1e2be04c7SGreg Kroah-Hartman /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2daedfb22SAlexei Starovoitov /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3daedfb22SAlexei Starovoitov * 4daedfb22SAlexei Starovoitov * This program is free software; you can redistribute it and/or 5daedfb22SAlexei Starovoitov * modify it under the terms of version 2 of the GNU General Public 6daedfb22SAlexei Starovoitov * License as published by the Free Software Foundation. 7daedfb22SAlexei Starovoitov */ 8daedfb22SAlexei Starovoitov #ifndef _UAPI__LINUX_BPF_H__ 9daedfb22SAlexei Starovoitov #define _UAPI__LINUX_BPF_H__ 10daedfb22SAlexei Starovoitov 11daedfb22SAlexei Starovoitov #include <linux/types.h> 12c15952dcSAlexei Starovoitov #include <linux/bpf_common.h> 13daedfb22SAlexei Starovoitov 14daedfb22SAlexei Starovoitov /* Extended instruction set based on top of classic BPF */ 15daedfb22SAlexei Starovoitov 16daedfb22SAlexei Starovoitov /* instruction classes */ 17d405c740SJiong Wang #define BPF_JMP32 0x06 /* jmp mode in word width */ 18daedfb22SAlexei Starovoitov #define BPF_ALU64 0x07 /* alu mode in double word width */ 19daedfb22SAlexei Starovoitov 20daedfb22SAlexei Starovoitov /* ld/ldx fields */ 21cb5f7334SJesper Dangaard Brouer #define BPF_DW 0x18 /* double word (64-bit) */ 2291c960b0SBrendan Jackman #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */ 2391c960b0SBrendan Jackman #define BPF_XADD 0xc0 /* exclusive add - legacy name */ 24daedfb22SAlexei Starovoitov 25daedfb22SAlexei Starovoitov /* alu/jmp fields */ 26daedfb22SAlexei Starovoitov #define BPF_MOV 0xb0 /* mov reg to reg */ 27daedfb22SAlexei Starovoitov #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 28daedfb22SAlexei Starovoitov 29daedfb22SAlexei Starovoitov /* change endianness of a register */ 30daedfb22SAlexei Starovoitov #define BPF_END 0xd0 /* flags for endianness conversion: */ 31daedfb22SAlexei Starovoitov #define BPF_TO_LE 0x00 /* convert to little-endian */ 32daedfb22SAlexei Starovoitov #define BPF_TO_BE 0x08 /* convert to big-endian */ 33daedfb22SAlexei Starovoitov #define BPF_FROM_LE BPF_TO_LE 34daedfb22SAlexei Starovoitov #define BPF_FROM_BE BPF_TO_BE 35daedfb22SAlexei Starovoitov 3692b31a9aSDaniel Borkmann /* jmp encodings */ 37daedfb22SAlexei Starovoitov #define BPF_JNE 0x50 /* jump != */ 3892b31a9aSDaniel Borkmann #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 3992b31a9aSDaniel Borkmann #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 40daedfb22SAlexei Starovoitov #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 41daedfb22SAlexei Starovoitov #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 4292b31a9aSDaniel Borkmann #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 4392b31a9aSDaniel Borkmann #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 44daedfb22SAlexei Starovoitov #define BPF_CALL 0x80 /* function call */ 45daedfb22SAlexei Starovoitov #define BPF_EXIT 0x90 /* function return */ 46daedfb22SAlexei Starovoitov 475ca419f2SBrendan Jackman /* atomic op type fields (stored in immediate) */ 485ffa2550SBrendan Jackman #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */ 495ffa2550SBrendan Jackman #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */ 505ffa2550SBrendan Jackman #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */ 515ca419f2SBrendan Jackman 52daedfb22SAlexei Starovoitov /* Register numbers */ 53daedfb22SAlexei Starovoitov enum { 54daedfb22SAlexei Starovoitov BPF_REG_0 = 0, 55daedfb22SAlexei Starovoitov BPF_REG_1, 56daedfb22SAlexei Starovoitov BPF_REG_2, 57daedfb22SAlexei Starovoitov BPF_REG_3, 58daedfb22SAlexei Starovoitov BPF_REG_4, 59daedfb22SAlexei Starovoitov BPF_REG_5, 60daedfb22SAlexei Starovoitov BPF_REG_6, 61daedfb22SAlexei Starovoitov BPF_REG_7, 62daedfb22SAlexei Starovoitov BPF_REG_8, 63daedfb22SAlexei Starovoitov BPF_REG_9, 64daedfb22SAlexei Starovoitov BPF_REG_10, 65daedfb22SAlexei Starovoitov __MAX_BPF_REG, 