a fake value used by BPF for the 153 * instruction pointer register. 154 * * ``<formatted msg>`` is the message formatted with 155 * *fmt*. 156 * 157 * The conversion specifiers supported by *fmt* are similar, but 158 * more limited than for printk(). They are **%d**, **%i**, 159 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 160 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 161 * of field, padding with zeroes, etc.) is available, and the 162 * helper will return **-EINVAL** (but print nothing) if it 163 * encounters an unknown specifier. 164 * 165 * Also, note that **bpf_trace_printk**\ () is slow, and should 166 * only be used for debugging purposes. For this reason, a notice 167 * block (spanning several lines) is printed to kernel logs and 168 * states that the helper should not be used "for production use" 169 * the first time this helper is used (or more precisely, when 170 * **trace_printk**\ () buffers are allocated). For passing values 171 * to user space, perf events should be preferred. 172 * 173 * Returns 174 * The number of bytes written to the buffer, or a negative error 175 * in case of failure. 176 */ 177 static long (* const bpf_trace_printk)(const char *fmt, __u32 fmt_size, ...) = (void *) 6; 178 179 /* 180 * bpf_get_prandom_u32 181 * 182 * Get a pseudo-random number. 183 * 184 * From a security point of view, this helper uses its own 185 * pseudo-random internal state, and cannot be used to infer the 186 * seed of other random functions in the kernel. However, it is 187 * essential to note that the generator used by the helper is not 188 * cryptographically secure. 189 * 190 * Returns 191 * A random 32-bit unsigned value. 192 */ 193 static __u32 (* const bpf_get_prandom_u32)(void) = (void *) 7; 194 195 /* 196 * bpf_get_smp_processor_id 197 * 198 * Get the SMP (symmetric multiprocessing) processor id. Note that 199 * all programs run with migration disabled, which means that the 200 * SMP processor id is stable during all the execution of the 201 * program. 202 * 203 * Returns 204 * The SMP id of the processor running the program. 205 */ 206 static __u32 (* const bpf_get_smp_processor_id)(void) = (void *) 8; 207 208 /* 209 * bpf_skb_store_bytes 210 * 211 * Store *len* bytes from address *from* into the packet 212 * associated to *skb*, at *offset*. *flags* are a combination of 213 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 214 * checksum for the packet after storing the bytes) and 215 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 216 * **->swhash** and *skb*\ **->l4hash** to 0). 217 * 218 * A call to this helper is susceptible to change the underlying 219 * packet buffer. Therefore, at load time, all checks on pointers 220 * previously done by the verifier are invalidated and must be 221 * performed again, if the helper is used in combination with 222 * direct packet access. 223 * 224 * Returns 225 * 0 on success, or a negative error in case of failure. 226 */ 227 static long (* const bpf_skb_store_bytes)(struct __sk_buff *skb, __u32 offset, const void *from, __u32 len, __u64 flags) = (void *) 9; 228 229 /* 230 * bpf_l3_csum_replace 231 * 232 * Recompute the layer 3 (e.g. IP) checksum for the packet 233 * associated to *skb*. Computation is incremental, so the helper 234 * must know the former value of the header field that was 235 * modified (*from*), the new value of this field (*to*), and the 236 * number of bytes (2 or 4) for this field, stored in *size*. 237 * Alternatively, it is possible to store the difference between 238 * the previous and the new values of the header field in *to*, by 239 * setting *from* and *size* to 0. For both methods, *offset* 240 * indicates the location of the IP checksum within the packet. 241 * 242 * This helper works in combination with **bpf_csum_diff**\ (), 243 * which does not update the checksum in-place, but offers more 244 * flexibility and can handle sizes larger than 2 or 4 for the 245 * checksum to update. 246 * 247 * A call to this helper is susceptible to change the underlying 248 * packet buffer. Therefore, at load time, all checks on pointers 249 * previously done by the verifier are invalidated and must be 250 * performed again, if the helper is used in combination with 251 * direct packet access. 252 * 253 * Returns 254 * 0 on success, or a negative error in case of failure. 255 */ 256 static long (* const bpf_l3_csum_replace)(struct __sk_buff *skb, __u32 offset, __u64 from, __u64 to, __u64 size) = (void *) 10; 257 258 /* 259 * bpf_l4_csum_replace 260 * 261 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 262 * packet associated to *skb*. Computation is incremental, so the 263 * helper must know the former value of the header field that was 264 * modified (*from*), the new value of this field (*to*), and the 265 * number of bytes (2 or 4) for this field, stored on the lowest 266 * four bits of *flags*. Alternatively, it is possible to store 267 * the difference between the previous and the new values of the 268 * header field in *to*, by setting *from* and the four lowest 269 * bits of *flags* to 0. For both methods, *offset* indicates the 270 * location of the IP checksum within the packet. In addition to 271 * the size of the field, *flags* can be added (bitwise OR) actual 272 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 273 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 274 * for updates resulting in a null checksum the value is set to 275 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 276 * the checksum is to be computed against a pseudo-header. 277 * 278 * This helper works in combination with **bpf_csum_diff**\ (), 279 * which does not update the checksum in-place, but offers more 280 * flexibility and can handle sizes larger than 2 or 4 for the 281 * checksum to update. 282 * 283 * A call to this helper is susceptible to change the underlying 284 * packet buffer. Therefore, at load time, all checks on pointers 285 * previously done by the verifier are invalidated and must be 286 * performed again, if the helper is used in combination with 287 * direct packet access. 288 * 289 * Returns 290 * 0 on success, or a negative error in case of failure. 291 */ 292 static long (* const bpf_l4_csum_replace)(struct __sk_buff *skb, __u32 offset, __u64 from, __u64 to, __u64 flags) = (void *) 11; 293 294 /* 295 * bpf_tail_call 296 * 297 * This special helper is used to trigger a "tail call", or in 298 * other words, to jump into another eBPF program. The same stack 299 * frame is used (but values on stack and in registers for the 300 * caller are not accessible to the callee). This mechanism allows 301 * for program chaining, either for raising the maximum number of 302 * available eBPF instructions, or to execute given programs in 303 * conditional blocks. For security reasons, there is an upper 304 * limit to the number of successive tail calls that can be 305 * performed. 306 * 307 * Upon call of this helper, the program attempts to jump into a 308 * program referenced at index *index* in *prog_array_map*, a 309 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 310 * *ctx*, a pointer to the context. 311 * 312 * If the call succeeds, the kernel immediately runs the first 313 * instruction of the new program. This is not a function call, 314 * and it never returns to the previous program. If the call 315 * fails, then the helper has no effect, and the caller continues 316 * to run its subsequent instructions. A call can fail if the 317 * destination program for the jump does not exist (i.e. *index* 318 * is superior to the number of entries in *prog_array_map*), or 319 * if the maximum number of tail calls has been reached for this 320 * chain of programs. This limit is defined in the kernel by the 321 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 322 * which is currently set to 33. 323 * 324 * Returns 325 * 0 on success, or a negative error in case of failure. 326 */ 327 static long (* const bpf_tail_call)(void *ctx, void *prog_array_map, __u32 index) = (void *) 12; 328 329 /* 330 * bpf_clone_redirect 331 * 332 * Clone and redirect the packet associated to *skb* to another 333 * net device of index *ifindex*. Both ingress and egress 334 * interfaces can be used for redirection. The **BPF_F_INGRESS** 335 * value in *flags* is used to make the distinction (ingress path 336 * is selected if the flag is present, egress path otherwise). 337 * This is the only flag supported for now. 338 * 339 * In comparison with **bpf_redirect**\ () helper, 340 * **bpf_clone_redirect**\ () has the associated cost of 341 * duplicating the packet buffer, but this can be executed out of 342 * the eBPF program. Conversely, **bpf_redirect**\ () is more 343 * efficient, but it is handled through an action code where the 344 * redirection happens only after the eBPF program has returned. 345 * 346 * A call to this helper is susceptible to change the underlying 347 * packet buffer. Therefore, at load time, all checks on pointers 348 * previously done by the verifier are invalidated and must be 349 * performed again, if the helper is used in combination with 350 * direct packet access. 351 * 352 * Returns 353 * 0 on success, or a negative error in case of failure. Positive 354 * error indicates a potential drop or congestion in the target 355 * device. The particular positive error codes are not defined. 356 */ 357 static long (* const bpf_clone_redirect)(struct __sk_buff *skb, __u32 ifindex, __u64 flags) = (void *) 13; 358 359 /* 360 * bpf_get_current_pid_tgid 361 * 362 * Get the current pid and tgid. 363 * 364 * Returns 365 * A 64-bit integer containing the current tgid and pid, and 366 * created as such: 367 * *current_task*\ **->tgid << 32 \|** 368 * *current_task*\ **->pid**. 369 */ 370 static __u64 (* const bpf_get_current_pid_tgid)(void) = (void *) 14; 371 372 /* 373 * bpf_get_current_uid_gid 374 * 375 * Get the current uid and gid. 376 * 377 * Returns 378 * A 64-bit integer containing the current GID and UID, and 379 * created as such: *current_gid* **<< 32 \|** *current_uid*. 380 */ 381 static __u64 (* const bpf_get_current_uid_gid)(void) = (void *) 15; 382 383 /* 384 * bpf_get_current_comm 385 * 386 * Copy the **comm** attribute of the current task into *buf* of 387 * *size_of_buf*. The **comm** attribute contains the name of 388 * the executable (excluding the path) for the current task. The 389 * *size_of_buf* must be strictly positive. On success, the 390 * helper makes sure that the *buf* is NUL-terminated. On failure, 391 * it is filled with zeroes. 392 * 393 * Returns 394 * 0 on success, or a negative error in case of failure. 395 */ 396 static long (* const bpf_get_current_comm)(void *buf, __u32 size_of_buf) = (void *) 16; 397 398 /* 399 * bpf_get_cgroup_classid 400 * 401 * Retrieve the classid for the current task, i.e. for the net_cls 402 * cgroup to which *skb* belongs. 403 * 404 * This helper can be used on TC egress path, but not on ingress. 405 * 406 * The net_cls cgroup provides an interface to tag network packets 407 * based on a user-provided identifier for all traffic coming from 408 * the tasks belonging to the related cgroup. See also the related 409 * kernel documentation, available from the Linux sources in file 410 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 411 * 412 * The Linux kernel has two versions for cgroups: there are 413 * cgroups v1 and cgroups v2. Both are available to users, who can 414 * use a mixture of them, but note that the net_cls cgroup is for 415 * cgroup v1 only. This makes it incompatible with BPF programs 416 * run on cgroups, which is a cgroup-v2-only feature (a socket can 417 * only hold data for one version of cgroups at a time). 418 * 419 * This helper is only available is the kernel was compiled with 420 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 421 * "**y**" or to "**m**". 422 * 423 * Returns 424 * The classid, or 0 for the default unconfigured classid. 425 */ 426 static __u32 (* const bpf_get_cgroup_classid)(struct __sk_buff *skb) = (void *) 17; 427 428 /* 429 * bpf_skb_vlan_push 430 * 431 * Push a *vlan_tci* (VLAN tag control information) of protocol 432 * *vlan_proto* to the packet associated to *skb*, then update 433 * the checksum. Note that if *vlan_proto* is different from 434 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 435 * be **ETH_P_8021Q**. 436 * 437 * A call to this helper is susceptible to change the underlying 438 * packet buffer. Therefore, at load time, all checks on pointers 439 * previously done by the verifier are invalidated and must be 440 * performed again, if the helper is used in combination with 441 * direct packet access. 442 * 443 * Returns 444 * 0 on success, or a negative error in case of failure. 445 */ 446 static long (* const bpf_skb_vlan_push)(struct __sk_buff *skb, __be16 vlan_proto, __u16 vlan_tci) = (void *) 18; 447 448 /* 449 * bpf_skb_vlan_pop 450 * 451 * Pop a VLAN header from the packet associated to *skb*. 452 * 453 * A call to this helper is susceptible to change the underlying 454 * packet buffer. Therefore, at load time, all checks on pointers 455 * previously done by the verifier are invalidated and must be 456 * performed again, if the helper is used in combination with 457 * direct packet access. 458 * 459 * Returns 460 * 0 on success, or a negative error in case of failure. 461 */ 462 static long (* const bpf_skb_vlan_pop)(struct __sk_buff *skb) = (void *) 19; 463 464 /* 465 * bpf_skb_get_tunnel_key 466 * 467 * Get tunnel metadata. This helper takes a pointer *key* to an 468 * empty **struct bpf_tunnel_key** of **size**, that will be 469 * filled with tunnel metadata for the packet associated to *skb*. 470 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 471 * indicates that the tunnel is based on IPv6 protocol instead of 472 * IPv4. 473 * 474 * The **struct bpf_tunnel_key** is an object that generalizes the 475 * principal parameters used by various tunneling protocols into a 476 * single struct. This way, it can be used to easily make a 477 * decision based on the contents of the encapsulation header, 478 * "summarized" in this struct. In particular, it holds the IP 479 * address of the remote end (IPv4 or IPv6, depending on the case) 480 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 481 * this struct exposes the *key*\ **->tunnel_id**, which is 482 * generally mapped to a VNI (Virtual Network Identifier), making 483 * it programmable together with the **bpf_skb_set_tunnel_key**\ 484 * () helper. 485 * 486 * Let's imagine that the following code is part of a program 487 * attached to the TC ingress interface, on one end of a GRE 488 * tunnel, and is supposed to filter out all messages coming from 489 * remote ends with IPv4 address other than 10.0.0.1: 490 * 491 * :: 492 * 493 * int ret; 494 * struct bpf_tunnel_key key = {}; 495 * 496 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 497 * if (ret < 0) 498 * return TC_ACT_SHOT; // drop packet 499 * 500 * if (key.remote_ipv4 != 0x0a000001) 501 * return TC_ACT_SHOT; // drop packet 502 * 503 * return TC_ACT_OK; // accept packet 504 * 505 * This interface can also be used with all encapsulation devices 506 * that can operate in "collect metadata" mode: instead of having 507 * one network device per specific configuration, the "collect 508 * metadata" mode only requires a single device where the 509 * configuration can be extracted from this helper. 510 * 511 * This can be used together with various tunnels such as VXLan, 512 * Geneve, GRE or IP in IP (IPIP). 513 * 514 * Returns 515 * 0 on success, or a negative error in case of failure. 516 */ 517 static long (* const bpf_skb_get_tunnel_key)(struct __sk_buff *skb, struct bpf_tunnel_key *key, __u32 size, __u64 flags) = (void *) 20; 518 519 /* 520 * bpf_skb_set_tunnel_key 521 * 522 * Populate tunnel metadata for packet associated to *skb.* The 523 * tunnel metadata is set to the contents of *key*, of *size*. The 524 * *flags* can be set to a combination of the following values: 525 * 526 * **BPF_F_TUNINFO_IPV6** 527 * Indicate that the tunnel is based on IPv6 protocol 528 * instead of IPv4. 529 * **BPF_F_ZERO_CSUM_TX** 530 * For IPv4 packets, add a flag to tunnel metadata 531 * indicating that checksum computation should be skipped 532 * and checksum set to zeroes. 533 * **BPF_F_DONT_FRAGMENT** 534 * Add a flag to tunnel metadata indicating that the 535 * packet should not be fragmented. 536 * **BPF_F_SEQ_NUMBER** 537 * Add a flag to tunnel metadata indicating that a 538 * sequence number should be added to tunnel header before 539 * sending the packet. This flag was added for GRE 540 * encapsulation, but might be used with other protocols 541 * as well in the future. 542 * **BPF_F_NO_TUNNEL_KEY** 543 * Add a flag to tunnel metadata indicating that no tunnel 544 * key should be set in the resulting tunnel header. 545 * 546 * Here is a typical usage on the transmit path: 547 * 548 * :: 549 * 550 * struct bpf_tunnel_key key; 551 * populate key ... 552 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 553 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 554 * 555 * See also the description of the **bpf_skb_get_tunnel_key**\ () 556 * helper for additional information. 557 * 558 * Returns 559 * 0 on success, or a negative error in case of failure. 560 */ 561 static long (* const bpf_skb_set_tunnel_key)(struct __sk_buff *skb, struct bpf_tunnel_key *key, __u32 size, __u64 flags) = (void *) 21; 562 563 /* 564 * bpf_perf_event_read 565 * 566 * Read the value of a perf event counter. This helper relies on a 567 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 568 * the perf event counter is selected when *map* is updated with 569 * perf event file descriptors. The *map* is an array whose size 570 * is the number of available CPUs, and each cell contains a value 571 * relative to one CPU. The value to retrieve is indicated by 572 * *flags*, that contains the index of the CPU to look up, masked 573 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 574 * **BPF_F_CURRENT_CPU** to indicate that the value for the 575 * current CPU should be retrieved. 576 * 577 * Note that before Linux 4.13, only hardware perf event can be 578 * retrieved. 579 * 580 * Also, be aware that the newer helper 581 * **bpf_perf_event_read_value**\ () is recommended over 582 * **bpf_perf_event_read**\ () in general. The latter has some ABI 583 * quirks where error and counter value are used as a return code 584 * (which is wrong to do since ranges may overlap). This issue is 585 * fixed with **bpf_perf_event_read_value**\ (), which at the same 586 * time provides more features over the **bpf_perf_event_read**\ 587 * () interface. Please refer to the description of 588 * **bpf_perf_event_read_value**\ () for details. 589 * 590 * Returns 591 * The value of the perf event counter read from the map, or a 592 * negative error code in case of failure. 593 */ 594 static __u64 (* const bpf_perf_event_read)(void *map, __u64 flags) = (void *) 22; 595 596 /* 597 * bpf_redirect 598 * 599 * Redirect the packet to another net device of index *ifindex*. 600 * This helper is somewhat similar to **bpf_clone_redirect**\ 601 * (), except that the packet is not cloned, which provides 602 * increased performance. 603 * 604 * Except for XDP, both ingress and egress interfaces can be used 605 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 606 * to make the distinction (ingress path is selected if the flag 607 * is present, egress path otherwise). Currently, XDP only 608 * supports redirection to the egress interface, and accepts no 609 * flag at all. 610 * 611 * The same effect can also be attained with the more generic 612 * **bpf_redirect_map**\ (), which uses a BPF map to store the 613 * redirect target instead of providing it directly to the helper. 614 * 615 * Returns 616 * For XDP, the helper returns **XDP_REDIRECT** on success or 617 * **XDP_ABORTED** on error. For other program types, the values 618 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 619 * error. 620 */ 621 static long (* const bpf_redirect)(__u32 ifindex, __u64 flags) = (void *) 23; 622 623 /* 624 * bpf_get_route_realm 625 * 626 * Retrieve the realm or the route, that is to say the 627 * **tclassid** field of the destination for the *skb*. The 628 * identifier retrieved is a user-provided tag, similar to the 629 * one used with the net_cls cgroup (see description for 630 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 631 * held by a route (a destination entry), not by a task. 632 * 633 * Retrieving this identifier works with the clsact TC egress hook 634 * (see also **tc-bpf(8)**), or alternatively on conventional 635 * classful egress qdiscs, but not on TC ingress path. In case of 636 * clsact TC egress hook, this has the advantage that, internally, 637 * the destination entry has not been dropped yet in the transmit 638 * path. Therefore, the destination entry does not need to be 639 * artificially held via **netif_keep_dst**\ () for a classful 640 * qdisc until the *skb* is freed. 641 * 642 * This helper is available only if the kernel was compiled with 643 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 644 * 645 * Returns 646 * The realm of the route for the packet associated to *skb*, or 0 647 * if none was found. 648 */ 649 static __u32 (* const bpf_get_route_realm)(struct __sk_buff *skb) = (void *) 24; 650 651 /* 652 * bpf_perf_event_output 653 * 654 * Write raw *data* blob into a special BPF perf event held by 655 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 656 * event must have the following attributes: **PERF_SAMPLE_RAW** 657 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 658 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 659 * 660 * The *flags* are used to indicate the index in *map* for which 661 * the value must be put, masked with **BPF_F_INDEX_MASK**. 662 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 663 * to indicate that the index of the current CPU core should be 664 * used. 665 * 666 * The value to write, of *size*, is passed through eBPF stack and 667 * pointed by *data*. 668 * 669 * The context of the program *ctx* needs also be passed to the 670 * helper. 671 * 672 * On user space, a program willing to read the values needs to 673 * call **perf_event_open**\ () on the perf event (either for 674 * one or for all CPUs) and to store the file descriptor into the 675 * *map*. This must be done before the eBPF program can send data 676 * into it. An example is available in file 677 * *samples/bpf/trace_output_user.c* in the Linux kernel source 678 * tree (the eBPF program counterpart is in 679 * *samples/bpf/trace_output_kern.c*). 680 * 681 * **bpf_perf_event_output**\ () achieves better performance 682 * than **bpf_trace_printk**\ () for sharing data with user 683 * space, and is much better suitable for streaming data from eBPF 684 * programs. 