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