66daedfb22SAlexei Starovoitov }; 67daedfb22SAlexei Starovoitov 68daedfb22SAlexei Starovoitov /* BPF has 10 general purpose 64-bit registers and stack frame. */ 69daedfb22SAlexei Starovoitov #define MAX_BPF_REG __MAX_BPF_REG 70daedfb22SAlexei Starovoitov 71daedfb22SAlexei Starovoitov struct bpf_insn { 72daedfb22SAlexei Starovoitov __u8 code; /* opcode */ 73daedfb22SAlexei Starovoitov __u8 dst_reg:4; /* dest register */ 74daedfb22SAlexei Starovoitov __u8 src_reg:4; /* source register */ 75daedfb22SAlexei Starovoitov __s16 off; /* signed offset */ 76daedfb22SAlexei Starovoitov __s32 imm; /* signed immediate constant */ 77daedfb22SAlexei Starovoitov }; 78daedfb22SAlexei Starovoitov 79b95a5c4dSDaniel Mack /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 80b95a5c4dSDaniel Mack struct bpf_lpm_trie_key { 81b95a5c4dSDaniel Mack __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 82*3024d95aSDaniel Borkmann __u8 data[0]; /* Arbitrary size */ 83b95a5c4dSDaniel Mack }; 84b95a5c4dSDaniel Mack 85de9cbbaaSRoman Gushchin struct bpf_cgroup_storage_key { 86de9cbbaaSRoman Gushchin __u64 cgroup_inode_id; /* cgroup inode id */ 876fc88c35SDave Marchevsky __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 88de9cbbaaSRoman Gushchin }; 89de9cbbaaSRoman Gushchin 905e7b3020SYonghong Song union bpf_iter_link_info { 915e7b3020SYonghong Song struct { 925e7b3020SYonghong Song __u32 map_fd; 935e7b3020SYonghong Song } map; 945e7b3020SYonghong Song }; 955e7b3020SYonghong Song 967799e4d9SJoe Stringer /* BPF syscall commands, see bpf(2) man-page for more details. */ 977799e4d9SJoe Stringer /** 987799e4d9SJoe Stringer * DOC: eBPF Syscall Preamble 997799e4d9SJoe Stringer * 1007799e4d9SJoe Stringer * The operation to be performed by the **bpf**\ () system call is determined 1017799e4d9SJoe Stringer * by the *cmd* argument. Each operation takes an accompanying argument, 1027799e4d9SJoe Stringer * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 1037799e4d9SJoe Stringer * below). The size argument is the size of the union pointed to by *attr*. 1047799e4d9SJoe Stringer */ 1057799e4d9SJoe Stringer /** 1067799e4d9SJoe Stringer * DOC: eBPF Syscall Commands 1077799e4d9SJoe Stringer * 1087799e4d9SJoe Stringer * BPF_MAP_CREATE 1097799e4d9SJoe Stringer * Description 1107799e4d9SJoe Stringer * Create a map and return a file descriptor that refers to the 1117799e4d9SJoe Stringer * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 1127799e4d9SJoe Stringer * is automatically enabled for the new file descriptor. 1137799e4d9SJoe Stringer * 1147799e4d9SJoe Stringer * Applying **close**\ (2) to the file descriptor returned by 1157799e4d9SJoe Stringer * **BPF_MAP_CREATE** will delete the map (but see NOTES). 1167799e4d9SJoe Stringer * 1177799e4d9SJoe Stringer * Return 1187799e4d9SJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 1197799e4d9SJoe Stringer * error occurred (in which case, *errno* is set appropriately). 1207799e4d9SJoe Stringer * 1217799e4d9SJoe Stringer * BPF_MAP_LOOKUP_ELEM 1227799e4d9SJoe Stringer * Description 1237799e4d9SJoe Stringer * Look up an element with a given *key* in the map referred to 1247799e4d9SJoe Stringer * by the file descriptor *map_fd*. 1257799e4d9SJoe Stringer * 1266690523bSJoe Stringer * The *flags* argument may be specified as one of the 1276690523bSJoe Stringer * following: 1286690523bSJoe Stringer * 1296690523bSJoe Stringer * **BPF_F_LOCK** 1306690523bSJoe Stringer * Look up the value of a spin-locked map without 1316690523bSJoe Stringer * returning the lock. This must be specified if the 1326690523bSJoe Stringer * elements contain a spinlock. 1336690523bSJoe Stringer * 1347799e4d9SJoe Stringer * Return 1357799e4d9SJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 1367799e4d9SJoe Stringer * is set appropriately. 