685 * 686 * Note that this helper is not restricted to tracing use cases 687 * and can be used with programs attached to TC or XDP as well, 688 * where it allows for passing data to user space listeners. Data 689 * can be: 690 * 691 * * Only custom structs, 692 * * Only the packet payload, or 693 * * A combination of both. 694 * 695 * Returns 696 * 0 on success, or a negative error in case of failure. 697 */ 698 static long (* const bpf_perf_event_output)(void *ctx, void *map, __u64 flags, void *data, __u64 size) = (void *) 25; 699 700 /* 701 * bpf_skb_load_bytes 702 * 703 * This helper was provided as an easy way to load data from a 704 * packet. It can be used to load *len* bytes from *offset* from 705 * the packet associated to *skb*, into the buffer pointed by 706 * *to*. 707 * 708 * Since Linux 4.7, usage of this helper has mostly been replaced 709 * by "direct packet access", enabling packet data to be 710 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 711 * pointing respectively to the first byte of packet data and to 712 * the byte after the last byte of packet data. However, it 713 * remains useful if one wishes to read large quantities of data 714 * at once from a packet into the eBPF stack. 715 * 716 * Returns 717 * 0 on success, or a negative error in case of failure. 718 */ 719 static long (* const bpf_skb_load_bytes)(const void *skb, __u32 offset, void *to, __u32 len) = (void *) 26; 720 721 /* 722 * bpf_get_stackid 723 * 724 * Walk a user or a kernel stack and return its id. To achieve 725 * this, the helper needs *ctx*, which is a pointer to the context 726 * on which the tracing program is executed, and a pointer to a 727 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 728 * 729 * The last argument, *flags*, holds the number of stack frames to 730 * skip (from 0 to 255), masked with 731 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 732 * a combination of the following flags: 733 * 734 * **BPF_F_USER_STACK** 735 * Collect a user space stack instead of a kernel stack. 736 * **BPF_F_FAST_STACK_CMP** 737 * Compare stacks by hash only. 738 * **BPF_F_REUSE_STACKID** 739 * If two different stacks hash into the same *stackid*, 740 * discard the old one. 741 * 742 * The stack id retrieved is a 32 bit long integer handle which 743 * can be further combined with other data (including other stack 744 * ids) and used as a key into maps. This can be useful for 745 * generating a variety of graphs (such as flame graphs or off-cpu 746 * graphs). 747 * 748 * For walking a stack, this helper is an improvement over 749 * **bpf_probe_read**\ (), which can be used with unrolled loops 750 * but is not efficient and consumes a lot of eBPF instructions. 751 * Instead, **bpf_get_stackid**\ () can collect up to 752 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 753 * this limit can be controlled with the **sysctl** program, and 754 * that it should be manually increased in order to profile long 755 * user stacks (such as stacks for Java programs). To do so, use: 756 * 757 * :: 758 * 759 * # sysctl kernel.perf_event_max_stack=<new value> 760 * 761 * Returns 762 * The positive or null stack id on success, or a negative error 763 * in case of failure. 764 */ 765 static long (* const bpf_get_stackid)(void *ctx, void *map, __u64 flags) = (void *) 27; 766 767 /* 768 * bpf_csum_diff 769 * 770 * Compute a checksum difference, from the raw buffer pointed by 771 * *from*, of length *from_size* (that must be a multiple of 4), 772 * towards the raw buffer pointed by *to*, of size *to_size* 773 * (same remark). An optional *seed* can be added to the value 774 * (this can be cascaded, the seed may come from a previous call 775 * to the helper). 776 * 777 * This is flexible enough to be used in several ways: 778 * 779 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 780 * checksum, it can be used when pushing new data. 781 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 782 * checksum, it can be used when removing data from a packet. 783 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 784 * can be used to compute a diff. Note that *from_size* and 785 * *to_size* do not need to be equal. 786 * 787 * This helper can be used in combination with 788 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 789 * which one can feed in the difference computed with 790 * **bpf_csum_diff**\ (). 791 * 792 * Returns 793 * The checksum result, or a negative error code in case of 794 * failure. 795 */ 796 static __s64 (* const bpf_csum_diff)(__be32 *from, __u32 from_size, __be32 *to, __u32 to_size, __wsum seed) = (void *) 28; 797 798 /* 799 * bpf_skb_get_tunnel_opt 800 * 801 * Retrieve tunnel options metadata for the packet associated to 802 * *skb*, and store the raw tunnel option data to the buffer *opt* 803 * of *size*. 804 * 805 * This helper can be used with encapsulation devices that can 806 * operate in "collect metadata" mode (please refer to the related 807 * note in the description of **bpf_skb_get_tunnel_key**\ () for 808 * more details). A particular example where this can be used is 809 * in combination with the Geneve encapsulation protocol, where it 810 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 811 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 812 * the eBPF program. This allows for full customization of these 813 * headers. 814 * 815 * Returns 816 * The size of the option data retrieved. 817 */ 818 static long (* const bpf_skb_get_tunnel_opt)(struct __sk_buff *skb, void *opt, __u32 size) = (void *) 29; 819 820 /* 821 * bpf_skb_set_tunnel_opt 822 * 823 * Set tunnel options metadata for the packet associated to *skb* 824 * to the option data contained in the raw buffer *opt* of *size*. 825 * 826 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 827 * helper for additional information. 828 * 829 * Returns 830 * 0 on success, or a negative error in case of failure. 831 */ 832 static long (* const bpf_skb_set_tunnel_opt)(struct __sk_buff *skb, void *opt, __u32 size) = (void *) 30; 833 834 /* 835 * bpf_skb_change_proto 836 * 837 * Change the protocol of the *skb* to *proto*. Currently 838 * supported are transition from IPv4 to IPv6, and from IPv6 to 839 * IPv4. The helper takes care of the groundwork for the 840 * transition, including resizing the socket buffer. The eBPF 841 * program is expected to fill the new headers, if any, via 842 * **skb_store_bytes**\ () and to recompute the checksums with 843 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 844 * (). The main case for this helper is to perform NAT64 845 * operations out of an eBPF program. 846 * 847 * Internally, the GSO type is marked as dodgy so that headers are 848 * checked and segments are recalculated by the GSO/GRO engine. 849 * The size for GSO target is adapted as well. 850 * 851 * All values for *flags* are reserved for future usage, and must 852 * be left at zero. 853 * 854 * A call to this helper is susceptible to change the underlying 855 * packet buffer. Therefore, at load time, all checks on pointers 856 * previously done by the verifier are invalidated and must be 857 * performed again, if the helper is used in combination with 858 * direct packet access. 859 * 860 * Returns 861 * 0 on success, or a negative error in case of failure. 862 */ 863 static long (* const bpf_skb_change_proto)(struct __sk_buff *skb, __be16 proto, __u64 flags) = (void *) 31; 864 865 /* 866 * bpf_skb_change_type 867 * 868 * Change the packet type for the packet associated to *skb*. This 869 * comes down to setting *skb*\ **->pkt_type** to *type*, except 870 * the eBPF program does not have a write access to *skb*\ 871 * **->pkt_type** beside this helper. Using a helper here allows 872 * for graceful handling of errors. 873 * 874 * The major use case is to change incoming *skb*s to 875 * **PACKET_HOST** in a programmatic way instead of having to 876 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 877 * example. 878 * 879 * Note that *type* only allows certain values. At this time, they 880 * are: 881 * 882 * **PACKET_HOST** 883 * Packet is for us. 884 * **PACKET_BROADCAST** 885 * Send packet to all. 886 * **PACKET_MULTICAST** 887 * Send packet to group. 888 * **PACKET_OTHERHOST** 889 * Send packet to someone else. 890 * 891 * Returns 892 * 0 on success, or a negative error in case of failure. 893 */ 894 static long (* const bpf_skb_change_type)(struct __sk_buff *skb, __u32 type) = (void *) 32; 895 896 /* 897 * bpf_skb_under_cgroup 898 * 899 * Check whether *skb* is a descendant of the cgroup2 held by 900 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 901 * 902 * Returns 903 * The return value depends on the result of the test, and can be: 904 * 905 * * 0, if the *skb* failed the cgroup2 descendant test. 906 * * 1, if the *skb* succeeded the cgroup2 descendant test. 907 * * A negative error code, if an error occurred. 908 */ 909 static long (* const bpf_skb_under_cgroup)(struct __sk_buff *skb, void *map, __u32 index) = (void *) 33; 910 911 /* 912 * bpf_get_hash_recalc 913 * 914 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 915 * not set, in particular if the hash was cleared due to mangling, 916 * recompute this hash. Later accesses to the hash can be done 917 * directly with *skb*\ **->hash**. 918 * 919 * Calling **bpf_set_hash_invalid**\ (), changing a packet 920 * prototype with **bpf_skb_change_proto**\ (), or calling 921 * **bpf_skb_store_bytes**\ () with the 922 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 923 * the hash and to trigger a new computation for the next call to 924 * **bpf_get_hash_recalc**\ (). 925 * 926 * Returns 927 * The 32-bit hash. 928 */ 929 static __u32 (* const bpf_get_hash_recalc)(struct __sk_buff *skb) = (void *) 34; 930 931 /* 932 * bpf_get_current_task 933 * 934 * Get the current task. 935 * 936 * Returns 937 * A pointer to the current task struct. 938 */ 939 static __u64 (* const bpf_get_current_task)(void) = (void *) 35; 940 941 /* 942 * bpf_probe_write_user 943 * 944 * Attempt in a safe way to write *len* bytes from the buffer 945 * *src* to *dst* in memory. It only works for threads that are in 946 * user context, and *dst* must be a valid user space address. 947 * 948 * This helper should not be used to implement any kind of 949 * security mechanism because of TOC-TOU attacks, but rather to 950 * debug, divert, and manipulate execution of semi-cooperative 951 * processes. 952 * 953 * Keep in mind that this feature is meant for experiments, and it 954 * has a risk of crashing the system and running programs. 955 * Therefore, when an eBPF program using this helper is attached, 956 * a warning including PID and process name is printed to kernel 957 * logs. 958 * 959 * Returns 960 * 0 on success, or a negative error in case of failure. 961 */ 962 static long (* const bpf_probe_write_user)(void *dst, const void *src, __u32 len) = (void *) 36; 963 964 /* 965 * bpf_current_task_under_cgroup 966 * 967 * Check whether the probe is being run is the context of a given 968 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 969 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 970 * 971 * Returns 972 * The return value depends on the result of the test, and can be: 973 * 974 * * 1, if current task belongs to the cgroup2. 975 * * 0, if current task does not belong to the cgroup2. 976 * * A negative error code, if an error occurred. 977 */ 978 static long (* const bpf_current_task_under_cgroup)(void *map, __u32 index) = (void *) 37; 979 980 /* 981 * bpf_skb_change_tail 982 * 983 * Resize (trim or grow) the packet associated to *skb* to the 984 * new *len*. The *flags* are reserved for future usage, and must 985 * be left at zero. 986 * 987 * The basic idea is that the helper performs the needed work to 988 * change the size of the packet, then the eBPF program rewrites 989 * the rest via helpers like **bpf_skb_store_bytes**\ (), 990 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 991 * and others. This helper is a slow path utility intended for 992 * replies with control messages. And because it is targeted for 993 * slow path, the helper itself can afford to be slow: it 994 * implicitly linearizes, unclones and drops offloads from the 995 * *skb*. 996 * 997 * A call to this helper is susceptible to change the underlying 998 * packet buffer. Therefore, at load time, all checks on pointers 999 * previously done by the verifier are invalidated and must be 1000 * performed again, if the helper is used in combination with 1001 * direct packet access. 1002 * 1003 * Returns 1004 * 0 on success, or a negative error in case of failure. 1005 */ 1006 static long (* const bpf_skb_change_tail)(struct __sk_buff *skb, __u32 len, __u64 flags) = (void *) 38; 1007 1008 /* 1009 * bpf_skb_pull_data 1010 * 1011 * Pull in non-linear data in case the *skb* is non-linear and not 1012 * all of *len* are part of the linear section. Make *len* bytes 1013 * from *skb* readable and writable. If a zero value is passed for 1014 * *len*, then all bytes in the linear part of *skb* will be made 1015 * readable and writable. 1016 * 1017 * This helper is only needed for reading and writing with direct 1018 * packet access. 1019 * 1020 * For direct packet access, testing that offsets to access 1021 * are within packet boundaries (test on *skb*\ **->data_end**) is 1022 * susceptible to fail if offsets are invalid, or if the requested 1023 * data is in non-linear parts of the *skb*. On failure the 1024 * program can just bail out, or in the case of a non-linear 1025 * buffer, use a helper to make the data available. The 1026 * **bpf_skb_load_bytes**\ () helper is a first solution to access 1027 * the data. Another one consists in using **bpf_skb_pull_data** 1028 * to pull in once the non-linear parts, then retesting and 1029 * eventually access the data. 1030 * 1031 * At the same time, this also makes sure the *skb* is uncloned, 1032 * which is a necessary condition for direct write. As this needs 1033 * to be an invariant for the write part only, the verifier 1034 * detects writes and adds a prologue that is calling 1035 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1036 * the very beginning in case it is indeed cloned. 1037 * 1038 * A call to this helper is susceptible to change the underlying 1039 * packet buffer. Therefore, at load time, all checks on pointers 1040 * previously done by the verifier are invalidated and must be 1041 * performed again, if the helper is used in combination with 1042 * direct packet access. 1043 * 1044 * Returns 1045 * 0 on success, or a negative error in case of failure. 1046 */ 1047 static long (* const bpf_skb_pull_data)(struct __sk_buff *skb, __u32 len) = (void *) 39; 1048 1049 /* 1050 * bpf_csum_update 1051 * 1052 * Add the checksum *csum* into *skb*\ **->csum** in case the 1053 * driver has supplied a checksum for the entire packet into that 1054 * field. Return an error otherwise. This helper is intended to be 1055 * used in combination with **bpf_csum_diff**\ (), in particular 1056 * when the checksum needs to be updated after data has been 1057 * written into the packet through direct packet access. 1058 * 1059 * Returns 1060 * The checksum on success, or a negative error code in case of 1061 * failure. 1062 */ 1063 static __s64 (* const bpf_csum_update)(struct __sk_buff *skb, __wsum csum) = (void *) 40; 1064 1065 /* 1066 * bpf_set_hash_invalid 1067 * 1068 * Invalidate the current *skb*\ **->hash**. It can be used after 1069 * mangling on headers through direct packet access, in order to 1070 * indicate that the hash is outdated and to trigger a 1071 * recalculation the next time the kernel tries to access this 1072 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1073 * 1074 * Returns 1075 * void. 1076 */ 1077 static void (* const bpf_set_hash_invalid)(struct __sk_buff *skb) = (void *) 41; 1078 1079 /* 1080 * bpf_get_numa_node_id 1081 * 1082 * Return the id of the current NUMA node. The primary use case 1083 * for this helper is the selection of sockets for the local NUMA 1084 * node, when the program is attached to sockets using the 1085 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1086 * but the helper is also available to other eBPF program types, 1087 * similarly to **bpf_get_smp_processor_id**\ (). 1088 * 1089 * Returns 1090 * The id of current NUMA node. 1091 */ 1092 static long (* const bpf_get_numa_node_id)(void) = (void *) 42; 1093 1094 /* 1095 * bpf_skb_change_head 1096 * 1097 * Grows headroom of packet associated to *skb* and adjusts the 1098 * offset of the MAC header accordingly, adding *len* bytes of 1099 * space. It automatically extends and reallocates memory as 1100 * required. 1101 * 1102 * This helper can be used on a layer 3 *skb* to push a MAC header 1103 * for redirection into a layer 2 device. 1104 * 1105 * All values for *flags* are reserved for future usage, and must 1106 * be left at zero. 1107 * 1108 * A call to this helper is susceptible to change the underlying 1109 * packet buffer. Therefore, at load time, all checks on pointers 1110 * previously done by the verifier are invalidated and must be 1111 * performed again, if the helper is used in combination with 1112 * direct packet access. 1113 * 1114 * Returns 1115 * 0 on success, or a negative error in case of failure. 1116 */ 1117 static long (* const bpf_skb_change_head)(struct __sk_buff *skb, __u32 len, __u64 flags) = (void *) 43; 1118 1119 /* 1120 * bpf_xdp_adjust_head 1121 * 1122 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1123 * it is possible to use a negative value for *delta*. This helper 1124 * can be used to prepare the packet for pushing or popping 1125 * headers. 1126 * 1127 * A call to this helper is susceptible to change the underlying 1128 * packet buffer. Therefore, at load time, all checks on pointers 1129 * previously done by the verifier are invalidated and must be 1130 * performed again, if the helper is used in combination with 1131 * direct packet access. 1132 * 1133 * Returns 1134 * 0 on success, or a negative error in case of failure. 1135 */ 1136 static long (* const bpf_xdp_adjust_head)(struct xdp_md *xdp_md, int delta) = (void *) 44; 1137 1138 /* 1139 * bpf_probe_read_str 1140 * 1141 * Copy a NUL terminated string from an unsafe kernel address 1142 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 1143 * more details. 1144 * 1145 * Generally, use **bpf_probe_read_user_str**\ () or 1146 * **bpf_probe_read_kernel_str**\ () instead. 1147 * 1148 * Returns 1149 * On success, the strictly positive length of the string, 1150 * including the trailing NUL character. On error, a negative 1151 * value. 1152 */ 1153 static long (* const bpf_probe_read_str)(void *dst, __u32 size, const void *unsafe_ptr) = (void *) 45; 1154 1155 /* 1156 * bpf_get_socket_cookie 1157 * 1158 * If the **struct sk_buff** pointed by *skb* has a known socket, 1159 * retrieve the cookie (generated by the kernel) of this socket. 1160 * If no cookie has been set yet, generate a new cookie. Once 1161 * generated, the socket cookie remains stable for the life of the 1162 * socket. This helper can be useful for monitoring per socket 1163 * networking traffic statistics as it provides a global socket 1164 * identifier that can be assumed unique. 1165 * 1166 * Returns 1167 * A 8-byte long unique number on success, or 0 if the socket 1168 * field is missing inside *skb*. 1169 */ 1170 static __u64 (* const bpf_get_socket_cookie)(void *ctx) = (void *) 46; 1171 1172 /* 1173 * bpf_get_socket_uid 1174 * 1175 * Get the owner UID of the socked associated to *skb*. 1176 * 1177 * Returns 1178 * The owner UID of the socket associated to *skb*. If the socket 1179 * is **NULL**, or if it is not a full socket (i.e. if it is a 1180 * time-wait or a request socket instead), **overflowuid** value 1181 * is returned (note that **overflowuid** might also be the actual 1182 * UID value for the socket). 1183 */ 1184 static __u32 (* const bpf_get_socket_uid)(struct __sk_buff *skb) = (void *) 47; 1185 1186 /* 1187 * bpf_set_hash 1188 * 1189 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1190 * to value *hash*. 1191 * 1192 * Returns 1193 * 0 1194 */ 1195 static long (* const bpf_set_hash)(struct __sk_buff *skb, __u32 hash) = (void *) 48; 1196 1197 /* 1198 * bpf_setsockopt 1199 * 1200 * Emulate a call to **setsockopt()** on the socket associated to 1201 * *bpf_socket*, which must be a full socket. The *level* at 1202 * which the option resides and the name *optname* of the option 1203 * must be specified, see **setsockopt(2)** for more information. 1204 * The option value of length *optlen* is pointed by *optval*. 1205 * 1206 * *bpf_socket* should be one of the following: 1207 * 1208 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1209 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 1210 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 1211 * 1212 * This helper actually implements a subset of **setsockopt()**. 1213 * It supports the following *level*\ s: 1214 * 1215 * * **SOL_SOCKET**, which supports the following *optname*\ s: 1216 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1217 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 1218 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**, 1219 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**. 1220 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 1221 * **TCP_CONGESTION**, **TCP_BPF_IW**, 1222 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 1223 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 1224 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**, 1225 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**, 1226 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**, 1227 * **TCP_BPF_RTO_MIN**. 1228 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 1229 * * **IPPROTO_IPV6**, which supports the following *optname*\ s: 1230 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**. 1231 * 1232 * Returns 1233 * 0 on success, or a negative error in case of failure. 1234 */ 1235 static long (* const bpf_setsockopt)(void *bpf_socket, int level, int optname, void *optval, int optlen) = (void *) 49; 1236 1237 /* 1238 * bpf_skb_adjust_room 1239 * 1240 * Grow or shrink the room for data in the packet associated to 1241 * *skb* by *len_diff*, and according to the selected *mode*. 1242 * 1243 * By default, the helper will reset any offloaded checksum 1244 * indicator of the skb to CHECKSUM_NONE. This can be avoided 1245 * by the following flag: 1246 * 1247 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 1248 * checksum data of the skb to CHECKSUM_NONE. 1249 * 1250 * There are two supported modes at this time: 1251 * 1252 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 1253 * (room space is added or removed between the layer 2 and 1254 * layer 3 headers). 1255 * 1256 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1257 * (room space is added or removed between the layer 3 and 1258 * layer 4 headers). 1259 * 1260 * The following flags are supported at this time: 1261 * 1262 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 1263 * Adjusting mss in this way is not allowed for datagrams. 1264 * 1265 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 1266 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1267 * Any new space is reserved to hold a tunnel header. 1268 * Configure skb offsets and other fields accordingly. 1269 * 1270 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 1271 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1272 * Use with ENCAP_L3 flags to further specify the tunnel type. 1273 * 1274 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 1275 * Use with ENCAP_L3/L4 flags to further specify the tunnel 1276 * type; *len* is the length of the inner MAC header. 1277 * 1278 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 1279 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 1280 * L2 type as Ethernet. 1281 * 1282 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**, 1283 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**: 1284 * Indicate the new IP header version after decapsulating the outer 1285 * IP header. Used when the inner and outer IP versions are different. 1286 * 1287 * A call to this helper is susceptible to change the underlying 1288 * packet buffer. Therefore, at load time, all checks on pointers 1289 * previously done by the verifier are invalidated and must be 1290 * performed again, if the helper is used in combination with 1291 * direct packet access. 1292 * 1293 * Returns 1294 * 0 on success, or a negative error in case of failure. 1295 */ 1296 static long (* const bpf_skb_adjust_room)(struct __sk_buff *skb, __s32 len_diff, __u32 mode, __u64 flags) = (void *) 50; 1297 1298 /* 1299 * bpf_redirect_map 1300 * 1301 * Redirect the packet to the endpoint referenced by *map* at 1302 * index *key*. Depending on its type, this *map* can contain 1303 * references to net devices (for forwarding packets through other 1304 * ports), or to CPUs (for redirecting XDP frames to another CPU; 1305 * but this is only implemented for native XDP (with driver 1306 * support) as of this writing). 1307 * 1308 * The lower two bits of *flags* are used as the return code if 1309 * the map lookup fails. This is so that the return value can be 1310 * one of the XDP program return codes up to **XDP_TX**, as chosen 1311 * by the caller. The higher bits of *flags* can be set to 1312 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 1313 * 1314 * With BPF_F_BROADCAST the packet will be broadcasted to all the 1315 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 1316 * interface will be excluded when do broadcasting. 1317 * 1318 * See also **bpf_redirect**\ (), which only supports redirecting 1319 * to an ifindex, but doesn't require a map to do so. 1320 * 1321 * Returns 1322 * **XDP_REDIRECT** on success, or the value of the two lower bits 1323 * of the *flags* argument on error. 1324 */ 1325 static long (* const bpf_redirect_map)(void *map, __u64 key, __u64 flags) = (void *) 51; 1326 1327 /* 1328 * bpf_sk_redirect_map 1329 * 1330 * Redirect the packet to the socket referenced by *map* (of type 1331 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1332 * egress interfaces can be used for redirection. The 1333 * **BPF_F_INGRESS** value in *flags* is used to make the 1334 * distinction (ingress path is selected if the flag is present, 1335 * egress path otherwise). This is the only flag supported for now. 1336 * 1337 * Returns 1338 * **SK_PASS** on success, or **SK_DROP** on error. 1339 */ 1340 static long (* const bpf_sk_redirect_map)(struct __sk_buff *skb, void *map, __u32 key, __u64 flags) = (void *) 52; 1341 1342 /* 1343 * bpf_sock_map_update 1344 * 1345 * Add an entry to, or update a *map* referencing sockets. The 1346 * *skops* is used as a new value for the entry associated to 1347 * *key*. *flags* is one of: 1348 * 1349 * **BPF_NOEXIST** 1350 * The entry for *key* must not exist in the map. 1351 * **BPF_EXIST** 1352 * The entry for *key* must already exist in the map. 1353 * **BPF_ANY** 1354 * No condition on the existence of the entry for *key*. 1355 * 1356 * If the *map* has eBPF programs (parser and verdict), those will 1357 * be inherited by the socket being added. If the socket is 1358 * already attached to eBPF programs, this results in an error. 1359 * 1360 * Returns 1361 * 0 on success, or a negative error in case of failure. 1362 */ 1363 static long (* const bpf_sock_map_update)(struct bpf_sock_ops *skops, void *map, void *key, __u64 flags) = (void *) 53; 1364 1365 /* 1366 * bpf_xdp_adjust_meta 1367 * 1368 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1369 * *delta* (which can be positive or negative). Note that this 1370 * operation modifies the address stored in *xdp_md*\ **->data**, 1371 * so the latter must be loaded only after the helper has been 1372 * called. 1373 * 1374 * The use of *xdp_md*\ **->data_meta** is optional and programs 1375 * are not required to use it. The rationale is that when the 1376 * packet is processed with XDP (e.g. as DoS filter), it is 1377 * possible to push further meta data along with it before passing 1378 * to the stack, and to give the guarantee that an ingress eBPF 1379 * program attached as a TC classifier on the same device can pick 1380 * this up for further post-processing. Since TC works with socket 1381 * buffers, it remains possible to set from XDP the **mark** or 1382 * **priority** pointers, or other pointers for the socket buffer. 1383 * Having this scratch space generic and programmable allows for 1384 * more flexibility as the user is free to store whatever meta 1385 * data they need. 1386 * 1387 * A call to this helper is susceptible to change the underlying 1388 * packet buffer. Therefore, at load time, all checks on pointers 1389 * previously done by the verifier are invalidated and must be 1390 * performed again, if the helper is used in combination with 1391 * direct packet access. 1392 * 1393 * Returns 1394 * 0 on success, or a negative error in case of failure. 1395 */ 1396 static long (* const bpf_xdp_adjust_meta)(struct xdp_md *xdp_md, int delta) = (void *) 54; 1397 1398 /* 1399 * bpf_perf_event_read_value 1400 * 1401 * Read the value of a perf event counter, and store it into *buf* 1402 * of size *buf_size*. This helper relies on a *map* of type 1403 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 1404 * counter is selected when *map* is updated with perf event file 1405 * descriptors. The *map* is an array whose size is the number of 1406 * available CPUs, and each cell contains a value relative to one 1407 * CPU. The value to retrieve is indicated by *flags*, that 1408 * contains the index of the CPU to look up, masked with 1409 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1410 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1411 * current CPU should be retrieved. 1412 * 1413 * This helper behaves in a way close to 1414 * **bpf_perf_event_read**\ () helper, save that instead of 1415 * just returning the value observed, it fills the *buf* 1416 * structure. This allows for additional data to be retrieved: in 1417 * particular, the enabled and running times (in *buf*\ 1418 * **->enabled** and *buf*\ **->running**, respectively) are 1419 * copied. In general, **bpf_perf_event_read_value**\ () is 1420 * recommended over **bpf_perf_event_read**\ (), which has some 1421 * ABI issues and provides fewer functionalities. 1422 * 1423 * These values are interesting, because hardware PMU (Performance 1424 * Monitoring Unit) counters are limited resources. When there are 1425 * more PMU based perf events opened than available counters, 1426 * kernel will multiplex these events so each event gets certain 1427 * percentage (but not all) of the PMU time. In case that 1428 * multiplexing happens, the number of samples or counter value 1429 * will not reflect the case compared to when no multiplexing 1430 * occurs. This makes comparison between different runs difficult. 1431 * Typically, the counter value should be normalized before 1432 * comparing to other experiments. The usual normalization is done 1433 * as follows. 1434 * 1435 * :: 1436 * 1437 * normalized_counter = counter * t_enabled / t_running 1438 * 1439 * Where t_enabled is the time enabled for event and t_running is 1440 * the time running for event since last normalization. The 1441 * enabled and running times are accumulated since the perf event 1442 * open. To achieve scaling factor between two invocations of an 1443 * eBPF program, users can use CPU id as the key (which is 1444 * typical for perf array usage model) to remember the previous 1445 * value and do the calculation inside the eBPF program. 1446 * 1447 * Returns 1448 * 0 on success, or a negative error in case of failure. 1449 */ 1450 static long (* const bpf_perf_event_read_value)(void *map, __u64 flags, struct bpf_perf_event_value *buf, __u32 buf_size) = (void *) 55; 1451 1452 /* 1453 * bpf_perf_prog_read_value 1454 * 1455 * For an eBPF program attached to a perf event, retrieve the 1456 * value of the event counter associated to *ctx* and store it in 1457 * the structure pointed by *buf* and of size *buf_size*. Enabled 1458 * and running times are also stored in the structure (see 1459 * description of helper **bpf_perf_event_read_value**\ () for 1460 * more details). 1461 * 1462 * Returns 1463 * 0 on success, or a negative error in case of failure. 1464 */ 1465 static long (* const bpf_perf_prog_read_value)(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, __u32 buf_size) = (void *) 56; 1466 1467 /* 1468 * bpf_getsockopt 1469 * 1470 * Emulate a call to **getsockopt()** on the socket associated to 1471 * *bpf_socket*, which must be a full socket. The *level* at 1472 * which the option resides and the name *optname* of the option 1473 * must be specified, see **getsockopt(2)** for more information. 1474 * The retrieved value is stored in the structure pointed by 1475 * *opval* and of length *optlen*. 1476 * 1477 * *bpf_socket* should be one of the following: 1478 * 1479 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1480 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 1481 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 1482 * 1483 * This helper actually implements a subset of **getsockopt()**. 1484 * It supports the same set of *optname*\ s that is supported by 1485 * the **bpf_setsockopt**\ () helper. The exceptions are 1486 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and 1487 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only. 1488 * 1489 * Returns 1490 * 0 on success, or a negative error in case of failure. 1491 */ 1492 static long (* const bpf_getsockopt)(void *bpf_socket, int level, int optname, void *optval, int optlen) = (void *) 57; 1493 1494 /* 1495 * bpf_override_return 1496 * 1497 * Used for error injection, this helper uses kprobes to override 1498 * the return value of the probed function, and to set it to *rc*. 1499 * The first argument is the context *regs* on which the kprobe 1500 * works. 1501 * 1502 * This helper works by setting the PC (program counter) 1503 * to an override function which is run in place of the original 1504 * probed function. This means the probed function is not run at 1505 * all. The replacement function just returns with the required 1506 * value. 1507 * 1508 * This helper has security implications, and thus is subject to 1509 * restrictions. It is only available if the kernel was compiled 1510 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 1511 * option, and in this case it only works on functions tagged with 1512 * **ALLOW_ERROR_INJECTION** in the kernel code. 1513 * 1514 * Also, the helper is only available for the architectures having 1515 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 1516 * x86 architecture is the only one to support this feature. 1517 * 1518 * Returns 1519 * 0 1520 */ 1521 static long (* const bpf_override_return)(struct pt_regs *regs, __u64 rc) = (void *) 58; 1522 1523 /* 1524 * bpf_sock_ops_cb_flags_set 1525 * 1526 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 1527 * for the full TCP socket associated to *bpf_sock_ops* to 1528 * *argval*. 1529 * 1530 * The primary use of this field is to determine if there should 1531 * be calls to eBPF programs of type 1532 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 1533 * code. A program of the same type can change its value, per 1534 * connection and as necessary, when the connection is 1535 * established. This field is directly accessible for reading, but 1536 * this helper must be used for updates in order to return an 1537 * error if an eBPF program tries to set a callback that is not 1538 * supported in the current kernel. 1539 * 1540 * *argval* is a flag array which can combine these flags: 1541 * 1542 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 1543 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 1544 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 1545 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 1546 * 1547 * Therefore, this function can be used to clear a callback flag by 1548 * setting the appropriate bit to zero. e.g. to disable the RTO 1549 * callback: 1550 * 1551 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1552 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1553 * 1554 * Here are some examples of where one could call such eBPF 1555 * program: 1556 * 1557 * * When RTO fires. 1558 * * When a packet is retransmitted. 1559 * * When the connection terminates. 1560 * * When a packet is sent. 1561 * * When a packet is received. 1562 * 1563 * Returns 1564 * Code **-EINVAL** if the socket is not a full TCP socket; 1565 * otherwise, a positive number containing the bits that could not 1566 * be set is returned (which comes down to 0 if all bits were set 1567 * as required). 1568 */ 1569 static long (* const bpf_sock_ops_cb_flags_set)(struct bpf_sock_ops *bpf_sock, int argval) = (void *) 59; 1570 1571 /* 1572 * bpf_msg_redirect_map 1573 * 1574 * This helper is used in programs implementing policies at the 1575 * socket level. If the message *msg* is allowed to pass (i.e. if 1576 * the verdict eBPF program returns **SK_PASS**), redirect it to 1577 * the socket referenced by *map* (of type 1578 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1579 * egress interfaces can be used for redirection. The 1580 * **BPF_F_INGRESS** value in *flags* is used to make the 1581 * distinction (ingress path is selected if the flag is present, 1582 * egress path otherwise). This is the only flag supported for now. 1583 * 1584 * Returns 1585 * **SK_PASS** on success, or **SK_DROP** on error. 1586 */ 1587 static long (* const bpf_msg_redirect_map)(struct sk_msg_md *msg, void *map, __u32 key, __u64 flags) = (void *) 60; 1588 1589 /* 1590 * bpf_msg_apply_bytes 1591 * 1592 * For socket policies, apply the verdict of the eBPF program to 1593 * the next *bytes* (number of bytes) of message *msg*. 1594 * 1595 * For example, this helper can be used in the following cases: 1596 * 1597 * * A single **sendmsg**\ () or **sendfile**\ () system call 1598 * contains multiple logical messages that the eBPF program is 1599 * supposed to read and for which it should apply a verdict. 1600 * * An eBPF program only cares to read the first *bytes* of a 1601 * *msg*. If the message has a large payload, then setting up 1602 * and calling the eBPF program repeatedly for all bytes, even 1603 * though the verdict is already known, would create unnecessary 1604 * overhead. 1605 * 1606 * When called from within an eBPF program, the helper sets a 1607 * counter internal to the BPF infrastructure, that is used to 1608 * apply the last verdict to the next *bytes*. If *bytes* is 1609 * smaller than the current data being processed from a 1610 * **sendmsg**\ () or **sendfile**\ () system call, the first 1611 * *bytes* will be sent and the eBPF program will be re-run with 1612 * the pointer for start of data pointing to byte number *bytes* 1613 * **+ 1**. If *bytes* is larger than the current data being 1614 * processed, then the eBPF verdict will be applied to multiple 1615 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1616 * consumed. 1617 * 1618 * Note that if a socket closes with the internal counter holding 1619 * a non-zero value, this is not a problem because data is not 1620 * being buffered for *bytes* and is sent as it is received. 1621 * 1622 * Returns 1623 * 0 1624 */ 1625 static long (* const bpf_msg_apply_bytes)(struct sk_msg_md *msg, __u32 bytes) = (void *) 61; 1626 1627 /* 1628 * bpf_msg_cork_bytes 1629 * 1630 * For socket policies, prevent the execution of the verdict eBPF 1631 * program for message *msg* until *bytes* (byte number) have been 1632 * accumulated. 1633 * 1634 * This can be used when one needs a specific number of bytes 1635 * before a verdict can be assigned, even if the data spans 1636 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1637 * case would be a user calling **sendmsg**\ () repeatedly with 1638 * 1-byte long message segments. Obviously, this is bad for 1639 * performance, but it is still valid. If the eBPF program needs 1640 * *bytes* bytes to validate a header, this helper can be used to 1641 * prevent the eBPF program to be called again until *bytes* have 1642 * been accumulated. 1643 * 1644 * Returns 1645 * 0 1646 */ 1647 static long (* const bpf_msg_cork_bytes)(struct sk_msg_md *msg, __u32 bytes) = (void *) 62; 1648 1649 /* 1650 * bpf_msg_pull_data 1651 * 1652 * For socket policies, pull in non-linear data from user space 1653 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1654 * **->data_end** to *start* and *end* bytes offsets into *msg*, 1655 * respectively. 1656 * 1657 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1658 * *msg* it can only parse data that the (**data**, **data_end**) 1659 * pointers have already consumed. For **sendmsg**\ () hooks this 1660 * is likely the first scatterlist element. But for calls relying 1661 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1662 * be the range (**0**, **0**) because the data is shared with 1663 * user space and by default the objective is to avoid allowing 1664 * user space to modify data while (or after) eBPF verdict is 1665 * being decided. This helper can be used to pull in data and to 1666 * set the start and end pointer to given values. Data will be 1667 * copied if necessary (i.e. if data was not linear and if start 1668 * and end pointers do not point to the same chunk). 1669 * 1670 * A call to this helper is susceptible to change the underlying 1671 * packet buffer. Therefore, at load time, all checks on pointers 1672 * previously done by the verifier are invalidated and must be 1673 * performed again, if the helper is used in combination with 1674 * direct packet access. 1675 * 1676 * All values for *flags* are reserved for future usage, and must 1677 * be left at zero. 1678 * 1679 * Returns 1680 * 0 on success, or a negative error in case of failure. 1681 */ 1682 static long (* const bpf_msg_pull_data)(struct sk_msg_md *msg, __u32 start, __u32 end, __u64 flags) = (void *) 63; 1683 1684 /* 1685 * bpf_bind 1686 * 1687 * Bind the socket associated to *ctx* to the address pointed by 1688 * *addr*, of length *addr_len*. This allows for making outgoing 1689 * connection from the desired IP address, which can be useful for 1690 * example when all processes inside a cgroup should use one 1691 * single IP address on a host that has multiple IP configured. 1692 * 1693 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 1694 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 1695 * **AF_INET6**). It's advised to pass zero port (**sin_port** 1696 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 1697 * behavior and lets the kernel efficiently pick up an unused 1698 * port as long as 4-tuple is unique. Passing non-zero port might 1699 * lead to degraded performance. 1700 * 1701 * Returns 1702 * 0 on success, or a negative error in case of failure. 1703 */ 1704 static long (* const bpf_bind)(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) = (void *) 64; 1705 1706 /* 1707 * bpf_xdp_adjust_tail 1708 * 1709 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 1710 * possible to both shrink and grow the packet tail. 1711 * Shrink done via *delta* being a negative integer. 1712 * 1713 * A call to this helper is susceptible to change the underlying 1714 * packet buffer. Therefore, at load time, all checks on pointers 1715 * previously done by the verifier are invalidated and must be 1716 * performed again, if the helper is used in combination with 1717 * direct packet access. 1718 * 1719 * Returns 1720 * 0 on success, or a negative error in case of failure. 1721 */ 1722 static long (* const bpf_xdp_adjust_tail)(struct xdp_md *xdp_md, int delta) = (void *) 65; 1723 1724 /* 1725 * bpf_skb_get_xfrm_state 1726 * 1727 * Retrieve the XFRM state (IP transform framework, see also 1728 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 1729 * 1730 * The retrieved value is stored in the **struct bpf_xfrm_state** 1731 * pointed by *xfrm_state* and of length *size*. 1732 * 1733 * All values for *flags* are reserved for future usage, and must 1734 * be left at zero. 1735 * 1736 * This helper is available only if the kernel was compiled with 1737 * **CONFIG_XFRM** configuration option. 1738 * 1739 * Returns 1740 * 0 on success, or a negative error in case of failure. 1741 */ 1742 static long (* const bpf_skb_get_xfrm_state)(struct __sk_buff *skb, __u32 index, struct bpf_xfrm_state *xfrm_state, __u32 size, __u64 flags) = (void *) 66; 1743 1744 /* 1745 * bpf_get_stack 1746 * 1747 * Return a user or a kernel stack in bpf program provided buffer. 1748 * To achieve this, the helper needs *ctx*, which is a pointer 1749 * to the context on which the tracing program is executed. 1750 * To store the stacktrace, the bpf program provides *buf* with 1751 * a nonnegative *size*. 1752 * 1753 * The last argument, *flags*, holds the number of stack frames to 1754 * skip (from 0 to 255), masked with 1755 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1756 * the following flags: 1757 * 1758 * **BPF_F_USER_STACK** 1759 * Collect a user space stack instead of a kernel stack. 1760 * **BPF_F_USER_BUILD_ID** 1761 * Collect (build_id, file_offset) instead of ips for user 1762 * stack, only valid if **BPF_F_USER_STACK** is also 1763 * specified. 1764 * 1765 * *file_offset* is an offset relative to the beginning 1766 * of the executable or shared object file backing the vma 1767 * which the *ip* falls in. It is *not* an offset relative 1768 * to that object's base address. Accordingly, it must be 1769 * adjusted by adding (sh_addr - sh_offset), where 1770 * sh_{addr,offset} correspond to the executable section 1771 * containing *file_offset* in the object, for comparisons 1772 * to symbols' st_value to be valid. 