1377799e4d9SJoe Stringer * 1387799e4d9SJoe Stringer * BPF_MAP_UPDATE_ELEM 1397799e4d9SJoe Stringer * Description 1407799e4d9SJoe Stringer * Create or update an element (key/value pair) in a specified map. 1417799e4d9SJoe Stringer * 1427799e4d9SJoe Stringer * The *flags* argument should be specified as one of the 1437799e4d9SJoe Stringer * following: 1447799e4d9SJoe Stringer * 1457799e4d9SJoe Stringer * **BPF_ANY** 1467799e4d9SJoe Stringer * Create a new element or update an existing element. 1477799e4d9SJoe Stringer * **BPF_NOEXIST** 1487799e4d9SJoe Stringer * Create a new element only if it did not exist. 1497799e4d9SJoe Stringer * **BPF_EXIST** 1507799e4d9SJoe Stringer * Update an existing element. 1516690523bSJoe Stringer * **BPF_F_LOCK** 1526690523bSJoe Stringer * Update a spin_lock-ed map element. 1537799e4d9SJoe Stringer * 1547799e4d9SJoe Stringer * Return 1557799e4d9SJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 1567799e4d9SJoe Stringer * is set appropriately. 1577799e4d9SJoe Stringer * 1587799e4d9SJoe Stringer * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 1597799e4d9SJoe Stringer * **E2BIG**, **EEXIST**, or **ENOENT**. 1607799e4d9SJoe Stringer * 1617799e4d9SJoe Stringer * **E2BIG** 1627799e4d9SJoe Stringer * The number of elements in the map reached the 1637799e4d9SJoe Stringer * *max_entries* limit specified at map creation time. 1647799e4d9SJoe Stringer * **EEXIST** 1657799e4d9SJoe Stringer * If *flags* specifies **BPF_NOEXIST** and the element 1667799e4d9SJoe Stringer * with *key* already exists in the map. 1677799e4d9SJoe Stringer * **ENOENT** 1687799e4d9SJoe Stringer * If *flags* specifies **BPF_EXIST** and the element with 1697799e4d9SJoe Stringer * *key* does not exist in the map. 1707799e4d9SJoe Stringer * 1717799e4d9SJoe Stringer * BPF_MAP_DELETE_ELEM 1727799e4d9SJoe Stringer * Description 1737799e4d9SJoe Stringer * Look up and delete an element by key in a specified map. 1747799e4d9SJoe Stringer * 1757799e4d9SJoe Stringer * Return 1767799e4d9SJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 1777799e4d9SJoe Stringer * is set appropriately. 1787799e4d9SJoe Stringer * 1797799e4d9SJoe Stringer * BPF_MAP_GET_NEXT_KEY 1807799e4d9SJoe Stringer * Description 1817799e4d9SJoe Stringer * Look up an element by key in a specified map and return the key 1827799e4d9SJoe Stringer * of the next element. Can be used to iterate over all elements 1837799e4d9SJoe Stringer * in the map. 1847799e4d9SJoe Stringer * 1857799e4d9SJoe Stringer * Return 1867799e4d9SJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 1877799e4d9SJoe Stringer * is set appropriately. 1887799e4d9SJoe Stringer * 1897799e4d9SJoe Stringer * The following cases can be used to iterate over all elements of 1907799e4d9SJoe Stringer * the map: 1917799e4d9SJoe Stringer * 1927799e4d9SJoe Stringer * * If *key* is not found, the operation returns zero and sets 1937799e4d9SJoe Stringer * the *next_key* pointer to the key of the first element. 1947799e4d9SJoe Stringer * * If *key* is found, the operation returns zero and sets the 1957799e4d9SJoe Stringer * *next_key* pointer to the key of the next element. 1967799e4d9SJoe Stringer * * If *key* is the last element, returns -1 and *errno* is set 1977799e4d9SJoe Stringer * to **ENOENT**. 1987799e4d9SJoe Stringer * 1997799e4d9SJoe Stringer * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 2007799e4d9SJoe Stringer * **EINVAL** on error. 2017799e4d9SJoe Stringer * 2027799e4d9SJoe Stringer * BPF_PROG_LOAD 2037799e4d9SJoe Stringer * Description 2047799e4d9SJoe Stringer * Verify and load an eBPF program, returning a new file 2057799e4d9SJoe Stringer * descriptor associated with the program. 2067799e4d9SJoe Stringer * 2077799e4d9SJoe Stringer * Applying **close**\ (2) to the file descriptor returned by 2087799e4d9SJoe Stringer * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 2097799e4d9SJoe Stringer * 2107799e4d9SJoe Stringer * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 2117799e4d9SJoe Stringer * automatically enabled for the new file descriptor. 