1773 * 1774 * **bpf_get_stack**\ () can collect up to 1775 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 1776 * to sufficient large buffer size. Note that 1777 * this limit can be controlled with the **sysctl** program, and 1778 * that it should be manually increased in order to profile long 1779 * user stacks (such as stacks for Java programs). To do so, use: 1780 * 1781 * :: 1782 * 1783 * # sysctl kernel.perf_event_max_stack=<new value> 1784 * 1785 * Returns 1786 * The non-negative copied *buf* length equal to or less than 1787 * *size* on success, or a negative error in case of failure. 1788 */ 1789 static long (* const bpf_get_stack)(void *ctx, void *buf, __u32 size, __u64 flags) = (void *) 67; 1790 1791 /* 1792 * bpf_skb_load_bytes_relative 1793 * 1794 * This helper is similar to **bpf_skb_load_bytes**\ () in that 1795 * it provides an easy way to load *len* bytes from *offset* 1796 * from the packet associated to *skb*, into the buffer pointed 1797 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 1798 * a fifth argument *start_header* exists in order to select a 1799 * base offset to start from. *start_header* can be one of: 1800 * 1801 * **BPF_HDR_START_MAC** 1802 * Base offset to load data from is *skb*'s mac header. 1803 * **BPF_HDR_START_NET** 1804 * Base offset to load data from is *skb*'s network header. 1805 * 1806 * In general, "direct packet access" is the preferred method to 1807 * access packet data, however, this helper is in particular useful 1808 * in socket filters where *skb*\ **->data** does not always point 1809 * to the start of the mac header and where "direct packet access" 1810 * is not available. 1811 * 1812 * Returns 1813 * 0 on success, or a negative error in case of failure. 1814 */ 1815 static long (* const bpf_skb_load_bytes_relative)(const void *skb, __u32 offset, void *to, __u32 len, __u32 start_header) = (void *) 68; 1816 1817 /* 1818 * bpf_fib_lookup 1819 * 1820 * Do FIB lookup in kernel tables using parameters in *params*. 1821 * If lookup is successful and result shows packet is to be 1822 * forwarded, the neighbor tables are searched for the nexthop. 1823 * If successful (ie., FIB lookup shows forwarding and nexthop 1824 * is resolved), the nexthop address is returned in ipv4_dst 1825 * or ipv6_dst based on family, smac is set to mac address of 1826 * egress device, dmac is set to nexthop mac address, rt_metric 1827 * is set to metric from route (IPv4/IPv6 only), and ifindex 1828 * is set to the device index of the nexthop from the FIB lookup. 1829 * 1830 * *plen* argument is the size of the passed in struct. 1831 * *flags* argument can be a combination of one or more of the 1832 * following values: 1833 * 1834 * **BPF_FIB_LOOKUP_DIRECT** 1835 * Do a direct table lookup vs full lookup using FIB 1836 * rules. 1837 * **BPF_FIB_LOOKUP_TBID** 1838 * Used with BPF_FIB_LOOKUP_DIRECT. 1839 * Use the routing table ID present in *params*->tbid 1840 * for the fib lookup. 1841 * **BPF_FIB_LOOKUP_OUTPUT** 1842 * Perform lookup from an egress perspective (default is 1843 * ingress). 1844 * **BPF_FIB_LOOKUP_SKIP_NEIGH** 1845 * Skip the neighbour table lookup. *params*->dmac 1846 * and *params*->smac will not be set as output. A common 1847 * use case is to call **bpf_redirect_neigh**\ () after 1848 * doing **bpf_fib_lookup**\ (). 1849 * **BPF_FIB_LOOKUP_SRC** 1850 * Derive and set source IP addr in *params*->ipv{4,6}_src 1851 * for the nexthop. If the src addr cannot be derived, 1852 * **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this 1853 * case, *params*->dmac and *params*->smac are not set either. 1854 * 1855 * *ctx* is either **struct xdp_md** for XDP programs or 1856 * **struct sk_buff** tc cls_act programs. 1857 * 1858 * Returns 1859 * * < 0 if any input argument is invalid 1860 * * 0 on success (packet is forwarded, nexthop neighbor exists) 1861 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 1862 * packet is not forwarded or needs assist from full stack 1863 * 1864 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 1865 * was exceeded and output params->mtu_result contains the MTU. 1866 */ 1867 static long (* const bpf_fib_lookup)(void *ctx, struct bpf_fib_lookup *params, int plen, __u32 flags) = (void *) 69; 1868 1869 /* 1870 * bpf_sock_hash_update 1871 * 1872 * Add an entry to, or update a sockhash *map* referencing sockets. 1873 * The *skops* is used as a new value for the entry associated to 1874 * *key*. *flags* is one of: 1875 * 1876 * **BPF_NOEXIST** 1877 * The entry for *key* must not exist in the map. 1878 * **BPF_EXIST** 1879 * The entry for *key* must already exist in the map. 1880 * **BPF_ANY** 1881 * No condition on the existence of the entry for *key*. 1882 * 1883 * If the *map* has eBPF programs (parser and verdict), those will 1884 * be inherited by the socket being added. If the socket is 1885 * already attached to eBPF programs, this results in an error. 1886 * 1887 * Returns 1888 * 0 on success, or a negative error in case of failure. 1889 */ 1890 static long (* const bpf_sock_hash_update)(struct bpf_sock_ops *skops, void *map, void *key, __u64 flags) = (void *) 70; 1891 1892 /* 1893 * bpf_msg_redirect_hash 1894 * 1895 * This helper is used in programs implementing policies at the 1896 * socket level. If the message *msg* is allowed to pass (i.e. if 1897 * the verdict eBPF program returns **SK_PASS**), redirect it to 1898 * the socket referenced by *map* (of type 1899 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 1900 * egress interfaces can be used for redirection. The 1901 * **BPF_F_INGRESS** value in *flags* is used to make the 1902 * distinction (ingress path is selected if the flag is present, 1903 * egress path otherwise). This is the only flag supported for now. 1904 * 1905 * Returns 1906 * **SK_PASS** on success, or **SK_DROP** on error. 1907 */ 1908 static long (* const bpf_msg_redirect_hash)(struct sk_msg_md *msg, void *map, void *key, __u64 flags) = (void *) 71; 1909 1910 /* 1911 * bpf_sk_redirect_hash 1912 * 1913 * This helper is used in programs implementing policies at the 1914 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 1915 * if the verdict eBPF program returns **SK_PASS**), redirect it 1916 * to the socket referenced by *map* (of type 1917 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 1918 * egress interfaces can be used for redirection. The 1919 * **BPF_F_INGRESS** value in *flags* is used to make the 1920 * distinction (ingress path is selected if the flag is present, 1921 * egress otherwise). This is the only flag supported for now. 1922 * 1923 * Returns 1924 * **SK_PASS** on success, or **SK_DROP** on error. 1925 */ 1926 static long (* const bpf_sk_redirect_hash)(struct __sk_buff *skb, void *map, void *key, __u64 flags) = (void *) 72; 1927 1928 /* 1929 * bpf_lwt_push_encap 1930 * 1931 * Encapsulate the packet associated to *skb* within a Layer 3 1932 * protocol header. This header is provided in the buffer at 1933 * address *hdr*, with *len* its size in bytes. *type* indicates 1934 * the protocol of the header and can be one of: 1935 * 1936 * **BPF_LWT_ENCAP_SEG6** 1937 * IPv6 encapsulation with Segment Routing Header 1938 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 1939 * the IPv6 header is computed by the kernel. 1940 * **BPF_LWT_ENCAP_SEG6_INLINE** 1941 * Only works if *skb* contains an IPv6 packet. Insert a 1942 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 1943 * the IPv6 header. 1944 * **BPF_LWT_ENCAP_IP** 1945 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 1946 * must be IPv4 or IPv6, followed by zero or more 1947 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 1948 * total bytes in all prepended headers. Please note that 1949 * if **skb_is_gso**\ (*skb*) is true, no more than two 1950 * headers can be prepended, and the inner header, if 1951 * present, should be either GRE or UDP/GUE. 1952 * 1953 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 1954 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 1955 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 1956 * **BPF_PROG_TYPE_LWT_XMIT**. 1957 * 1958 * A call to this helper is susceptible to change the underlying 1959 * packet buffer. Therefore, at load time, all checks on pointers 1960 * previously done by the verifier are invalidated and must be 1961 * performed again, if the helper is used in combination with 1962 * direct packet access. 1963 * 1964 * Returns 1965 * 0 on success, or a negative error in case of failure. 1966 */ 1967 static long (* const bpf_lwt_push_encap)(struct __sk_buff *skb, __u32 type, void *hdr, __u32 len) = (void *) 73; 1968 1969 /* 1970 * bpf_lwt_seg6_store_bytes 1971 * 1972 * Store *len* bytes from address *from* into the packet 1973 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 1974 * inside the outermost IPv6 Segment Routing Header can be 1975 * modified through this helper. 1976 * 1977 * A call to this helper is susceptible to change the underlying 1978 * packet buffer. Therefore, at load time, all checks on pointers 1979 * previously done by the verifier are invalidated and must be 1980 * performed again, if the helper is used in combination with 1981 * direct packet access. 1982 * 1983 * Returns 1984 * 0 on success, or a negative error in case of failure. 1985 */ 1986 static long (* const bpf_lwt_seg6_store_bytes)(struct __sk_buff *skb, __u32 offset, const void *from, __u32 len) = (void *) 74; 1987 1988 /* 1989 * bpf_lwt_seg6_adjust_srh 1990 * 1991 * Adjust the size allocated to TLVs in the outermost IPv6 1992 * Segment Routing Header contained in the packet associated to 1993 * *skb*, at position *offset* by *delta* bytes. Only offsets 1994 * after the segments are accepted. *delta* can be as well 1995 * positive (growing) as negative (shrinking). 1996 * 1997 * A call to this helper is susceptible to change the underlying 1998 * packet buffer. Therefore, at load time, all checks on pointers 1999 * previously done by the verifier are invalidated and must be 2000 * performed again, if the helper is used in combination with 2001 * direct packet access. 2002 * 2003 * Returns 2004 * 0 on success, or a negative error in case of failure. 2005 */ 2006 static long (* const bpf_lwt_seg6_adjust_srh)(struct __sk_buff *skb, __u32 offset, __s32 delta) = (void *) 75; 2007 2008 /* 2009 * bpf_lwt_seg6_action 2010 * 2011 * Apply an IPv6 Segment Routing action of type *action* to the 2012 * packet associated to *skb*. Each action takes a parameter 2013 * contained at address *param*, and of length *param_len* bytes. 2014 * *action* can be one of: 2015 * 2016 * **SEG6_LOCAL_ACTION_END_X** 2017 * End.X action: Endpoint with Layer-3 cross-connect. 2018 * Type of *param*: **struct in6_addr**. 2019 * **SEG6_LOCAL_ACTION_END_T** 2020 * End.T action: Endpoint with specific IPv6 table lookup. 2021 * Type of *param*: **int**. 2022 * **SEG6_LOCAL_ACTION_END_B6** 2023 * End.B6 action: Endpoint bound to an SRv6 policy. 2024 * Type of *param*: **struct ipv6_sr_hdr**. 2025 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2026 * End.B6.Encap action: Endpoint bound to an SRv6 2027 * encapsulation policy. 2028 * Type of *param*: **struct ipv6_sr_hdr**. 2029 * 2030 * A call to this helper is susceptible to change the underlying 2031 * packet buffer. Therefore, at load time, all checks on pointers 2032 * previously done by the verifier are invalidated and must be 2033 * performed again, if the helper is used in combination with 2034 * direct packet access. 2035 * 2036 * Returns 2037 * 0 on success, or a negative error in case of failure. 2038 */ 2039 static long (* const bpf_lwt_seg6_action)(struct __sk_buff *skb, __u32 action, void *param, __u32 param_len) = (void *) 76; 2040 2041 /* 2042 * bpf_rc_repeat 2043 * 2044 * This helper is used in programs implementing IR decoding, to 2045 * report a successfully decoded repeat key message. This delays 2046 * the generation of a key up event for previously generated 2047 * key down event. 2048 * 2049 * Some IR protocols like NEC have a special IR message for 2050 * repeating last button, for when a button is held down. 2051 * 2052 * The *ctx* should point to the lirc sample as passed into 2053 * the program. 2054 * 2055 * This helper is only available is the kernel was compiled with 2056 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2057 * "**y**". 2058 * 2059 * Returns 2060 * 0 2061 */ 2062 static long (* const bpf_rc_repeat)(void *ctx) = (void *) 77; 2063 2064 /* 2065 * bpf_rc_keydown 2066 * 2067 * This helper is used in programs implementing IR decoding, to 2068 * report a successfully decoded key press with *scancode*, 2069 * *toggle* value in the given *protocol*. The scancode will be 2070 * translated to a keycode using the rc keymap, and reported as 2071 * an input key down event. After a period a key up event is 2072 * generated. This period can be extended by calling either 2073 * **bpf_rc_keydown**\ () again with the same values, or calling 2074 * **bpf_rc_repeat**\ (). 2075 * 2076 * Some protocols include a toggle bit, in case the button was 2077 * released and pressed again between consecutive scancodes. 2078 * 2079 * The *ctx* should point to the lirc sample as passed into 2080 * the program. 2081 * 2082 * The *protocol* is the decoded protocol number (see 2083 * **enum rc_proto** for some predefined values). 2084 * 2085 * This helper is only available is the kernel was compiled with 2086 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2087 * "**y**". 2088 * 2089 * Returns 2090 * 0 2091 */ 2092 static long (* const bpf_rc_keydown)(void *ctx, __u32 protocol, __u64 scancode, __u32 toggle) = (void *) 78; 2093 2094 /* 2095 * bpf_skb_cgroup_id 2096 * 2097 * Return the cgroup v2 id of the socket associated with the *skb*. 2098 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2099 * helper for cgroup v1 by providing a tag resp. identifier that 2100 * can be matched on or used for map lookups e.g. to implement 2101 * policy. The cgroup v2 id of a given path in the hierarchy is 2102 * exposed in user space through the f_handle API in order to get 2103 * to the same 64-bit id. 2104 * 2105 * This helper can be used on TC egress path, but not on ingress, 2106 * and is available only if the kernel was compiled with the 2107 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2108 * 2109 * Returns 2110 * The id is returned or 0 in case the id could not be retrieved. 2111 */ 2112 static __u64 (* const bpf_skb_cgroup_id)(struct __sk_buff *skb) = (void *) 79; 2113 2114 /* 2115 * bpf_get_current_cgroup_id 2116 * 2117 * Get the current cgroup id based on the cgroup within which 2118 * the current task is running. 2119 * 2120 * Returns 2121 * A 64-bit integer containing the current cgroup id based 2122 * on the cgroup within which the current task is running. 2123 */ 2124 static __u64 (* const bpf_get_current_cgroup_id)(void) = (void *) 80; 2125 2126 /* 2127 * bpf_get_local_storage 2128 * 2129 * Get the pointer to the local storage area. 2130 * The type and the size of the local storage is defined 2131 * by the *map* argument. 2132 * The *flags* meaning is specific for each map type, 2133 * and has to be 0 for cgroup local storage. 2134 * 2135 * Depending on the BPF program type, a local storage area 2136 * can be shared between multiple instances of the BPF program, 2137 * running simultaneously. 2138 * 2139 * A user should care about the synchronization by himself. 2140 * For example, by using the **BPF_ATOMIC** instructions to alter 2141 * the shared data. 2142 * 2143 * Returns 2144 * A pointer to the local storage area. 2145 */ 2146 static void *(* const bpf_get_local_storage)(void *map, __u64 flags) = (void *) 81; 2147 2148 /* 2149 * bpf_sk_select_reuseport 2150 * 2151 * Select a **SO_REUSEPORT** socket from a 2152 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 2153 * It checks the selected socket is matching the incoming 2154 * request in the socket buffer. 2155 * 2156 * Returns 2157 * 0 on success, or a negative error in case of failure. 2158 */ 2159 static long (* const bpf_sk_select_reuseport)(struct sk_reuseport_md *reuse, void *map, void *key, __u64 flags) = (void *) 82; 2160 2161 /* 2162 * bpf_skb_ancestor_cgroup_id 2163 * 2164 * Return id of cgroup v2 that is ancestor of cgroup associated 2165 * with the *skb* at the *ancestor_level*. The root cgroup is at 2166 * *ancestor_level* zero and each step down the hierarchy 2167 * increments the level. If *ancestor_level* == level of cgroup 2168 * associated with *skb*, then return value will be same as that 2169 * of **bpf_skb_cgroup_id**\ (). 2170 * 2171 * The helper is useful to implement policies based on cgroups 2172 * that are upper in hierarchy than immediate cgroup associated 2173 * with *skb*. 2174 * 2175 * The format of returned id and helper limitations are same as in 2176 * **bpf_skb_cgroup_id**\ (). 2177 * 2178 * Returns 2179 * The id is returned or 0 in case the id could not be retrieved. 2180 */ 2181 static __u64 (* const bpf_skb_ancestor_cgroup_id)(struct __sk_buff *skb, int ancestor_level) = (void *) 83; 2182 2183 /* 2184 * bpf_sk_lookup_tcp 2185 * 2186 * Look for TCP socket matching *tuple*, optionally in a child 2187 * network namespace *netns*. The return value must be checked, 2188 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2189 * 2190 * The *ctx* should point to the context of the program, such as 2191 * the skb or socket (depending on the hook in use). This is used 2192 * to determine the base network namespace for the lookup. 2193 * 2194 * *tuple_size* must be one of: 2195 * 2196 * **sizeof**\ (*tuple*\ **->ipv4**) 2197 * Look for an IPv4 socket. 2198 * **sizeof**\ (*tuple*\ **->ipv6**) 2199 * Look for an IPv6 socket. 2200 * 2201 * If the *netns* is a negative signed 32-bit integer, then the 2202 * socket lookup table in the netns associated with the *ctx* 2203 * will be used. For the TC hooks, this is the netns of the device 2204 * in the skb. For socket hooks, this is the netns of the socket. 2205 * If *netns* is any other signed 32-bit value greater than or 2206 * equal to zero then it specifies the ID of the netns relative to 2207 * the netns associated with the *ctx*. *netns* values beyond the 2208 * range of 32-bit integers are reserved for future use. 2209 * 2210 * All values for *flags* are reserved for future usage, and must 2211 * be left at zero. 2212 * 2213 * This helper is available only if the kernel was compiled with 2214 * **CONFIG_NET** configuration option. 2215 * 2216 * Returns 2217 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2218 * For sockets with reuseport option, the **struct bpf_sock** 2219 * result is from *reuse*\ **->socks**\ [] using the hash of the 2220 * tuple. 2221 */ 2222 static struct bpf_sock *(* const bpf_sk_lookup_tcp)(void *ctx, struct bpf_sock_tuple *tuple, __u32 tuple_size, __u64 netns, __u64 flags) = (void *) 84; 2223 2224 /* 2225 * bpf_sk_lookup_udp 2226 * 2227 * Look for UDP socket matching *tuple*, optionally in a child 2228 * network namespace *netns*. The return value must be checked, 2229 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2230 * 2231 * The *ctx* should point to the context of the program, such as 2232 * the skb or socket (depending on the hook in use). This is used 2233 * to determine the base network namespace for the lookup. 2234 * 2235 * *tuple_size* must be one of: 2236 * 2237 * **sizeof**\ (*tuple*\ **->ipv4**) 2238 * Look for an IPv4 socket. 2239 * **sizeof**\ (*tuple*\ **->ipv6**) 2240 * Look for an IPv6 socket. 2241 * 2242 * If the *netns* is a negative signed 32-bit integer, then the 2243 * socket lookup table in the netns associated with the *ctx* 2244 * will be used. For the TC hooks, this is the netns of the device 2245 * in the skb. For socket hooks, this is the netns of the socket. 2246 * If *netns* is any other signed 32-bit value greater than or 2247 * equal to zero then it specifies the ID of the netns relative to 2248 * the netns associated with the *ctx*. *netns* values beyond the 2249 * range of 32-bit integers are reserved for future use. 2250 * 2251 * All values for *flags* are reserved for future usage, and must 2252 * be left at zero. 2253 * 2254 * This helper is available only if the kernel was compiled with 2255 * **CONFIG_NET** configuration option. 2256 * 2257 * Returns 2258 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2259 * For sockets with reuseport option, the **struct bpf_sock** 2260 * result is from *reuse*\ **->socks**\ [] using the hash of the 2261 * tuple. 2262 */ 2263 static struct bpf_sock *(* const bpf_sk_lookup_udp)(void *ctx, struct bpf_sock_tuple *tuple, __u32 tuple_size, __u64 netns, __u64 flags) = (void *) 85; 2264 2265 /* 2266 * bpf_sk_release 2267 * 2268 * Release the reference held by *sock*. *sock* must be a 2269 * non-**NULL** pointer that was returned from 2270 * **bpf_sk_lookup_xxx**\ (). 2271 * 2272 * Returns 2273 * 0 on success, or a negative error in case of failure. 2274 */ 2275 static long (* const bpf_sk_release)(void *sock) = (void *) 86; 2276 2277 /* 2278 * bpf_map_push_elem 2279 * 2280 * Push an element *value* in *map*. *flags* is one of: 2281 * 2282 * **BPF_EXIST** 2283 * If the queue/stack is full, the oldest element is 2284 * removed to make room for this. 2285 * 2286 * Returns 2287 * 0 on success, or a negative error in case of failure. 2288 */ 2289 static long (* const bpf_map_push_elem)(void *map, const void *value, __u64 flags) = (void *) 87; 2290 2291 /* 2292 * bpf_map_pop_elem 2293 * 2294 * Pop an element from *map*. 2295 * 2296 * Returns 2297 * 0 on success, or a negative error in case of failure. 2298 */ 2299 static long (* const bpf_map_pop_elem)(void *map, void *value) = (void *) 88; 2300 2301 /* 2302 * bpf_map_peek_elem 2303 * 2304 * Get an element from *map* without removing it. 2305 * 2306 * Returns 2307 * 0 on success, or a negative error in case of failure. 2308 */ 2309 static long (* const bpf_map_peek_elem)(void *map, void *value) = (void *) 89; 2310 2311 /* 2312 * bpf_msg_push_data 2313 * 2314 * For socket policies, insert *len* bytes into *msg* at offset 2315 * *start*. 2316 * 2317 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2318 * *msg* it may want to insert metadata or options into the *msg*. 2319 * This can later be read and used by any of the lower layer BPF 2320 * hooks. 2321 * 2322 * This helper may fail if under memory pressure (a malloc 2323 * fails) in these cases BPF programs will get an appropriate 2324 * error and BPF programs will need to handle them. 2325 * 2326 * Returns 2327 * 0 on success, or a negative error in case of failure. 2328 */ 2329 static long (* const bpf_msg_push_data)(struct sk_msg_md *msg, __u32 start, __u32 len, __u64 flags) = (void *) 90; 2330 2331 /* 2332 * bpf_msg_pop_data 2333 * 2334 * Will remove *len* bytes from a *msg* starting at byte *start*. 2335 * This may result in **ENOMEM** errors under certain situations if 2336 * an allocation and copy are required due to a full ring buffer. 2337 * However, the helper will try to avoid doing the allocation 2338 * if possible. Other errors can occur if input parameters are 2339 * invalid either due to *start* byte not being valid part of *msg* 2340 * payload and/or *pop* value being to large. 2341 * 2342 * Returns 2343 * 0 on success, or a negative error in case of failure. 2344 */ 2345 static long (* const bpf_msg_pop_data)(struct sk_msg_md *msg, __u32 start, __u32 len, __u64 flags) = (void *) 91; 2346 2347 /* 2348 * bpf_rc_pointer_rel 2349 * 2350 * This helper is used in programs implementing IR decoding, to 2351 * report a successfully decoded pointer movement. 2352 * 2353 * The *ctx* should point to the lirc sample as passed into 2354 * the program. 2355 * 2356 * This helper is only available is the kernel was compiled with 2357 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2358 * "**y**". 2359 * 2360 * Returns 2361 * 0 2362 */ 2363 static long (* const bpf_rc_pointer_rel)(void *ctx, __s32 rel_x, __s32 rel_y) = (void *) 92; 2364 2365 /* 2366 * bpf_spin_lock 2367 * 2368 * Acquire a spinlock represented by the pointer *lock*, which is 2369 * stored as part of a value of a map. Taking the lock allows to 2370 * safely update the rest of the fields in that value. The 2371 * spinlock can (and must) later be released with a call to 2372 * **bpf_spin_unlock**\ (\ *lock*\ ). 2373 * 2374 * Spinlocks in BPF programs come with a number of restrictions 2375 * and constraints: 2376 * 2377 * * **bpf_spin_lock** objects are only allowed inside maps of 2378 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 2379 * list could be extended in the future). 2380 * * BTF description of the map is mandatory. 2381 * * The BPF program can take ONE lock at a time, since taking two 2382 * or more could cause dead locks. 2383 * * Only one **struct bpf_spin_lock** is allowed per map element. 2384 * * When the lock is taken, calls (either BPF to BPF or helpers) 2385 * are not allowed. 2386 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 2387 * allowed inside a spinlock-ed region. 2388 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 2389 * the lock, on all execution paths, before it returns. 2390 * * The BPF program can access **struct bpf_spin_lock** only via 2391 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 2392 * helpers. Loading or storing data into the **struct 2393 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 2394 * * To use the **bpf_spin_lock**\ () helper, the BTF description 2395 * of the map value must be a struct and have **struct 2396 * bpf_spin_lock** *anyname*\ **;** field at the top level. 2397 * Nested lock inside another struct is not allowed. 2398 * * The **struct bpf_spin_lock** *lock* field in a map value must 2399 * be aligned on a multiple of 4 bytes in that value. 2400 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 2401 * the **bpf_spin_lock** field to user space. 2402 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 2403 * a BPF program, do not update the **bpf_spin_lock** field. 2404 * * **bpf_spin_lock** cannot be on the stack or inside a 2405 * networking packet (it can only be inside of a map values). 2406 * * **bpf_spin_lock** is available to root only. 2407 * * Tracing programs and socket filter programs cannot use 2408 * **bpf_spin_lock**\ () due to insufficient preemption checks 2409 * (but this may change in the future). 2410 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 2411 * 2412 * Returns 2413 * 0 2414 */ 2415 static long (* const bpf_spin_lock)(struct bpf_spin_lock *lock) = (void *) 93; 2416 2417 /* 2418 * bpf_spin_unlock 2419 * 2420 * Release the *lock* previously locked by a call to 2421 * **bpf_spin_lock**\ (\ *lock*\ ). 2422 * 2423 * Returns 2424 * 0 2425 */ 2426 static long (* const bpf_spin_unlock)(struct bpf_spin_lock *lock) = (void *) 94; 2427 2428 /* 2429 * bpf_sk_fullsock 2430 * 2431 * This helper gets a **struct bpf_sock** pointer such 2432 * that all the fields in this **bpf_sock** can be accessed. 2433 * 2434 * Returns 2435 * A **struct bpf_sock** pointer on success, or **NULL** in 2436 * case of failure. 2437 */ 2438 static struct bpf_sock *(* const bpf_sk_fullsock)(struct bpf_sock *sk) = (void *) 95; 2439 2440 /* 2441 * bpf_tcp_sock 2442 * 2443 * This helper gets a **struct bpf_tcp_sock** pointer from a 2444 * **struct bpf_sock** pointer. 2445 * 2446 * Returns 2447 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2448 * case of failure. 2449 */ 2450 static struct bpf_tcp_sock *(* const bpf_tcp_sock)(struct bpf_sock *sk) = (void *) 96; 2451 2452 /* 2453 * bpf_skb_ecn_set_ce 2454 * 2455 * Set ECN (Explicit Congestion Notification) field of IP header 2456 * to **CE** (Congestion Encountered) if current value is **ECT** 2457 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 2458 * and IPv4. 2459 * 2460 * Returns 2461 * 1 if the **CE** flag is set (either by the current helper call 2462 * or because it was already present), 0 if it is not set. 2463 */ 2464 static long (* const bpf_skb_ecn_set_ce)(struct __sk_buff *skb) = (void *) 97; 2465 2466 /* 2467 * bpf_get_listener_sock 2468 * 2469 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 2470 * **bpf_sk_release**\ () is unnecessary and not allowed. 2471 * 2472 * Returns 2473 * A **struct bpf_sock** pointer on success, or **NULL** in 2474 * case of failure. 2475 */ 2476 static struct bpf_sock *(* const bpf_get_listener_sock)(struct bpf_sock *sk) = (void *) 98; 2477 2478 /* 2479 * bpf_skc_lookup_tcp 2480 * 2481 * Look for TCP socket matching *tuple*, optionally in a child 2482 * network namespace *netns*. The return value must be checked, 2483 * and if non-**NULL**, released via **bpf_sk_release**\ (). 2484 * 2485 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 2486 * that it also returns timewait or request sockets. Use 2487 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 2488 * full structure. 2489 * 2490 * This helper is available only if the kernel was compiled with 2491 * **CONFIG_NET** configuration option. 2492 * 2493 * Returns 2494 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2495 * For sockets with reuseport option, the **struct bpf_sock** 2496 * result is from *reuse*\ **->socks**\ [] using the hash of the 2497 * tuple. 2498 */ 2499 static struct bpf_sock *(* const bpf_skc_lookup_tcp)(void *ctx, struct bpf_sock_tuple *tuple, __u32 tuple_size, __u64 netns, __u64 flags) = (void *) 99; 2500 2501 /* 2502 * bpf_tcp_check_syncookie 2503 * 2504 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 2505 * the listening socket in *sk*. 2506 * 2507 * *iph* points to the start of the IPv4 or IPv6 header, while 2508 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2509 * **sizeof**\ (**struct ipv6hdr**). 2510 * 2511 * *th* points to the start of the TCP header, while *th_len* 2512 * contains the length of the TCP header (at least 2513 * **sizeof**\ (**struct tcphdr**)). 2514 * 2515 * Returns 2516 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 2517 * error otherwise. 2518 */ 2519 static long (* const bpf_tcp_check_syncookie)(void *sk, void *iph, __u32 iph_len, struct tcphdr *th, __u32 th_len) = (void *) 100; 2520 2521 /* 2522 * bpf_sysctl_get_name 2523 * 2524 * Get name of sysctl in /proc/sys/ and copy it into provided by 2525 * program buffer *buf* of size *buf_len*. 2526 * 2527 * The buffer is always NUL terminated, unless it's zero-sized. 2528 * 2529 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2530 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2531 * only (e.g. "tcp_mem"). 2532 * 2533 * Returns 2534 * Number of character copied (not including the trailing NUL). 2535 * 2536 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2537 * truncated name in this case). 2538 */ 2539 static long (* const bpf_sysctl_get_name)(struct bpf_sysctl *ctx, char *buf, unsigned long buf_len, __u64 flags) = (void *) 101; 2540 2541 /* 2542 * bpf_sysctl_get_current_value 2543 * 2544 * Get current value of sysctl as it is presented in /proc/sys 2545 * (incl. newline, etc), and copy it as a string into provided 2546 * by program buffer *buf* of size *buf_len*. 2547 * 2548 * The whole value is copied, no matter what file position user 2549 * space issued e.g. sys_read at. 2550 * 2551 * The buffer is always NUL terminated, unless it's zero-sized. 2552 * 2553 * Returns 2554 * Number of character copied (not including the trailing NUL). 2555 * 2556 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2557 * truncated name in this case). 2558 * 2559 * **-EINVAL** if current value was unavailable, e.g. because 2560 * sysctl is uninitialized and read returns -EIO for it. 2561 */ 2562 static long (* const bpf_sysctl_get_current_value)(struct bpf_sysctl *ctx, char *buf, unsigned long buf_len) = (void *) 102; 2563 2564 /* 2565 * bpf_sysctl_get_new_value 2566 * 2567 * Get new value being written by user space to sysctl (before 2568 * the actual write happens) and copy it as a string into 2569 * provided by program buffer *buf* of size *buf_len*. 2570 * 2571 * User space may write new value at file position > 0. 2572 * 2573 * The buffer is always NUL terminated, unless it's zero-sized. 2574 * 2575 * Returns 2576 * Number of character copied (not including the trailing NUL). 2577 * 2578 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2579 * truncated name in this case). 2580 * 2581 * **-EINVAL** if sysctl is being read. 2582 */ 2583 static long (* const bpf_sysctl_get_new_value)(struct bpf_sysctl *ctx, char *buf, unsigned long buf_len) = (void *) 103; 2584 2585 /* 2586 * bpf_sysctl_set_new_value 2587 * 2588 * Override new value being written by user space to sysctl with 2589 * value provided by program in buffer *buf* of size *buf_len*. 2590 * 2591 * *buf* should contain a string in same form as provided by user 2592 * space on sysctl write. 2593 * 2594 * User space may write new value at file position > 0. To override 2595 * the whole sysctl value file position should be set to zero. 2596 * 2597 * Returns 2598 * 0 on success. 2599 * 2600 * **-E2BIG** if the *buf_len* is too big. 2601 * 2602 * **-EINVAL** if sysctl is being read. 2603 */ 2604 static long (* const bpf_sysctl_set_new_value)(struct bpf_sysctl *ctx, const char *buf, unsigned long buf_len) = (void *) 104; 2605 2606 /* 2607 * bpf_strtol 2608 * 2609 * Convert the initial part of the string from buffer *buf* of 2610 * size *buf_len* to a long integer according to the given base 2611 * and save the result in *res*. 2612 * 2613 * The string may begin with an arbitrary amount of white space 2614 * (as determined by **isspace**\ (3)) followed by a single 2615 * optional '**-**' sign. 2616 * 2617 * Five least significant bits of *flags* encode base, other bits 2618 * are currently unused. 2619 * 2620 * Base must be either 8, 10, 16 or 0 to detect it automatically 2621 * similar to user space **strtol**\ (3). 2622 * 2623 * Returns 2624 * Number of characters consumed on success. Must be positive but 2625 * no more than *buf_len*. 2626 * 2627 * **-EINVAL** if no valid digits were found or unsupported base 2628 * was provided. 2629 * 2630 * **-ERANGE** if resulting value was out of range. 2631 */ 2632 static long (* const bpf_strtol)(const char *buf, unsigned long buf_len, __u64 flags, long *res) = (void *) 105; 2633 2634 /* 2635 * bpf_strtoul 2636 * 2637 * Convert the initial part of the string from buffer *buf* of 2638 * size *buf_len* to an unsigned long integer according to the 2639 * given base and save the result in *res*. 2640 * 2641 * The string may begin with an arbitrary amount of white space 2642 * (as determined by **isspace**\ (3)). 2643 * 2644 * Five least significant bits of *flags* encode base, other bits 2645 * are currently unused. 2646 * 2647 * Base must be either 8, 10, 16 or 0 to detect it automatically 2648 * similar to user space **strtoul**\ (3). 2649 * 2650 * Returns 2651 * Number of characters consumed on success. Must be positive but 2652 * no more than *buf_len*. 2653 * 2654 * **-EINVAL** if no valid digits were found or unsupported base 2655 * was provided. 2656 * 2657 * **-ERANGE** if resulting value was out of range. 2658 */ 2659 static long (* const bpf_strtoul)(const char *buf, unsigned long buf_len, __u64 flags, unsigned long *res) = (void *) 106; 2660 2661 /* 2662 * bpf_sk_storage_get 2663 * 2664 * Get a bpf-local-storage from a *sk*. 2665 * 2666 * Logically, it could be thought of getting the value from 2667 * a *map* with *sk* as the **key**. From this 2668 * perspective, the usage is not much different from 2669 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 2670 * helper enforces the key must be a full socket and the map must 2671 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 2672 * 2673 * Underneath, the value is stored locally at *sk* instead of 2674 * the *map*. The *map* is used as the bpf-local-storage 2675 * "type". The bpf-local-storage "type" (i.e. the *map*) is 2676 * searched against all bpf-local-storages residing at *sk*. 2677 * 2678 * *sk* is a kernel **struct sock** pointer for LSM program. 2679 * *sk* is a **struct bpf_sock** pointer for other program types. 2680 * 2681 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 2682 * used such that a new bpf-local-storage will be 2683 * created if one does not exist. *value* can be used 2684 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 2685 * the initial value of a bpf-local-storage. If *value* is 2686 * **NULL**, the new bpf-local-storage will be zero initialized. 2687 * 2688 * Returns 2689 * A bpf-local-storage pointer is returned on success. 2690 * 2691 * **NULL** if not found or there was an error in adding 2692 * a new bpf-local-storage. 2693 */ 2694 static void *(* const bpf_sk_storage_get)(void *map, void *sk, void *value, __u64 flags) = (void *) 107; 2695 2696 /* 2697 * bpf_sk_storage_delete 2698 * 2699 * Delete a bpf-local-storage from a *sk*. 2700 * 2701 * Returns 2702 * 0 on success. 2703 * 2704 * **-ENOENT** if the bpf-local-storage cannot be found. 2705 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 2706 */ 2707 static long (* const bpf_sk_storage_delete)(void *map, void *sk) = (void *) 108; 2708 2709 /* 2710 * bpf_send_signal 2711 * 2712 * Send signal *sig* to the process of the current task. 2713 * The signal may be delivered to any of this process's threads. 2714 * 2715 * Returns 2716 * 0 on success or successfully queued. 2717 * 2718 * **-EBUSY** if work queue under nmi is full. 2719 * 2720 * **-EINVAL** if *sig* is invalid. 2721 * 2722 * **-EPERM** if no permission to send the *sig*. 2723 * 2724 * **-EAGAIN** if bpf program can try again. 2725 */ 2726 static long (* const bpf_send_signal)(__u32 sig) = (void *) 109; 2727 2728 /* 2729 * bpf_tcp_gen_syncookie 2730 * 2731 * Try to issue a SYN cookie for the packet with corresponding 2732 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 2733 * 2734 * *iph* points to the start of the IPv4 or IPv6 header, while 2735 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 2736 * **sizeof**\ (**struct ipv6hdr**). 2737 * 2738 * *th* points to the start of the TCP header, while *th_len* 2739 * contains the length of the TCP header with options (at least 2740 * **sizeof**\ (**struct tcphdr**)). 2741 * 2742 * Returns 2743 * On success, lower 32 bits hold the generated SYN cookie in 2744 * followed by 16 bits which hold the MSS value for that cookie, 2745 * and the top 16 bits are unused. 2746 * 2747 * On failure, the returned value is one of the following: 2748 * 2749 * **-EINVAL** SYN cookie cannot be issued due to error 2750 * 2751 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 2752 * 2753 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 2754 * 2755 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 2756 */ 2757 static __s64 (* const bpf_tcp_gen_syncookie)(void *sk, void *iph, __u32 iph_len, struct tcphdr *th, __u32 th_len) = (void *) 110; 2758 2759 /* 2760 * bpf_skb_output 2761 * 2762 * Write raw *data* blob into a special BPF perf event held by 2763 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2764 * event must have the following attributes: **PERF_SAMPLE_RAW** 2765 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2766 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2767 * 2768 * The *flags* are used to indicate the index in *map* for which 2769 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2770 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2771 * to indicate that the index of the current CPU core should be 2772 * used. 2773 * 2774 * The value to write, of *size*, is passed through eBPF stack and 2775 * pointed by *data*. 2776 * 2777 * *ctx* is a pointer to in-kernel struct sk_buff. 2778 * 2779 * This helper is similar to **bpf_perf_event_output**\ () but 2780 * restricted to raw_tracepoint bpf programs. 2781 * 2782 * Returns 2783 * 0 on success, or a negative error in case of failure. 2784 */ 2785 static long (* const bpf_skb_output)(void *ctx, void *map, __u64 flags, void *data, __u64 size) = (void *) 111; 2786 2787 /* 2788 * bpf_probe_read_user 2789 * 2790 * Safely attempt to read *size* bytes from user space address 2791 * *unsafe_ptr* and store the data in *dst*. 2792 * 2793 * Returns 2794 * 0 on success, or a negative error in case of failure. 2795 */ 2796 static long (* const bpf_probe_read_user)(void *dst, __u32 size, const void *unsafe_ptr) = (void *) 112; 2797 2798 /* 2799 * bpf_probe_read_kernel 2800 * 2801 * Safely attempt to read *size* bytes from kernel space address 2802 * *unsafe_ptr* and store the data in *dst*. 2803 * 2804 * Returns 2805 * 0 on success, or a negative error in case of failure. 2806 */ 2807 static long (* const bpf_probe_read_kernel)(void *dst, __u32 size, const void *unsafe_ptr) = (void *) 113; 2808 2809 /* 2810 * bpf_probe_read_user_str 2811 * 2812 * Copy a NUL terminated string from an unsafe user address 2813 * *unsafe_ptr* to *dst*. The *size* should include the 2814 * terminating NUL byte. In case the string length is smaller than 2815 * *size*, the target is not padded with further NUL bytes. If the 2816 * string length is larger than *size*, just *size*-1 bytes are 2817 * copied and the last byte is set to NUL. 2818 * 2819 * On success, returns the number of bytes that were written, 2820 * including the terminal NUL. This makes this helper useful in 2821 * tracing programs for reading strings, and more importantly to 2822 * get its length at runtime. See the following snippet: 2823 * 2824 * :: 2825 * 2826 * SEC("kprobe/sys_open") 2827 * void bpf_sys_open(struct pt_regs *ctx) 2828 * { 2829 * char buf[PATHLEN]; // PATHLEN is defined to 256 2830 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 2831 * ctx->di); 2832 * 2833 * // Consume buf, for example push it to 2834 * // userspace via bpf_perf_event_output(); we 2835 * // can use res (the string length) as event 2836 * // size, after checking its boundaries. 2837 * } 2838 * 2839 * In comparison, using **bpf_probe_read_user**\ () helper here 2840 * instead to read the string would require to estimate the length 2841 * at compile time, and would often result in copying more memory 2842 * than necessary. 2843 * 2844 * Another useful use case is when parsing individual process 2845 * arguments or individual environment variables navigating 2846 * *current*\ **->mm->arg_start** and *current*\ 2847 * **->mm->env_start**: using this helper and the return value, 2848 * one can quickly iterate at the right offset of the memory area. 2849 * 2850 * Returns 2851 * On success, the strictly positive length of the output string, 2852 * including the trailing NUL character. On error, a negative 2853 * value. 2854 */ 2855 static long (* const bpf_probe_read_user_str)(void *dst, __u32 size, const void *unsafe_ptr) = (void *) 114; 2856 2857 /* 2858 * bpf_probe_read_kernel_str 2859 * 2860 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 2861 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 2862 * 2863 * Returns 2864 * On success, the strictly positive length of the string, including 2865 * the trailing NUL character. On error, a negative value. 2866 */ 2867 static long (* const bpf_probe_read_kernel_str)(void *dst, __u32 size, const void *unsafe_ptr) = (void *) 115; 2868 2869 /* 2870 * bpf_tcp_send_ack 2871 * 2872 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 2873 * *rcv_nxt* is the ack_seq to be sent out. 2874 * 2875 * Returns 2876 * 0 on success, or a negative error in case of failure. 2877 */ 2878 static long (* const bpf_tcp_send_ack)(void *tp, __u32 rcv_nxt) = (void *) 116; 2879 2880 /* 2881 * bpf_send_signal_thread 2882 * 2883 * Send signal *sig* to the thread corresponding to the current task. 2884 * 2885 * Returns 2886 * 0 on success or successfully queued. 2887 * 2888 * **-EBUSY** if work queue under nmi is full. 2889 * 2890 * **-EINVAL** if *sig* is invalid. 2891 * 2892 * **-EPERM** if no permission to send the *sig*. 2893 * 2894 * **-EAGAIN** if bpf program can try again. 2895 */ 2896 static long (* const bpf_send_signal_thread)(__u32 sig) = (void *) 117; 2897 2898 /* 2899 * bpf_jiffies64 2900 * 2901 * Obtain the 64bit jiffies 2902 * 2903 * Returns 2904 * The 64 bit jiffies 2905 */ 2906 static __u64 (* const bpf_jiffies64)(void) = (void *) 118; 2907 2908 /* 2909 * bpf_read_branch_records 2910 * 2911 * For an eBPF program attached to a perf event, retrieve the 2912 * branch records (**struct perf_branch_entry**) associated to *ctx* 2913 * and store it in the buffer pointed by *buf* up to size 2914 * *size* bytes. 2915 * 2916 * Returns 2917 * On success, number of bytes written to *buf*. On error, a 2918 * negative value. 2919 * 2920 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 2921 * instead return the number of bytes required to store all the 2922 * branch entries. If this flag is set, *buf* may be NULL. 2923 * 2924 * **-EINVAL** if arguments invalid or **size** not a multiple 2925 * of **sizeof**\ (**struct perf_branch_entry**\ ). 2926 * 2927 * **-ENOENT** if architecture does not support branch records. 2928 */ 2929 static long (* const bpf_read_branch_records)(struct bpf_perf_event_data *ctx, void *buf, __u32 size, __u64 flags) = (void *) 119; 2930 2931 /* 2932 * bpf_get_ns_current_pid_tgid 2933 * 2934 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 2935 * *namespace* will be returned in *nsdata*. 2936 * 2937 * Returns 2938 * 0 on success, or one of the following in case of failure: 2939 * 2940 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 2941 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 2942 * 2943 * **-ENOENT** if pidns does not exists for the current task. 2944 */ 2945 static long (* const bpf_get_ns_current_pid_tgid)(__u64 dev, __u64 ino, struct bpf_pidns_info *nsdata, __u32 size) = (void *) 120; 2946 2947 /* 2948 * bpf_xdp_output 2949 * 2950 * Write raw *data* blob into a special BPF perf event held by 2951 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2952 * event must have the following attributes: **PERF_SAMPLE_RAW** 2953 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2954 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2955 * 2956 * The *flags* are used to indicate the index in *map* for which 2957 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2958 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2959 * to indicate that the index of the current CPU core should be 2960 * used. 2961 * 2962 * The value to write, of *size*, is passed through eBPF stack and 2963 * pointed by *data*. 