2127799e4d9SJoe Stringer * 2137799e4d9SJoe Stringer * Return 2147799e4d9SJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 2157799e4d9SJoe Stringer * error occurred (in which case, *errno* is set appropriately). 2167799e4d9SJoe Stringer * 217f67c9cbfSJoe Stringer * BPF_OBJ_PIN 218f67c9cbfSJoe Stringer * Description 219f67c9cbfSJoe Stringer * Pin an eBPF program or map referred by the specified *bpf_fd* 220f67c9cbfSJoe Stringer * to the provided *pathname* on the filesystem. 221f67c9cbfSJoe Stringer * 2228aacb3c8SJoe Stringer * The *pathname* argument must not contain a dot ("."). 2238aacb3c8SJoe Stringer * 2248aacb3c8SJoe Stringer * On success, *pathname* retains a reference to the eBPF object, 2258aacb3c8SJoe Stringer * preventing deallocation of the object when the original 2268aacb3c8SJoe Stringer * *bpf_fd* is closed. This allow the eBPF object to live beyond 2278aacb3c8SJoe Stringer * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 2288aacb3c8SJoe Stringer * process. 2298aacb3c8SJoe Stringer * 2308aacb3c8SJoe Stringer * Applying **unlink**\ (2) or similar calls to the *pathname* 2318aacb3c8SJoe Stringer * unpins the object from the filesystem, removing the reference. 2328aacb3c8SJoe Stringer * If no other file descriptors or filesystem nodes refer to the 2338aacb3c8SJoe Stringer * same object, it will be deallocated (see NOTES). 2348aacb3c8SJoe Stringer * 2358aacb3c8SJoe Stringer * The filesystem type for the parent directory of *pathname* must 2368aacb3c8SJoe Stringer * be **BPF_FS_MAGIC**. 2378aacb3c8SJoe Stringer * 238f67c9cbfSJoe Stringer * Return 239f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 240f67c9cbfSJoe Stringer * is set appropriately. 241f67c9cbfSJoe Stringer * 242f67c9cbfSJoe Stringer * BPF_OBJ_GET 243f67c9cbfSJoe Stringer * Description 244f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF object pinned to the 245f67c9cbfSJoe Stringer * specified *pathname*. 246f67c9cbfSJoe Stringer * 247f67c9cbfSJoe Stringer * Return 248f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 249f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 250f67c9cbfSJoe Stringer * 251f67c9cbfSJoe Stringer * BPF_PROG_ATTACH 252f67c9cbfSJoe Stringer * Description 253f67c9cbfSJoe Stringer * Attach an eBPF program to a *target_fd* at the specified 254f67c9cbfSJoe Stringer * *attach_type* hook. 255f67c9cbfSJoe Stringer * 25632e76b18SJoe Stringer * The *attach_type* specifies the eBPF attachment point to 25732e76b18SJoe Stringer * attach the program to, and must be one of *bpf_attach_type* 25832e76b18SJoe Stringer * (see below). 25932e76b18SJoe Stringer * 26032e76b18SJoe Stringer * The *attach_bpf_fd* must be a valid file descriptor for a 26132e76b18SJoe Stringer * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 26232e76b18SJoe Stringer * or sock_ops type corresponding to the specified *attach_type*. 26332e76b18SJoe Stringer * 26432e76b18SJoe Stringer * The *target_fd* must be a valid file descriptor for a kernel 26532e76b18SJoe Stringer * object which depends on the attach type of *attach_bpf_fd*: 26632e76b18SJoe Stringer * 26732e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_DEVICE**, 26832e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SKB**, 26932e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK**, 27032e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 27132e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 27232e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 27332e76b18SJoe Stringer * **BPF_PROG_TYPE_SOCK_OPS** 27432e76b18SJoe Stringer * 27532e76b18SJoe Stringer * Control Group v2 hierarchy with the eBPF controller 27632e76b18SJoe Stringer * enabled. Requires the kernel to be compiled with 27732e76b18SJoe Stringer * **CONFIG_CGROUP_BPF**. 27832e76b18SJoe Stringer * 27932e76b18SJoe Stringer * **BPF_PROG_TYPE_FLOW_DISSECTOR** 28032e76b18SJoe Stringer * 28132e76b18SJoe Stringer * Network namespace (eg /proc/self/ns/net). 