2964 * 2965 * *ctx* is a pointer to in-kernel struct xdp_buff. 2966 * 2967 * This helper is similar to **bpf_perf_eventoutput**\ () but 2968 * restricted to raw_tracepoint bpf programs. 2969 * 2970 * Returns 2971 * 0 on success, or a negative error in case of failure. 2972 */ 2973 static long (* const bpf_xdp_output)(void *ctx, void *map, __u64 flags, void *data, __u64 size) = (void *) 121; 2974 2975 /* 2976 * bpf_get_netns_cookie 2977 * 2978 * Retrieve the cookie (generated by the kernel) of the network 2979 * namespace the input *ctx* is associated with. The network 2980 * namespace cookie remains stable for its lifetime and provides 2981 * a global identifier that can be assumed unique. If *ctx* is 2982 * NULL, then the helper returns the cookie for the initial 2983 * network namespace. The cookie itself is very similar to that 2984 * of **bpf_get_socket_cookie**\ () helper, but for network 2985 * namespaces instead of sockets. 2986 * 2987 * Returns 2988 * A 8-byte long opaque number. 2989 */ 2990 static __u64 (* const bpf_get_netns_cookie)(void *ctx) = (void *) 122; 2991 2992 /* 2993 * bpf_get_current_ancestor_cgroup_id 2994 * 2995 * Return id of cgroup v2 that is ancestor of the cgroup associated 2996 * with the current task at the *ancestor_level*. The root cgroup 2997 * is at *ancestor_level* zero and each step down the hierarchy 2998 * increments the level. If *ancestor_level* == level of cgroup 2999 * associated with the current task, then return value will be the 3000 * same as that of **bpf_get_current_cgroup_id**\ (). 3001 * 3002 * The helper is useful to implement policies based on cgroups 3003 * that are upper in hierarchy than immediate cgroup associated 3004 * with the current task. 3005 * 3006 * The format of returned id and helper limitations are same as in 3007 * **bpf_get_current_cgroup_id**\ (). 3008 * 3009 * Returns 3010 * The id is returned or 0 in case the id could not be retrieved. 3011 */ 3012 static __u64 (* const bpf_get_current_ancestor_cgroup_id)(int ancestor_level) = (void *) 123; 3013 3014 /* 3015 * bpf_sk_assign 3016 * 3017 * Helper is overloaded depending on BPF program type. This 3018 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 3019 * **BPF_PROG_TYPE_SCHED_ACT** programs. 3020 * 3021 * Assign the *sk* to the *skb*. When combined with appropriate 3022 * routing configuration to receive the packet towards the socket, 3023 * will cause *skb* to be delivered to the specified socket. 3024 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3025 * **bpf_clone_redirect**\ () or other methods outside of BPF may 3026 * interfere with successful delivery to the socket. 3027 * 3028 * This operation is only valid from TC ingress path. 3029 * 3030 * The *flags* argument must be zero. 3031 * 3032 * Returns 3033 * 0 on success, or a negative error in case of failure: 3034 * 3035 * **-EINVAL** if specified *flags* are not supported. 3036 * 3037 * **-ENOENT** if the socket is unavailable for assignment. 3038 * 3039 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 3040 * 3041 * **-EOPNOTSUPP** if the operation is not supported, for example 3042 * a call from outside of TC ingress. 3043 */ 3044 static long (* const bpf_sk_assign)(void *ctx, void *sk, __u64 flags) = (void *) 124; 3045 3046 /* 3047 * bpf_ktime_get_boot_ns 3048 * 3049 * Return the time elapsed since system boot, in nanoseconds. 3050 * Does include the time the system was suspended. 3051 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 3052 * 3053 * Returns 3054 * Current *ktime*. 3055 */ 3056 static __u64 (* const bpf_ktime_get_boot_ns)(void) = (void *) 125; 3057 3058 /* 3059 * bpf_seq_printf 3060 * 3061 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 3062 * out the format string. 3063 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 3064 * the format string itself. The *data* and *data_len* are format string 3065 * arguments. The *data* are a **u64** array and corresponding format string 3066 * values are stored in the array. For strings and pointers where pointees 3067 * are accessed, only the pointer values are stored in the *data* array. 3068 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 3069 * 3070 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 3071 * Reading kernel memory may fail due to either invalid address or 3072 * valid address but requiring a major memory fault. If reading kernel memory 3073 * fails, the string for **%s** will be an empty string, and the ip 3074 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 3075 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 3076 * 3077 * Returns 3078 * 0 on success, or a negative error in case of failure: 3079 * 3080 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 3081 * by returning 1 from bpf program. 3082 * 3083 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 3084 * 3085 * **-E2BIG** if *fmt* contains too many format specifiers. 3086 * 3087 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3088 */ 3089 static long (* const bpf_seq_printf)(struct seq_file *m, const char *fmt, __u32 fmt_size, const void *data, __u32 data_len) = (void *) 126; 3090 3091 /* 3092 * bpf_seq_write 3093 * 3094 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 3095 * The *m* represents the seq_file. The *data* and *len* represent the 3096 * data to write in bytes. 3097 * 3098 * Returns 3099 * 0 on success, or a negative error in case of failure: 3100 * 3101 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3102 */ 3103 static long (* const bpf_seq_write)(struct seq_file *m, const void *data, __u32 len) = (void *) 127; 3104 3105 /* 3106 * bpf_sk_cgroup_id 3107 * 3108 * Return the cgroup v2 id of the socket *sk*. 3109 * 3110 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 3111 * returned from **bpf_sk_lookup_xxx**\ (), 3112 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 3113 * same as in **bpf_skb_cgroup_id**\ (). 3114 * 3115 * This helper is available only if the kernel was compiled with 3116 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 3117 * 3118 * Returns 3119 * The id is returned or 0 in case the id could not be retrieved. 3120 */ 3121 static __u64 (* const bpf_sk_cgroup_id)(void *sk) = (void *) 128; 3122 3123 /* 3124 * bpf_sk_ancestor_cgroup_id 3125 * 3126 * Return id of cgroup v2 that is ancestor of cgroup associated 3127 * with the *sk* at the *ancestor_level*. The root cgroup is at 3128 * *ancestor_level* zero and each step down the hierarchy 3129 * increments the level. If *ancestor_level* == level of cgroup 3130 * associated with *sk*, then return value will be same as that 3131 * of **bpf_sk_cgroup_id**\ (). 3132 * 3133 * The helper is useful to implement policies based on cgroups 3134 * that are upper in hierarchy than immediate cgroup associated 3135 * with *sk*. 3136 * 3137 * The format of returned id and helper limitations are same as in 3138 * **bpf_sk_cgroup_id**\ (). 3139 * 3140 * Returns 3141 * The id is returned or 0 in case the id could not be retrieved. 3142 */ 3143 static __u64 (* const bpf_sk_ancestor_cgroup_id)(void *sk, int ancestor_level) = (void *) 129; 3144 3145 /* 3146 * bpf_ringbuf_output 3147 * 3148 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 3149 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3150 * of new data availability is sent. 3151 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3152 * of new data availability is sent unconditionally. 3153 * If **0** is specified in *flags*, an adaptive notification 3154 * of new data availability is sent. 3155 * 3156 * An adaptive notification is a notification sent whenever the user-space 3157 * process has caught up and consumed all available payloads. In case the user-space 3158 * process is still processing a previous payload, then no notification is needed 3159 * as it will process the newly added payload automatically. 3160 * 3161 * Returns 3162 * 0 on success, or a negative error in case of failure. 3163 */ 3164 static long (* const bpf_ringbuf_output)(void *ringbuf, void *data, __u64 size, __u64 flags) = (void *) 130; 3165 3166 /* 3167 * bpf_ringbuf_reserve 3168 * 3169 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 3170 * *flags* must be 0. 3171 * 3172 * Returns 3173 * Valid pointer with *size* bytes of memory available; NULL, 3174 * otherwise. 3175 */ 3176 static void *(* const bpf_ringbuf_reserve)(void *ringbuf, __u64 size, __u64 flags) = (void *) 131; 3177 3178 /* 3179 * bpf_ringbuf_submit 3180 * 3181 * Submit reserved ring buffer sample, pointed to by *data*. 3182 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3183 * of new data availability is sent. 3184 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3185 * of new data availability is sent unconditionally. 3186 * If **0** is specified in *flags*, an adaptive notification 3187 * of new data availability is sent. 3188 * 3189 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 3190 * 3191 * Returns 3192 * Nothing. Always succeeds. 3193 */ 3194 static void (* const bpf_ringbuf_submit)(void *data, __u64 flags) = (void *) 132; 3195 3196 /* 3197 * bpf_ringbuf_discard 3198 * 3199 * Discard reserved ring buffer sample, pointed to by *data*. 3200 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3201 * of new data availability is sent. 3202 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3203 * of new data availability is sent unconditionally. 3204 * If **0** is specified in *flags*, an adaptive notification 3205 * of new data availability is sent. 3206 * 3207 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 3208 * 3209 * Returns 3210 * Nothing. Always succeeds. 3211 */ 3212 static void (* const bpf_ringbuf_discard)(void *data, __u64 flags) = (void *) 133; 3213 3214 /* 3215 * bpf_ringbuf_query 3216 * 3217 * Query various characteristics of provided ring buffer. What 3218 * exactly is queries is determined by *flags*: 3219 * 3220 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 3221 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 3222 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 3223 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 3224 * 3225 * Data returned is just a momentary snapshot of actual values 3226 * and could be inaccurate, so this facility should be used to 3227 * power heuristics and for reporting, not to make 100% correct 3228 * calculation. 3229 * 3230 * Returns 3231 * Requested value, or 0, if *flags* are not recognized. 3232 */ 3233 static __u64 (* const bpf_ringbuf_query)(void *ringbuf, __u64 flags) = (void *) 134; 3234 3235 /* 3236 * bpf_csum_level 3237 * 3238 * Change the skbs checksum level by one layer up or down, or 3239 * reset it entirely to none in order to have the stack perform 3240 * checksum validation. The level is applicable to the following 3241 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 3242 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 3243 * through **bpf_skb_adjust_room**\ () helper with passing in 3244 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 3245 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 3246 * the UDP header is removed. Similarly, an encap of the latter 3247 * into the former could be accompanied by a helper call to 3248 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 3249 * skb is still intended to be processed in higher layers of the 3250 * stack instead of just egressing at tc. 3251 * 3252 * There are three supported level settings at this time: 3253 * 3254 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 3255 * with CHECKSUM_UNNECESSARY. 3256 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 3257 * with CHECKSUM_UNNECESSARY. 3258 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 3259 * sets CHECKSUM_NONE to force checksum validation by the stack. 3260 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 3261 * skb->csum_level. 3262 * 3263 * Returns 3264 * 0 on success, or a negative error in case of failure. In the 3265 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 3266 * is returned or the error code -EACCES in case the skb is not 3267 * subject to CHECKSUM_UNNECESSARY. 3268 */ 3269 static long (* const bpf_csum_level)(struct __sk_buff *skb, __u64 level) = (void *) 135; 3270 3271 /* 3272 * bpf_skc_to_tcp6_sock 3273 * 3274 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 3275 * 3276 * Returns 3277 * *sk* if casting is valid, or **NULL** otherwise. 3278 */ 3279 static struct tcp6_sock *(* const bpf_skc_to_tcp6_sock)(void *sk) = (void *) 136; 3280 3281 /* 3282 * bpf_skc_to_tcp_sock 3283 * 3284 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 3285 * 3286 * Returns 3287 * *sk* if casting is valid, or **NULL** otherwise. 3288 */ 3289 static struct tcp_sock *(* const bpf_skc_to_tcp_sock)(void *sk) = (void *) 137; 3290 3291 /* 3292 * bpf_skc_to_tcp_timewait_sock 3293 * 3294 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 3295 * 3296 * Returns 3297 * *sk* if casting is valid, or **NULL** otherwise. 3298 */ 3299 static struct tcp_timewait_sock *(* const bpf_skc_to_tcp_timewait_sock)(void *sk) = (void *) 138; 3300 3301 /* 3302 * bpf_skc_to_tcp_request_sock 3303 * 3304 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 3305 * 3306 * Returns 3307 * *sk* if casting is valid, or **NULL** otherwise. 3308 */ 3309 static struct tcp_request_sock *(* const bpf_skc_to_tcp_request_sock)(void *sk) = (void *) 139; 3310 3311 /* 3312 * bpf_skc_to_udp6_sock 3313 * 3314 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 3315 * 3316 * Returns 3317 * *sk* if casting is valid, or **NULL** otherwise. 3318 */ 3319 static struct udp6_sock *(* const bpf_skc_to_udp6_sock)(void *sk) = (void *) 140; 3320 3321 /* 3322 * bpf_get_task_stack 3323 * 3324 * Return a user or a kernel stack in bpf program provided buffer. 3325 * Note: the user stack will only be populated if the *task* is 3326 * the current task; all other tasks will return -EOPNOTSUPP. 3327 * To achieve this, the helper needs *task*, which is a valid 3328 * pointer to **struct task_struct**. To store the stacktrace, the 3329 * bpf program provides *buf* with a nonnegative *size*. 3330 * 3331 * The last argument, *flags*, holds the number of stack frames to 3332 * skip (from 0 to 255), masked with 3333 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3334 * the following flags: 3335 * 3336 * **BPF_F_USER_STACK** 3337 * Collect a user space stack instead of a kernel stack. 3338 * The *task* must be the current task. 3339 * **BPF_F_USER_BUILD_ID** 3340 * Collect buildid+offset instead of ips for user stack, 3341 * only valid if **BPF_F_USER_STACK** is also specified. 3342 * 3343 * **bpf_get_task_stack**\ () can collect up to 3344 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3345 * to sufficient large buffer size. Note that 3346 * this limit can be controlled with the **sysctl** program, and 3347 * that it should be manually increased in order to profile long 3348 * user stacks (such as stacks for Java programs). To do so, use: 3349 * 3350 * :: 3351 * 3352 * # sysctl kernel.perf_event_max_stack=<new value> 3353 * 3354 * Returns 3355 * The non-negative copied *buf* length equal to or less than 3356 * *size* on success, or a negative error in case of failure. 3357 */ 3358 static long (* const bpf_get_task_stack)(struct task_struct *task, void *buf, __u32 size, __u64 flags) = (void *) 141; 3359 3360 /* 3361 * bpf_load_hdr_opt 3362 * 3363 * Load header option. Support reading a particular TCP header 3364 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 3365 * 3366 * If *flags* is 0, it will search the option from the 3367 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 3368 * has details on what skb_data contains under different 3369 * *skops*\ **->op**. 3370 * 3371 * The first byte of the *searchby_res* specifies the 3372 * kind that it wants to search. 3373 * 3374 * If the searching kind is an experimental kind 3375 * (i.e. 253 or 254 according to RFC6994). It also 3376 * needs to specify the "magic" which is either 3377 * 2 bytes or 4 bytes. It then also needs to 3378 * specify the size of the magic by using 3379 * the 2nd byte which is "kind-length" of a TCP 3380 * header option and the "kind-length" also 3381 * includes the first 2 bytes "kind" and "kind-length" 3382 * itself as a normal TCP header option also does. 3383 * 3384 * For example, to search experimental kind 254 with 3385 * 2 byte magic 0xeB9F, the searchby_res should be 3386 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 3387 * 3388 * To search for the standard window scale option (3), 3389 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 3390 * Note, kind-length must be 0 for regular option. 3391 * 3392 * Searching for No-Op (0) and End-of-Option-List (1) are 3393 * not supported. 3394 * 3395 * *len* must be at least 2 bytes which is the minimal size 3396 * of a header option. 3397 * 3398 * Supported flags: 3399 * 3400 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 3401 * saved_syn packet or the just-received syn packet. 3402 * 3403 * 3404 * Returns 3405 * > 0 when found, the header option is copied to *searchby_res*. 3406 * The return value is the total length copied. On failure, a 3407 * negative error code is returned: 3408 * 3409 * **-EINVAL** if a parameter is invalid. 3410 * 3411 * **-ENOMSG** if the option is not found. 3412 * 3413 * **-ENOENT** if no syn packet is available when 3414 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 3415 * 3416 * **-ENOSPC** if there is not enough space. Only *len* number of 3417 * bytes are copied. 3418 * 3419 * **-EFAULT** on failure to parse the header options in the 3420 * packet. 3421 * 3422 * **-EPERM** if the helper cannot be used under the current 3423 * *skops*\ **->op**. 3424 */ 3425 static long (* const bpf_load_hdr_opt)(struct bpf_sock_ops *skops, void *searchby_res, __u32 len, __u64 flags) = (void *) 142; 3426 3427 /* 3428 * bpf_store_hdr_opt 3429 * 3430 * Store header option. The data will be copied 3431 * from buffer *from* with length *len* to the TCP header. 3432 * 3433 * The buffer *from* should have the whole option that 3434 * includes the kind, kind-length, and the actual 3435 * option data. The *len* must be at least kind-length 3436 * long. The kind-length does not have to be 4 byte 3437 * aligned. The kernel will take care of the padding 3438 * and setting the 4 bytes aligned value to th->doff. 3439 * 3440 * This helper will check for duplicated option 3441 * by searching the same option in the outgoing skb. 3442 * 3443 * This helper can only be called during 3444 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 3445 * 3446 * 3447 * Returns 3448 * 0 on success, or negative error in case of failure: 3449 * 3450 * **-EINVAL** If param is invalid. 3451 * 3452 * **-ENOSPC** if there is not enough space in the header. 3453 * Nothing has been written 3454 * 3455 * **-EEXIST** if the option already exists. 3456 * 3457 * **-EFAULT** on failure to parse the existing header options. 3458 * 3459 * **-EPERM** if the helper cannot be used under the current 3460 * *skops*\ **->op**. 3461 */ 3462 static long (* const bpf_store_hdr_opt)(struct bpf_sock_ops *skops, const void *from, __u32 len, __u64 flags) = (void *) 143; 3463 3464 /* 3465 * bpf_reserve_hdr_opt 3466 * 3467 * Reserve *len* bytes for the bpf header option. The 3468 * space will be used by **bpf_store_hdr_opt**\ () later in 3469 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 3470 * 3471 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 3472 * the total number of bytes will be reserved. 3473 * 3474 * This helper can only be called during 3475 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 3476 * 3477 * 3478 * Returns 3479 * 0 on success, or negative error in case of failure: 3480 * 3481 * **-EINVAL** if a parameter is invalid. 3482 * 3483 * **-ENOSPC** if there is not enough space in the header. 3484 * 3485 * **-EPERM** if the helper cannot be used under the current 3486 * *skops*\ **->op**. 3487 */ 3488 static long (* const bpf_reserve_hdr_opt)(struct bpf_sock_ops *skops, __u32 len, __u64 flags) = (void *) 144; 3489 3490 /* 3491 * bpf_inode_storage_get 3492 * 3493 * Get a bpf_local_storage from an *inode*. 3494 * 3495 * Logically, it could be thought of as getting the value from 3496 * a *map* with *inode* as the **key**. From this 3497 * perspective, the usage is not much different from 3498 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 3499 * helper enforces the key must be an inode and the map must also 3500 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 3501 * 3502 * Underneath, the value is stored locally at *inode* instead of 3503 * the *map*. The *map* is used as the bpf-local-storage 3504 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3505 * searched against all bpf_local_storage residing at *inode*. 3506 * 3507 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 3508 * used such that a new bpf_local_storage will be 3509 * created if one does not exist. *value* can be used 3510 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 3511 * the initial value of a bpf_local_storage. If *value* is 3512 * **NULL**, the new bpf_local_storage will be zero initialized. 3513 * 3514 * Returns 3515 * A bpf_local_storage pointer is returned on success. 3516 * 3517 * **NULL** if not found or there was an error in adding 3518 * a new bpf_local_storage. 3519 */ 3520 static void *(* const bpf_inode_storage_get)(void *map, void *inode, void *value, __u64 flags) = (void *) 145; 3521 3522 /* 3523 * bpf_inode_storage_delete 3524 * 3525 * Delete a bpf_local_storage from an *inode*. 3526 * 3527 * Returns 3528 * 0 on success. 3529 * 3530 * **-ENOENT** if the bpf_local_storage cannot be found. 3531 */ 3532 static int (* const bpf_inode_storage_delete)(void *map, void *inode) = (void *) 146; 3533 3534 /* 3535 * bpf_d_path 3536 * 3537 * Return full path for given **struct path** object, which 3538 * needs to be the kernel BTF *path* object. The path is 3539 * returned in the provided buffer *buf* of size *sz* and 3540 * is zero terminated. 3541 * 3542 * 3543 * Returns 3544 * On success, the strictly positive length of the string, 3545 * including the trailing NUL character. On error, a negative 3546 * value. 3547 */ 3548 static long (* const bpf_d_path)(struct path *path, char *buf, __u32 sz) = (void *) 147; 3549 3550 /* 3551 * bpf_copy_from_user 3552 * 3553 * Read *size* bytes from user space address *user_ptr* and store 3554 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 3555 * 3556 * Returns 3557 * 0 on success, or a negative error in case of failure. 3558 */ 3559 static long (* const bpf_copy_from_user)(void *dst, __u32 size, const void *user_ptr) = (void *) 148; 3560 3561 /* 3562 * bpf_snprintf_btf 3563 * 3564 * Use BTF to store a string representation of *ptr*->ptr in *str*, 3565 * using *ptr*->type_id. This value should specify the type 3566 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 3567 * can be used to look up vmlinux BTF type ids. Traversing the 3568 * data structure using BTF, the type information and values are 3569 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 3570 * the pointer data is carried out to avoid kernel crashes during 3571 * operation. Smaller types can use string space on the stack; 3572 * larger programs can use map data to store the string 3573 * representation. 3574 * 3575 * The string can be subsequently shared with userspace via 3576 * bpf_perf_event_output() or ring buffer interfaces. 3577 * bpf_trace_printk() is to be avoided as it places too small 3578 * a limit on string size to be useful. 3579 * 3580 * *flags* is a combination of 3581 * 3582 * **BTF_F_COMPACT** 3583 * no formatting around type information 3584 * **BTF_F_NONAME** 3585 * no struct/union member names/types 3586 * **BTF_F_PTR_RAW** 3587 * show raw (unobfuscated) pointer values; 3588 * equivalent to printk specifier %px. 3589 * **BTF_F_ZERO** 3590 * show zero-valued struct/union members; they 3591 * are not displayed by default 3592 * 3593 * 3594 * Returns 3595 * The number of bytes that were written (or would have been 3596 * written if output had to be truncated due to string size), 3597 * or a negative error in cases of failure. 3598 */ 3599 static long (* const bpf_snprintf_btf)(char *str, __u32 str_size, struct btf_ptr *ptr, __u32 btf_ptr_size, __u64 flags) = (void *) 149; 3600 3601 /* 3602 * bpf_seq_printf_btf 3603 * 3604 * Use BTF to write to seq_write a string representation of 3605 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 3606 * *flags* are identical to those used for bpf_snprintf_btf. 3607 * 3608 * Returns 3609 * 0 on success or a negative error in case of failure. 3610 */ 3611 static long (* const bpf_seq_printf_btf)(struct seq_file *m, struct btf_ptr *ptr, __u32 ptr_size, __u64 flags) = (void *) 150; 3612 3613 /* 3614 * bpf_skb_cgroup_classid 3615 * 3616 * See **bpf_get_cgroup_classid**\ () for the main description. 3617 * This helper differs from **bpf_get_cgroup_classid**\ () in that 3618 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 3619 * associated socket instead of the current process. 3620 * 3621 * Returns 3622 * The id is returned or 0 in case the id could not be retrieved. 3623 */ 3624 static __u64 (* const bpf_skb_cgroup_classid)(struct __sk_buff *skb) = (void *) 151; 3625 3626 /* 3627 * bpf_redirect_neigh 3628 * 3629 * Redirect the packet to another net device of index *ifindex* 3630 * and fill in L2 addresses from neighboring subsystem. This helper 3631 * is somewhat similar to **bpf_redirect**\ (), except that it 3632 * populates L2 addresses as well, meaning, internally, the helper 3633 * relies on the neighbor lookup for the L2 address of the nexthop. 3634 * 3635 * The helper will perform a FIB lookup based on the skb's 3636 * networking header to get the address of the next hop, unless 3637 * this is supplied by the caller in the *params* argument. The 3638 * *plen* argument indicates the len of *params* and should be set 3639 * to 0 if *params* is NULL. 3640 * 3641 * The *flags* argument is reserved and must be 0. The helper is 3642 * currently only supported for tc BPF program types, and enabled 3643 * for IPv4 and IPv6 protocols. 3644 * 3645 * Returns 3646 * The helper returns **TC_ACT_REDIRECT** on success or 3647 * **TC_ACT_SHOT** on error. 3648 */ 3649 static long (* const bpf_redirect_neigh)(__u32 ifindex, struct bpf_redir_neigh *params, int plen, __u64 flags) = (void *) 152; 3650 3651 /* 3652 * bpf_per_cpu_ptr 3653 * 3654 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 3655 * pointer to the percpu kernel variable on *cpu*. A ksym is an 3656 * extern variable decorated with '__ksym'. For ksym, there is a 3657 * global var (either static or global) defined of the same name 3658 * in the kernel. The ksym is percpu if the global var is percpu. 3659 * The returned pointer points to the global percpu var on *cpu*. 3660 * 3661 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 3662 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 3663 * happens if *cpu* is larger than nr_cpu_ids. The caller of 3664 * bpf_per_cpu_ptr() must check the returned value. 3665 * 3666 * Returns 3667 * A pointer pointing to the kernel percpu variable on *cpu*, or 3668 * NULL, if *cpu* is invalid. 3669 */ 3670 static void *(* const bpf_per_cpu_ptr)(const void *percpu_ptr, __u32 cpu) = (void *) 153; 3671 3672 /* 3673 * bpf_this_cpu_ptr 3674 * 3675 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 3676 * pointer to the percpu kernel variable on this cpu. See the 3677 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 3678 * 3679 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 3680 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 3681 * never return NULL. 3682 * 3683 * Returns 3684 * A pointer pointing to the kernel percpu variable on this cpu. 3685 */ 3686 static void *(* const bpf_this_cpu_ptr)(const void *percpu_ptr) = (void *) 154; 3687 3688 /* 3689 * bpf_redirect_peer 3690 * 3691 * Redirect the packet to another net device of index *ifindex*. 3692 * This helper is somewhat similar to **bpf_redirect**\ (), except 3693 * that the redirection happens to the *ifindex*' peer device and 3694 * the netns switch takes place from ingress to ingress without 3695 * going through the CPU's backlog queue. 3696 * 3697 * The *flags* argument is reserved and must be 0. The helper is 3698 * currently only supported for tc BPF program types at the 3699 * ingress hook and for veth and netkit target device types. The 3700 * peer device must reside in a different network namespace. 3701 * 3702 * Returns 3703 * The helper returns **TC_ACT_REDIRECT** on success or 3704 * **TC_ACT_SHOT** on error. 3705 */ 3706 static long (* const bpf_redirect_peer)(__u32 ifindex, __u64 flags) = (void *) 155; 3707 3708 /* 3709 * bpf_task_storage_get 3710 * 3711 * Get a bpf_local_storage from the *task*. 3712 * 3713 * Logically, it could be thought of as getting the value from 3714 * a *map* with *task* as the **key**. From this 3715 * perspective, the usage is not much different from 3716 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 3717 * helper enforces the key must be a task_struct and the map must also 3718 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 3719 * 3720 * Underneath, the value is stored locally at *task* instead of 3721 * the *map*. The *map* is used as the bpf-local-storage 3722 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3723 * searched against all bpf_local_storage residing at *task*. 3724 * 3725 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 3726 * used such that a new bpf_local_storage will be 3727 * created if one does not exist. *value* can be used 3728 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 3729 * the initial value of a bpf_local_storage. If *value* is 3730 * **NULL**, the new bpf_local_storage will be zero initialized. 3731 * 3732 * Returns 3733 * A bpf_local_storage pointer is returned on success. 3734 * 3735 * **NULL** if not found or there was an error in adding 3736 * a new bpf_local_storage. 3737 */ 3738 static void *(* const bpf_task_storage_get)(void *map, struct task_struct *task, void *value, __u64 flags) = (void *) 156; 3739 3740 /* 3741 * bpf_task_storage_delete 3742 * 3743 * Delete a bpf_local_storage from a *task*. 3744 * 3745 * Returns 3746 * 0 on success. 3747 * 3748 * **-ENOENT** if the bpf_local_storage cannot be found. 3749 */ 3750 static long (* const bpf_task_storage_delete)(void *map, struct task_struct *task) = (void *) 157; 3751 3752 /* 3753 * bpf_get_current_task_btf 3754 * 3755 * Return a BTF pointer to the "current" task. 3756 * This pointer can also be used in helpers that accept an 3757 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 3758 * 3759 * Returns 3760 * Pointer to the current task. 3761 */ 3762 static struct task_struct *(* const bpf_get_current_task_btf)(void) = (void *) 158; 3763 3764 /* 3765 * bpf_bprm_opts_set 3766 * 3767 * Set or clear certain options on *bprm*: 3768 * 3769 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 3770 * which sets the **AT_SECURE** auxv for glibc. The bit 3771 * is cleared if the flag is not specified. 3772 * 3773 * Returns 3774 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 3775 */ 3776 static long (* const bpf_bprm_opts_set)(struct linux_binprm *bprm, __u64 flags) = (void *) 159; 3777 3778 /* 3779 * bpf_ktime_get_coarse_ns 3780 * 3781 * Return a coarse-grained version of the time elapsed since 3782 * system boot, in nanoseconds. Does not include time the system 3783 * was suspended. 3784 * 3785 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 3786 * 3787 * Returns 3788 * Current *ktime*. 3789 */ 3790 static __u64 (* const bpf_ktime_get_coarse_ns)(void) = (void *) 160; 3791 3792 /* 3793 * bpf_ima_inode_hash 3794 * 3795 * Returns the stored IMA hash of the *inode* (if it's available). 3796 * If the hash is larger than *size*, then only *size* 3797 * bytes will be copied to *dst* 3798 * 3799 * Returns 3800 * The **hash_algo** is returned on success, 3801 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 3802 * invalid arguments are passed. 3803 */ 3804 static long (* const bpf_ima_inode_hash)(struct inode *inode, void *dst, __u32 size) = (void *) 161; 3805 3806 /* 3807 * bpf_sock_from_file 3808 * 3809 * If the given file represents a socket, returns the associated 3810 * socket. 3811 * 3812 * Returns 3813 * A pointer to a struct socket on success or NULL if the file is 3814 * not a socket. 3815 */ 3816 static struct socket *(* const bpf_sock_from_file)(struct file *file) = (void *) 162; 3817 3818 /* 3819 * bpf_check_mtu 3820 * 3821 * Check packet size against exceeding MTU of net device (based 3822 * on *ifindex*). This helper will likely be used in combination 3823 * with helpers that adjust/change the packet size. 3824 * 3825 * The argument *len_diff* can be used for querying with a planned 3826 * size change. This allows to check MTU prior to changing packet 3827 * ctx. Providing a *len_diff* adjustment that is larger than the 3828 * actual packet size (resulting in negative packet size) will in 3829 * principle not exceed the MTU, which is why it is not considered 3830 * a failure. Other BPF helpers are needed for performing the 3831 * planned size change; therefore the responsibility for catching 3832 * a negative packet size belongs in those helpers. 3833 * 3834 * Specifying *ifindex* zero means the MTU check is performed 3835 * against the current net device. This is practical if this isn't 3836 * used prior to redirect. 3837 * 3838 * On input *mtu_len* must be a valid pointer, else verifier will 3839 * reject BPF program. If the value *mtu_len* is initialized to 3840 * zero then the ctx packet size is use. When value *mtu_len* is 3841 * provided as input this specify the L3 length that the MTU check 3842 * is done against. Remember XDP and TC length operate at L2, but 3843 * this value is L3 as this correlate to MTU and IP-header tot_len 3844 * values which are L3 (similar behavior as bpf_fib_lookup). 3845 * 3846 * The Linux kernel route table can configure MTUs on a more 3847 * specific per route level, which is not provided by this helper. 3848 * For route level MTU checks use the **bpf_fib_lookup**\ () 3849 * helper. 3850 * 3851 * *ctx* is either **struct xdp_md** for XDP programs or 3852 * **struct sk_buff** for tc cls_act programs. 3853 * 3854 * The *flags* argument can be a combination of one or more of the 3855 * following values: 3856 * 3857 * **BPF_MTU_CHK_SEGS** 3858 * This flag will only works for *ctx* **struct sk_buff**. 3859 * If packet context contains extra packet segment buffers 3860 * (often knows as GSO skb), then MTU check is harder to 3861 * check at this point, because in transmit path it is 3862 * possible for the skb packet to get re-segmented 3863 * (depending on net device features). This could still be 3864 * a MTU violation, so this flag enables performing MTU 3865 * check against segments, with a different violation 3866 * return code to tell it apart. Check cannot use len_diff. 3867 * 3868 * On return *mtu_len* pointer contains the MTU value of the net 3869 * device. Remember the net device configured MTU is the L3 size, 3870 * which is returned here and XDP and TC length operate at L2. 3871 * Helper take this into account for you, but remember when using 3872 * MTU value in your BPF-code. 3873 * 3874 * 3875 * Returns 3876 * * 0 on success, and populate MTU value in *mtu_len* pointer. 3877 * 3878 * * < 0 if any input argument is invalid (*mtu_len* not updated) 3879 * 3880 * MTU violations return positive values, but also populate MTU 3881 * value in *mtu_len* pointer, as this can be needed for 3882 * implementing PMTU handing: 3883 * 3884 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 3885 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 3886 */ 3887 static long (* const bpf_check_mtu)(void *ctx, __u32 ifindex, __u32 *mtu_len, __s32 len_diff, __u64 flags) = (void *) 163; 3888 3889 /* 3890 * bpf_for_each_map_elem 3891 * 3892 * For each element in **map**, call **callback_fn** function with 3893 * **map**, **callback_ctx** and other map-specific parameters. 3894 * The **callback_fn** should be a static function and 3895 * the **callback_ctx** should be a pointer to the stack. 3896 * The **flags** is used to control certain aspects of the helper. 3897 * Currently, the **flags** must be 0. 3898 * 3899 * The following are a list of supported map types and their 3900 * respective expected callback signatures: 3901 * 3902 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 3903 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 3904 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 3905 * 3906 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 3907 * 3908 * For per_cpu maps, the map_value is the value on the cpu where the 3909 * bpf_prog is running. 3910 * 3911 * If **callback_fn** return 0, the helper will continue to the next 3912 * element. If return value is 1, the helper will skip the rest of 3913 * elements and return. Other return values are not used now. 3914 * 3915 * 3916 * Returns 3917 * The number of traversed map elements for success, **-EINVAL** for 3918 * invalid **flags**. 3919 */ 3920 static long (* const bpf_for_each_map_elem)(void *map, void *callback_fn, void *callback_ctx, __u64 flags) = (void *) 164; 3921 3922 /* 3923 * bpf_snprintf 3924 * 3925 * Outputs a string into the **str** buffer of size **str_size** 3926 * based on a format string stored in a read-only map pointed by 3927 * **fmt**. 3928 * 3929 * Each format specifier in **fmt** corresponds to one u64 element 3930 * in the **data** array. For strings and pointers where pointees 3931 * are accessed, only the pointer values are stored in the *data* 3932 * array. The *data_len* is the size of *data* in bytes - must be 3933 * a multiple of 8. 3934 * 3935 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 3936 * memory. Reading kernel memory may fail due to either invalid 3937 * address or valid address but requiring a major memory fault. If 3938 * reading kernel memory fails, the string for **%s** will be an 3939 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 3940 * Not returning error to bpf program is consistent with what 3941 * **bpf_trace_printk**\ () does for now. 3942 * 3943 * 3944 * Returns 3945 * The strictly positive length of the formatted string, including 3946 * the trailing zero character. If the return value is greater than 3947 * **str_size**, **str** contains a truncated string, guaranteed to 3948 * be zero-terminated except when **str_size** is 0. 3949 * 3950 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 3951 */ 3952 static long (* const bpf_snprintf)(char *str, __u32 str_size, const char *fmt, __u64 *data, __u32 data_len) = (void *) 165; 3953 3954 /* 3955 * bpf_sys_bpf 3956 * 3957 * Execute bpf syscall with given arguments. 3958 * 3959 * Returns 3960 * A syscall result. 3961 */ 3962 static long (* const bpf_sys_bpf)(__u32 cmd, void *attr, __u32 attr_size) = (void *) 166; 3963 3964 /* 3965 * bpf_btf_find_by_name_kind 3966 * 3967 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 3968 * 3969 * Returns 3970 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 3971 */ 3972 static long (* const bpf_btf_find_by_name_kind)(char *name, int name_sz, __u32 kind, int flags) = (void *) 167; 3973 3974 /* 3975 * bpf_sys_close 3976 * 3977 * Execute close syscall for given FD. 3978 * 3979 * Returns 3980 * A syscall result. 3981 */ 3982 static long (* const bpf_sys_close)(__u32 fd) = (void *) 168; 3983 3984 /* 3985 * bpf_timer_init 3986 * 3987 * Initialize the timer. 3988 * First 4 bits of *flags* specify clockid. 3989 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 3990 * All other bits of *flags* are reserved. 3991 * The verifier will reject the program if *timer* is not from 3992 * the same *map*. 3993 * 3994 * Returns 3995 * 0 on success. 3996 * **-EBUSY** if *timer* is already initialized. 3997 * **-EINVAL** if invalid *flags* are passed. 3998 * **-EPERM** if *timer* is in a map that doesn't have any user references. 3999 * The user space should either hold a file descriptor to a map with timers 4000 * or pin such map in bpffs. When map is unpinned or file descriptor is 4001 * closed all timers in the map will be cancelled and freed. 4002 */ 4003 static long (* const bpf_timer_init)(struct bpf_timer *timer, void *map, __u64 flags) = (void *) 169; 4004 4005 /* 4006 * bpf_timer_set_callback 4007 * 4008 * Configure the timer to call *callback_fn* static function. 4009 * 4010 * Returns 4011 * 0 on success. 4012 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4013 * **-EPERM** if *timer* is in a map that doesn't have any user references. 4014 * The user space should either hold a file descriptor to a map with timers 4015 * or pin such map in bpffs. When map is unpinned or file descriptor is 4016 * closed all timers in the map will be cancelled and freed. 4017 */ 4018 static long (* const bpf_timer_set_callback)(struct bpf_timer *timer, void *callback_fn) = (void *) 170; 4019 4020 /* 4021 * bpf_timer_start 4022 * 4023 * Set timer expiration N nanoseconds from the current time. The 4024 * configured callback will be invoked in soft irq context on some cpu 4025 * and will not repeat unless another bpf_timer_start() is made. 4026 * In such case the next invocation can migrate to a different cpu. 4027 * Since struct bpf_timer is a field inside map element the map 4028 * owns the timer. The bpf_timer_set_callback() will increment refcnt 4029 * of BPF program to make sure that callback_fn code stays valid. 4030 * When user space reference to a map reaches zero all timers 4031 * in a map are cancelled and corresponding program's refcnts are 4032 * decremented. This is done to make sure that Ctrl-C of a user 4033 * process doesn't leave any timers running. If map is pinned in 4034 * bpffs the callback_fn can re-arm itself indefinitely. 4035 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 4036 * cancel and free the timer in the given map element. 4037 * The map can contain timers that invoke callback_fn-s from different 4038 * programs. The same callback_fn can serve different timers from 4039 * different maps if key/value layout matches across maps. 4040 * Every bpf_timer_set_callback() can have different callback_fn. 4041 * 4042 * *flags* can be one of: 4043 * 4044 * **BPF_F_TIMER_ABS** 4045 * Start the timer in absolute expire value instead of the 4046 * default relative one. 4047 * **BPF_F_TIMER_CPU_PIN** 4048 * Timer will be pinned to the CPU of the caller. 4049 * 4050 * 4051 * Returns 4052 * 0 on success. 4053 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 4054 * or invalid *flags* are passed. 4055 */ 4056 static long (* const bpf_timer_start)(struct bpf_timer *timer, __u64 nsecs, __u64 flags) = (void *) 171; 4057 4058 /* 4059 * bpf_timer_cancel 4060 * 4061 * Cancel the timer and wait for callback_fn to finish if it was running. 4062 * 4063 * Returns 4064 * 0 if the timer was not active. 4065 * 1 if the timer was active. 4066 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4067 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 4068 * own timer which would have led to a deadlock otherwise. 4069 */ 4070 static long (* const bpf_timer_cancel)(struct bpf_timer *timer) = (void *) 172; 4071 4072 /* 4073 * bpf_get_func_ip 4074 * 4075 * Get address of the traced function (for tracing and kprobe programs). 4076 * 4077 * When called for kprobe program attached as uprobe it returns 4078 * probe address for both entry and return uprobe. 4079 * 4080 * 4081 * Returns 4082 * Address of the traced function for kprobe. 4083 * 0 for kprobes placed within the function (not at the entry). 4084 * Address of the probe for uprobe and return uprobe. 4085 */ 4086 static __u64 (* const bpf_get_func_ip)(void *ctx) = (void *) 173; 4087 4088 /* 4089 * bpf_get_attach_cookie 4090 * 4091 * Get bpf_cookie value provided (optionally) during the program 4092 * attachment. It might be different for each individual 4093 * attachment, even if BPF program itself is the same. 4094 * Expects BPF program context *ctx* as a first argument. 4095 * 4096 * Supported for the following program types: 4097 * - kprobe/uprobe; 4098 * - tracepoint; 4099 * - perf_event. 4100 * 4101 * Returns 4102 * Value specified by user at BPF link creation/attachment time 4103 * or 0, if it was not specified. 4104 */ 4105 static __u64 (* const bpf_get_attach_cookie)(void *ctx) = (void *) 174; 4106 4107 /* 4108 * bpf_task_pt_regs 4109 * 4110 * Get the struct pt_regs associated with **task**. 4111 * 4112 * Returns 4113 * A pointer to struct pt_regs. 4114 */ 4115 static long (* const bpf_task_pt_regs)(struct task_struct *task) = (void *) 175; 4116 4117 /* 4118 * bpf_get_branch_snapshot 4119 * 4120 * Get branch trace from hardware engines like Intel LBR. The 4121 * hardware engine is stopped shortly after the helper is 4122 * called. Therefore, the user need to filter branch entries 4123 * based on the actual use case. To capture branch trace 4124 * before the trigger point of the BPF program, the helper 4125 * should be called at the beginning of the BPF program. 4126 * 4127 * The data is stored as struct perf_branch_entry into output 4128 * buffer *entries*. *size* is the size of *entries* in bytes. 