28232e76b18SJoe Stringer * 28332e76b18SJoe Stringer * **BPF_PROG_TYPE_LIRC_MODE2** 28432e76b18SJoe Stringer * 28532e76b18SJoe Stringer * LIRC device path (eg /dev/lircN). Requires the kernel 28632e76b18SJoe Stringer * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 28732e76b18SJoe Stringer * 28832e76b18SJoe Stringer * **BPF_PROG_TYPE_SK_SKB**, 28932e76b18SJoe Stringer * **BPF_PROG_TYPE_SK_MSG** 29032e76b18SJoe Stringer * 29132e76b18SJoe Stringer * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 29232e76b18SJoe Stringer * 293f67c9cbfSJoe Stringer * Return 294f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 295f67c9cbfSJoe Stringer * is set appropriately. 296f67c9cbfSJoe Stringer * 297f67c9cbfSJoe Stringer * BPF_PROG_DETACH 298f67c9cbfSJoe Stringer * Description 299f67c9cbfSJoe Stringer * Detach the eBPF program associated with the *target_fd* at the 300f67c9cbfSJoe Stringer * hook specified by *attach_type*. The program must have been 301f67c9cbfSJoe Stringer * previously attached using **BPF_PROG_ATTACH**. 302f67c9cbfSJoe Stringer * 303f67c9cbfSJoe Stringer * Return 304f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 305f67c9cbfSJoe Stringer * is set appropriately. 306f67c9cbfSJoe Stringer * 307f67c9cbfSJoe Stringer * BPF_PROG_TEST_RUN 308f67c9cbfSJoe Stringer * Description 3092a3fdca4SJoe Stringer * Run the eBPF program associated with the *prog_fd* a *repeat* 3102a3fdca4SJoe Stringer * number of times against a provided program context *ctx_in* and 3112a3fdca4SJoe Stringer * data *data_in*, and return the modified program context 3122a3fdca4SJoe Stringer * *ctx_out*, *data_out* (for example, packet data), result of the 3132a3fdca4SJoe Stringer * execution *retval*, and *duration* of the test run. 314f67c9cbfSJoe Stringer * 315f3c45326SJoe Stringer * The sizes of the buffers provided as input and output 316f3c45326SJoe Stringer * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 317f3c45326SJoe Stringer * be provided in the corresponding variables *ctx_size_in*, 318f3c45326SJoe Stringer * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 319f3c45326SJoe Stringer * of these parameters are not provided (ie set to NULL), the 320f3c45326SJoe Stringer * corresponding size field must be zero. 321f3c45326SJoe Stringer * 322f3c45326SJoe Stringer * Some program types have particular requirements: 323f3c45326SJoe Stringer * 324f3c45326SJoe Stringer * **BPF_PROG_TYPE_SK_LOOKUP** 325f3c45326SJoe Stringer * *data_in* and *data_out* must be NULL. 326f3c45326SJoe Stringer * 327f3c45326SJoe Stringer * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 328f3c45326SJoe Stringer * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 329f3c45326SJoe Stringer * 330f3c45326SJoe Stringer * *ctx_out*, *data_in* and *data_out* must be NULL. 331f3c45326SJoe Stringer * *repeat* must be zero. 332f3c45326SJoe Stringer * 333e40fbbf0SUsama Arif * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 334e40fbbf0SUsama Arif * 335f67c9cbfSJoe Stringer * Return 336f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 337f67c9cbfSJoe Stringer * is set appropriately. 338f67c9cbfSJoe Stringer * 3392a3fdca4SJoe Stringer * **ENOSPC** 3402a3fdca4SJoe Stringer * Either *data_size_out* or *ctx_size_out* is too small. 3412a3fdca4SJoe Stringer * **ENOTSUPP** 3422a3fdca4SJoe Stringer * This command is not supported by the program type of 3432a3fdca4SJoe Stringer * the program referred to by *prog_fd*. 3442a3fdca4SJoe Stringer * 345f67c9cbfSJoe Stringer * BPF_PROG_GET_NEXT_ID 346f67c9cbfSJoe Stringer * Description 347f67c9cbfSJoe Stringer * Fetch the next eBPF program currently loaded into the kernel. 