4129 * *flags* is reserved for now and must be zero. 4130 * 4131 * 4132 * Returns 4133 * On success, number of bytes written to *buf*. On error, a 4134 * negative value. 4135 * 4136 * **-EINVAL** if *flags* is not zero. 4137 * 4138 * **-ENOENT** if architecture does not support branch records. 4139 */ 4140 static long (* const bpf_get_branch_snapshot)(void *entries, __u32 size, __u64 flags) = (void *) 176; 4141 4142 /* 4143 * bpf_trace_vprintk 4144 * 4145 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 4146 * to format and can handle more format args as a result. 4147 * 4148 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 4149 * 4150 * Returns 4151 * The number of bytes written to the buffer, or a negative error 4152 * in case of failure. 4153 */ 4154 static long (* const bpf_trace_vprintk)(const char *fmt, __u32 fmt_size, const void *data, __u32 data_len) = (void *) 177; 4155 4156 /* 4157 * bpf_skc_to_unix_sock 4158 * 4159 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 4160 * 4161 * Returns 4162 * *sk* if casting is valid, or **NULL** otherwise. 4163 */ 4164 static struct unix_sock *(* const bpf_skc_to_unix_sock)(void *sk) = (void *) 178; 4165 4166 /* 4167 * bpf_kallsyms_lookup_name 4168 * 4169 * Get the address of a kernel symbol, returned in *res*. *res* is 4170 * set to 0 if the symbol is not found. 4171 * 4172 * Returns 4173 * On success, zero. On error, a negative value. 4174 * 4175 * **-EINVAL** if *flags* is not zero. 4176 * 4177 * **-EINVAL** if string *name* is not the same size as *name_sz*. 4178 * 4179 * **-ENOENT** if symbol is not found. 4180 * 4181 * **-EPERM** if caller does not have permission to obtain kernel address. 4182 */ 4183 static long (* const bpf_kallsyms_lookup_name)(const char *name, int name_sz, int flags, __u64 *res) = (void *) 179; 4184 4185 /* 4186 * bpf_find_vma 4187 * 4188 * Find vma of *task* that contains *addr*, call *callback_fn* 4189 * function with *task*, *vma*, and *callback_ctx*. 4190 * The *callback_fn* should be a static function and 4191 * the *callback_ctx* should be a pointer to the stack. 4192 * The *flags* is used to control certain aspects of the helper. 4193 * Currently, the *flags* must be 0. 4194 * 4195 * The expected callback signature is 4196 * 4197 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 4198 * 4199 * 4200 * Returns 4201 * 0 on success. 4202 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 4203 * **-EBUSY** if failed to try lock mmap_lock. 4204 * **-EINVAL** for invalid **flags**. 4205 */ 4206 static long (* const bpf_find_vma)(struct task_struct *task, __u64 addr, void *callback_fn, void *callback_ctx, __u64 flags) = (void *) 180; 4207 4208 /* 4209 * bpf_loop 4210 * 4211 * For **nr_loops**, call **callback_fn** function 4212 * with **callback_ctx** as the context parameter. 4213 * The **callback_fn** should be a static function and 4214 * the **callback_ctx** should be a pointer to the stack. 4215 * The **flags** is used to control certain aspects of the helper. 4216 * Currently, the **flags** must be 0. Currently, nr_loops is 4217 * limited to 1 << 23 (~8 million) loops. 4218 * 4219 * long (\*callback_fn)(u32 index, void \*ctx); 4220 * 4221 * where **index** is the current index in the loop. The index 4222 * is zero-indexed. 4223 * 4224 * If **callback_fn** returns 0, the helper will continue to the next 4225 * loop. If return value is 1, the helper will skip the rest of 4226 * the loops and return. Other return values are not used now, 4227 * and will be rejected by the verifier. 4228 * 4229 * 4230 * Returns 4231 * The number of loops performed, **-EINVAL** for invalid **flags**, 4232 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 4233 */ 4234 static long (* const bpf_loop)(__u32 nr_loops, void *callback_fn, void *callback_ctx, __u64 flags) = (void *) 181; 4235 4236 /* 4237 * bpf_strncmp 4238 * 4239 * Do strncmp() between **s1** and **s2**. **s1** doesn't need 4240 * to be null-terminated and **s1_sz** is the maximum storage 4241 * size of **s1**. **s2** must be a read-only string. 4242 * 4243 * Returns 4244 * An integer less than, equal to, or greater than zero 4245 * if the first **s1_sz** bytes of **s1** is found to be 4246 * less than, to match, or be greater than **s2**. 4247 */ 4248 static long (* const bpf_strncmp)(const char *s1, __u32 s1_sz, const char *s2) = (void *) 182; 4249 4250 /* 4251 * bpf_get_func_arg 4252 * 4253 * Get **n**-th argument register (zero based) of the traced function (for tracing programs) 4254 * returned in **value**. 4255 * 4256 * 4257 * Returns 4258 * 0 on success. 4259 * **-EINVAL** if n >= argument register count of traced function. 4260 */ 4261 static long (* const bpf_get_func_arg)(void *ctx, __u32 n, __u64 *value) = (void *) 183; 4262 4263 /* 4264 * bpf_get_func_ret 4265 * 4266 * Get return value of the traced function (for tracing programs) 4267 * in **value**. 4268 * 4269 * 4270 * Returns 4271 * 0 on success. 4272 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 4273 */ 4274 static long (* const bpf_get_func_ret)(void *ctx, __u64 *value) = (void *) 184; 4275 4276 /* 4277 * bpf_get_func_arg_cnt 4278 * 4279 * Get number of registers of the traced function (for tracing programs) where 4280 * function arguments are stored in these registers. 4281 * 4282 * 4283 * Returns 4284 * The number of argument registers of the traced function. 4285 */ 4286 static long (* const bpf_get_func_arg_cnt)(void *ctx) = (void *) 185; 4287 4288 /* 4289 * bpf_get_retval 4290 * 4291 * Get the BPF program's return value that will be returned to the upper layers. 4292 * 4293 * This helper is currently supported by cgroup programs and only by the hooks 4294 * where BPF program's return value is returned to the userspace via errno. 4295 * 4296 * Returns 4297 * The BPF program's return value. 4298 */ 4299 static int (* const bpf_get_retval)(void) = (void *) 186; 4300 4301 /* 4302 * bpf_set_retval 4303 * 4304 * Set the BPF program's return value that will be returned to the upper layers. 4305 * 4306 * This helper is currently supported by cgroup programs and only by the hooks 4307 * where BPF program's return value is returned to the userspace via errno. 4308 * 4309 * Note that there is the following corner case where the program exports an error 4310 * via bpf_set_retval but signals success via 'return 1': 4311 * 4312 * bpf_set_retval(-EPERM); 4313 * return 1; 4314 * 4315 * In this case, the BPF program's return value will use helper's -EPERM. This 4316 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case. 4317 * 4318 * 4319 * Returns 4320 * 0 on success, or a negative error in case of failure. 4321 */ 4322 static int (* const bpf_set_retval)(int retval) = (void *) 187; 4323 4324 /* 4325 * bpf_xdp_get_buff_len 4326 * 4327 * Get the total size of a given xdp buff (linear and paged area) 4328 * 4329 * Returns 4330 * The total size of a given xdp buffer. 4331 */ 4332 static __u64 (* const bpf_xdp_get_buff_len)(struct xdp_md *xdp_md) = (void *) 188; 4333 4334 /* 4335 * bpf_xdp_load_bytes 4336 * 4337 * This helper is provided as an easy way to load data from a 4338 * xdp buffer. It can be used to load *len* bytes from *offset* from 4339 * the frame associated to *xdp_md*, into the buffer pointed by 4340 * *buf*. 4341 * 4342 * Returns 4343 * 0 on success, or a negative error in case of failure. 4344 */ 4345 static long (* const bpf_xdp_load_bytes)(struct xdp_md *xdp_md, __u32 offset, void *buf, __u32 len) = (void *) 189; 4346 4347 /* 4348 * bpf_xdp_store_bytes 4349 * 4350 * Store *len* bytes from buffer *buf* into the frame 4351 * associated to *xdp_md*, at *offset*. 4352 * 4353 * Returns 4354 * 0 on success, or a negative error in case of failure. 4355 */ 4356 static long (* const bpf_xdp_store_bytes)(struct xdp_md *xdp_md, __u32 offset, void *buf, __u32 len) = (void *) 190; 4357 4358 /* 4359 * bpf_copy_from_user_task 4360 * 4361 * Read *size* bytes from user space address *user_ptr* in *tsk*'s 4362 * address space, and stores the data in *dst*. *flags* is not 4363 * used yet and is provided for future extensibility. This helper 4364 * can only be used by sleepable programs. 4365 * 4366 * Returns 4367 * 0 on success, or a negative error in case of failure. On error 4368 * *dst* buffer is zeroed out. 4369 */ 4370 static long (* const bpf_copy_from_user_task)(void *dst, __u32 size, const void *user_ptr, struct task_struct *tsk, __u64 flags) = (void *) 191; 4371 4372 /* 4373 * bpf_skb_set_tstamp 4374 * 4375 * Change the __sk_buff->tstamp_type to *tstamp_type* 4376 * and set *tstamp* to the __sk_buff->tstamp together. 4377 * 4378 * If there is no need to change the __sk_buff->tstamp_type, 4379 * the tstamp value can be directly written to __sk_buff->tstamp 4380 * instead. 4381 * 4382 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 4383 * will be kept during bpf_redirect_*(). A non zero 4384 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 4385 * *tstamp_type*. 4386 * 4387 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 4388 * with a zero *tstamp*. 4389 * 4390 * Only IPv4 and IPv6 skb->protocol are supported. 4391 * 4392 * This function is most useful when it needs to set a 4393 * mono delivery time to __sk_buff->tstamp and then 4394 * bpf_redirect_*() to the egress of an iface. For example, 4395 * changing the (rcv) timestamp in __sk_buff->tstamp at 4396 * ingress to a mono delivery time and then bpf_redirect_*() 4397 * to sch_fq@phy-dev. 4398 * 4399 * Returns 4400 * 0 on success. 4401 * **-EINVAL** for invalid input 4402 * **-EOPNOTSUPP** for unsupported protocol 4403 */ 4404 static long (* const bpf_skb_set_tstamp)(struct __sk_buff *skb, __u64 tstamp, __u32 tstamp_type) = (void *) 192; 4405 4406 /* 4407 * bpf_ima_file_hash 4408 * 4409 * Returns a calculated IMA hash of the *file*. 4410 * If the hash is larger than *size*, then only *size* 4411 * bytes will be copied to *dst* 4412 * 4413 * Returns 4414 * The **hash_algo** is returned on success, 4415 * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if 4416 * invalid arguments are passed. 4417 */ 4418 static long (* const bpf_ima_file_hash)(struct file *file, void *dst, __u32 size) = (void *) 193; 4419 4420 /* 4421 * bpf_kptr_xchg 4422 * 4423 * Exchange kptr at pointer *map_value* with *ptr*, and return the 4424 * old value. *ptr* can be NULL, otherwise it must be a referenced 4425 * pointer which will be released when this helper is called. 4426 * 4427 * Returns 4428 * The old value of kptr (which can be NULL). The returned pointer 4429 * if not NULL, is a reference which must be released using its 4430 * corresponding release function, or moved into a BPF map before 4431 * program exit. 4432 */ 4433 static void *(* const bpf_kptr_xchg)(void *map_value, void *ptr) = (void *) 194; 4434 4435 /* 4436 * bpf_map_lookup_percpu_elem 4437 * 4438 * Perform a lookup in *percpu map* for an entry associated to 4439 * *key* on *cpu*. 4440 * 4441 * Returns 4442 * Map value associated to *key* on *cpu*, or **NULL** if no entry 4443 * was found or *cpu* is invalid. 4444 */ 4445 static void *(* const bpf_map_lookup_percpu_elem)(void *map, const void *key, __u32 cpu) = (void *) 195; 4446 4447 /* 4448 * bpf_skc_to_mptcp_sock 4449 * 4450 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 4451 * 4452 * Returns 4453 * *sk* if casting is valid, or **NULL** otherwise. 4454 */ 4455 static struct mptcp_sock *(* const bpf_skc_to_mptcp_sock)(void *sk) = (void *) 196; 4456 4457 /* 4458 * bpf_dynptr_from_mem 4459 * 4460 * Get a dynptr to local memory *data*. 4461 * 4462 * *data* must be a ptr to a map value. 4463 * The maximum *size* supported is DYNPTR_MAX_SIZE. 4464 * *flags* is currently unused. 4465 * 4466 * Returns 4467 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 4468 * -EINVAL if flags is not 0. 4469 */ 4470 static long (* const bpf_dynptr_from_mem)(void *data, __u32 size, __u64 flags, struct bpf_dynptr *ptr) = (void *) 197; 4471 4472 /* 4473 * bpf_ringbuf_reserve_dynptr 4474 * 4475 * Reserve *size* bytes of payload in a ring buffer *ringbuf* 4476 * through the dynptr interface. *flags* must be 0. 4477 * 4478 * Please note that a corresponding bpf_ringbuf_submit_dynptr or 4479 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 4480 * reservation fails. This is enforced by the verifier. 4481 * 4482 * Returns 4483 * 0 on success, or a negative error in case of failure. 4484 */ 4485 static long (* const bpf_ringbuf_reserve_dynptr)(void *ringbuf, __u32 size, __u64 flags, struct bpf_dynptr *ptr) = (void *) 198; 4486 4487 /* 4488 * bpf_ringbuf_submit_dynptr 4489 * 4490 * Submit reserved ring buffer sample, pointed to by *data*, 4491 * through the dynptr interface. This is a no-op if the dynptr is 4492 * invalid/null. 4493 * 4494 * For more information on *flags*, please see 4495 * 'bpf_ringbuf_submit'. 4496 * 4497 * Returns 4498 * Nothing. Always succeeds. 4499 */ 4500 static void (* const bpf_ringbuf_submit_dynptr)(struct bpf_dynptr *ptr, __u64 flags) = (void *) 199; 4501 4502 /* 4503 * bpf_ringbuf_discard_dynptr 4504 * 4505 * Discard reserved ring buffer sample through the dynptr 4506 * interface. This is a no-op if the dynptr is invalid/null. 4507 * 4508 * For more information on *flags*, please see 4509 * 'bpf_ringbuf_discard'. 4510 * 4511 * Returns 4512 * Nothing. Always succeeds. 4513 */ 4514 static void (* const bpf_ringbuf_discard_dynptr)(struct bpf_dynptr *ptr, __u64 flags) = (void *) 200; 4515 4516 /* 4517 * bpf_dynptr_read 4518 * 4519 * Read *len* bytes from *src* into *dst*, starting from *offset* 4520 * into *src*. 4521 * *flags* is currently unused. 4522 * 4523 * Returns 4524 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 4525 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 4526 * *flags* is not 0. 4527 */ 4528 static long (* const bpf_dynptr_read)(void *dst, __u32 len, const struct bpf_dynptr *src, __u32 offset, __u64 flags) = (void *) 201; 4529 4530 /* 4531 * bpf_dynptr_write 4532 * 4533 * Write *len* bytes from *src* into *dst*, starting from *offset* 4534 * into *dst*. 4535 * 4536 * *flags* must be 0 except for skb-type dynptrs. 4537 * 4538 * For skb-type dynptrs: 4539 * * All data slices of the dynptr are automatically 4540 * invalidated after **bpf_dynptr_write**\ (). This is 4541 * because writing may pull the skb and change the 4542 * underlying packet buffer. 4543 * 4544 * * For *flags*, please see the flags accepted by 4545 * **bpf_skb_store_bytes**\ (). 4546 * 4547 * Returns 4548 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 4549 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 4550 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs, 4551 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ (). 4552 */ 4553 static long (* const bpf_dynptr_write)(const struct bpf_dynptr *dst, __u32 offset, void *src, __u32 len, __u64 flags) = (void *) 202; 4554 4555 /* 4556 * bpf_dynptr_data 4557 * 4558 * Get a pointer to the underlying dynptr data. 4559 * 4560 * *len* must be a statically known value. The returned data slice 4561 * is invalidated whenever the dynptr is invalidated. 4562 * 4563 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should 4564 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr. 4565 * 4566 * Returns 4567 * Pointer to the underlying dynptr data, NULL if the dynptr is 4568 * read-only, if the dynptr is invalid, or if the offset and length 4569 * is out of bounds. 4570 */ 4571 static void *(* const bpf_dynptr_data)(const struct bpf_dynptr *ptr, __u32 offset, __u32 len) = (void *) 203; 4572 4573 /* 4574 * bpf_tcp_raw_gen_syncookie_ipv4 4575 * 4576 * Try to issue a SYN cookie for the packet with corresponding 4577 * IPv4/TCP headers, *iph* and *th*, without depending on a 4578 * listening socket. 4579 * 4580 * *iph* points to the IPv4 header. 4581 * 4582 * *th* points to the start of the TCP header, while *th_len* 4583 * contains the length of the TCP header (at least 4584 * **sizeof**\ (**struct tcphdr**)). 4585 * 4586 * Returns 4587 * On success, lower 32 bits hold the generated SYN cookie in 4588 * followed by 16 bits which hold the MSS value for that cookie, 4589 * and the top 16 bits are unused. 4590 * 4591 * On failure, the returned value is one of the following: 4592 * 4593 * **-EINVAL** if *th_len* is invalid. 4594 */ 4595 static __s64 (* const bpf_tcp_raw_gen_syncookie_ipv4)(struct iphdr *iph, struct tcphdr *th, __u32 th_len) = (void *) 204; 4596 4597 /* 4598 * bpf_tcp_raw_gen_syncookie_ipv6 4599 * 4600 * Try to issue a SYN cookie for the packet with corresponding 4601 * IPv6/TCP headers, *iph* and *th*, without depending on a 4602 * listening socket. 4603 * 4604 * *iph* points to the IPv6 header. 4605 * 4606 * *th* points to the start of the TCP header, while *th_len* 4607 * contains the length of the TCP header (at least 4608 * **sizeof**\ (**struct tcphdr**)). 4609 * 4610 * Returns 4611 * On success, lower 32 bits hold the generated SYN cookie in 4612 * followed by 16 bits which hold the MSS value for that cookie, 4613 * and the top 16 bits are unused. 4614 * 4615 * On failure, the returned value is one of the following: 4616 * 4617 * **-EINVAL** if *th_len* is invalid. 4618 * 4619 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 4620 */ 4621 static __s64 (* const bpf_tcp_raw_gen_syncookie_ipv6)(struct ipv6hdr *iph, struct tcphdr *th, __u32 th_len) = (void *) 205; 4622 4623 /* 4624 * bpf_tcp_raw_check_syncookie_ipv4 4625 * 4626 * Check whether *iph* and *th* contain a valid SYN cookie ACK 4627 * without depending on a listening socket. 4628 * 4629 * *iph* points to the IPv4 header. 4630 * 4631 * *th* points to the TCP header. 4632 * 4633 * Returns 4634 * 0 if *iph* and *th* are a valid SYN cookie ACK. 4635 * 4636 * On failure, the returned value is one of the following: 4637 * 4638 * **-EACCES** if the SYN cookie is not valid. 4639 */ 4640 static long (* const bpf_tcp_raw_check_syncookie_ipv4)(struct iphdr *iph, struct tcphdr *th) = (void *) 206; 4641 4642 /* 4643 * bpf_tcp_raw_check_syncookie_ipv6 4644 * 4645 * Check whether *iph* and *th* contain a valid SYN cookie ACK 4646 * without depending on a listening socket. 4647 * 4648 * *iph* points to the IPv6 header. 4649 * 4650 * *th* points to the TCP header. 4651 * 4652 * Returns 4653 * 0 if *iph* and *th* are a valid SYN cookie ACK. 4654 * 4655 * On failure, the returned value is one of the following: 4656 * 4657 * **-EACCES** if the SYN cookie is not valid. 4658 * 4659 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 4660 */ 4661 static long (* const bpf_tcp_raw_check_syncookie_ipv6)(struct ipv6hdr *iph, struct tcphdr *th) = (void *) 207; 4662 4663 /* 4664 * bpf_ktime_get_tai_ns 4665 * 4666 * A nonsettable system-wide clock derived from wall-clock time but 4667 * ignoring leap seconds. This clock does not experience 4668 * discontinuities and backwards jumps caused by NTP inserting leap 4669 * seconds as CLOCK_REALTIME does. 4670 * 4671 * See: **clock_gettime**\ (**CLOCK_TAI**) 4672 * 4673 * Returns 4674 * Current *ktime*. 4675 */ 4676 static __u64 (* const bpf_ktime_get_tai_ns)(void) = (void *) 208; 4677 4678 /* 4679 * bpf_user_ringbuf_drain 4680 * 4681 * Drain samples from the specified user ring buffer, and invoke 4682 * the provided callback for each such sample: 4683 * 4684 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx); 4685 * 4686 * If **callback_fn** returns 0, the helper will continue to try 4687 * and drain the next sample, up to a maximum of 4688 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1, 4689 * the helper will skip the rest of the samples and return. Other 4690 * return values are not used now, and will be rejected by the 4691 * verifier. 4692 * 4693 * Returns 4694 * The number of drained samples if no error was encountered while 4695 * draining samples, or 0 if no samples were present in the ring 4696 * buffer. If a user-space producer was epoll-waiting on this map, 4697 * and at least one sample was drained, they will receive an event 4698 * notification notifying them of available space in the ring 4699 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this 4700 * function, no wakeup notification will be sent. If the 4701 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will 4702 * be sent even if no sample was drained. 4703 * 4704 * On failure, the returned value is one of the following: 4705 * 4706 * **-EBUSY** if the ring buffer is contended, and another calling 4707 * context was concurrently draining the ring buffer. 4708 * 4709 * **-EINVAL** if user-space is not properly tracking the ring 4710 * buffer due to the producer position not being aligned to 8 4711 * bytes, a sample not being aligned to 8 bytes, or the producer 4712 * position not matching the advertised length of a sample. 4713 * 4714 * **-E2BIG** if user-space has tried to publish a sample which is 4715 * larger than the size of the ring buffer, or which cannot fit 4716 * within a struct bpf_dynptr. 4717 */ 4718 static long (* const bpf_user_ringbuf_drain)(void *map, void *callback_fn, void *ctx, __u64 flags) = (void *) 209; 4719 4720 /* 4721 * bpf_cgrp_storage_get 4722 * 4723 * Get a bpf_local_storage from the *cgroup*. 4724 * 4725 * Logically, it could be thought of as getting the value from 4726 * a *map* with *cgroup* as the **key**. From this 4727 * perspective, the usage is not much different from 4728 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this 4729 * helper enforces the key must be a cgroup struct and the map must also 4730 * be a **BPF_MAP_TYPE_CGRP_STORAGE**. 4731 * 4732 * In reality, the local-storage value is embedded directly inside of the 4733 * *cgroup* object itself, rather than being located in the 4734 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is 4735 * queried for some *map* on a *cgroup* object, the kernel will perform an 4736 * O(n) iteration over all of the live local-storage values for that 4737 * *cgroup* object until the local-storage value for the *map* is found. 4738 * 4739 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4740 * used such that a new bpf_local_storage will be 4741 * created if one does not exist. *value* can be used 4742 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4743 * the initial value of a bpf_local_storage. If *value* is 4744 * **NULL**, the new bpf_local_storage will be zero initialized. 4745 * 4746 * Returns 4747 * A bpf_local_storage pointer is returned on success. 4748 * 4749 * **NULL** if not found or there was an error in adding 4750 * a new bpf_local_storage. 4751 */ 4752 static void *(* const bpf_cgrp_storage_get)(void *map, struct cgroup *cgroup, void *value, __u64 flags) = (void *) 210; 4753 4754 /* 4755 * bpf_cgrp_storage_delete 4756 * 4757 * Delete a bpf_local_storage from a *cgroup*. 4758 * 4759 * Returns 4760 * 0 on success. 4761 * 4762 * **-ENOENT** if the bpf_local_storage cannot be found. 4763 */ 4764 static long (* const bpf_cgrp_storage_delete)(void *map, struct cgroup *cgroup) = (void *) 211; 4765 4766 4767