348f67c9cbfSJoe Stringer * 349f67c9cbfSJoe Stringer * Looks for the eBPF program with an id greater than *start_id* 350f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF programs 351f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 352f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 353f67c9cbfSJoe Stringer * 354f67c9cbfSJoe Stringer * Return 355f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 356f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 357f67c9cbfSJoe Stringer * 358f67c9cbfSJoe Stringer * BPF_MAP_GET_NEXT_ID 359f67c9cbfSJoe Stringer * Description 360f67c9cbfSJoe Stringer * Fetch the next eBPF map currently loaded into the kernel. 361f67c9cbfSJoe Stringer * 362f67c9cbfSJoe Stringer * Looks for the eBPF map with an id greater than *start_id* 363f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF maps 364f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 365f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 366f67c9cbfSJoe Stringer * 367f67c9cbfSJoe Stringer * Return 368f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 369f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 370f67c9cbfSJoe Stringer * 371f67c9cbfSJoe Stringer * BPF_PROG_GET_FD_BY_ID 372f67c9cbfSJoe Stringer * Description 373f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF program corresponding to 374f67c9cbfSJoe Stringer * *prog_id*. 375f67c9cbfSJoe Stringer * 376f67c9cbfSJoe Stringer * Return 377f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 378f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 379f67c9cbfSJoe Stringer * 380f67c9cbfSJoe Stringer * BPF_MAP_GET_FD_BY_ID 381f67c9cbfSJoe Stringer * Description 382f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF map corresponding to 383f67c9cbfSJoe Stringer * *map_id*. 384f67c9cbfSJoe Stringer * 385f67c9cbfSJoe Stringer * Return 386f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 387f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 388f67c9cbfSJoe Stringer * 389f67c9cbfSJoe Stringer * BPF_OBJ_GET_INFO_BY_FD 390f67c9cbfSJoe Stringer * Description 391f67c9cbfSJoe Stringer * Obtain information about the eBPF object corresponding to 392f67c9cbfSJoe Stringer * *bpf_fd*. 393f67c9cbfSJoe Stringer * 394f67c9cbfSJoe Stringer * Populates up to *info_len* bytes of *info*, which will be in 395f67c9cbfSJoe Stringer * one of the following formats depending on the eBPF object type 396f67c9cbfSJoe Stringer * of *bpf_fd*: 397f67c9cbfSJoe Stringer * 398f67c9cbfSJoe Stringer * * **struct bpf_prog_info** 399f67c9cbfSJoe Stringer * * **struct bpf_map_info** 400f67c9cbfSJoe Stringer * * **struct bpf_btf_info** 401f67c9cbfSJoe Stringer * * **struct bpf_link_info** 402f67c9cbfSJoe Stringer * 403f67c9cbfSJoe Stringer * Return 404f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 405f67c9cbfSJoe Stringer * is set appropriately. 406f67c9cbfSJoe Stringer * 407f67c9cbfSJoe Stringer * BPF_PROG_QUERY 408f67c9cbfSJoe Stringer * Description 409f67c9cbfSJoe Stringer * Obtain information about eBPF programs associated with the 410f67c9cbfSJoe Stringer * specified *attach_type* hook. 411f67c9cbfSJoe Stringer * 4125d999994SJoe Stringer * The *target_fd* must be a valid file descriptor for a kernel 4135d999994SJoe Stringer * object which depends on the attach type of *attach_bpf_fd*: 4145d999994SJoe Stringer * 4155d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_DEVICE**, 4165d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SKB**, 4175d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK**, 4185d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 4195d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 4205d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 4215d999994SJoe Stringer * **BPF_PROG_TYPE_SOCK_OPS** 4225d999994SJoe Stringer * 4235d999994SJoe Stringer * Control Group v2 hierarchy with the eBPF controller 4245d999994SJoe Stringer * enabled. Requires the kernel to be compiled with 4255d999994SJoe Stringer * **CONFIG_CGROUP_BPF**. 4265d999994SJoe Stringer * 4275d999994SJoe Stringer * **BPF_PROG_TYPE_FLOW_DISSECTOR** 4285d999994SJoe Stringer * 4295d999994SJoe Stringer * Network namespace (eg /proc/self/ns/net). 4305d999994SJoe Stringer * 4315d999994SJoe Stringer * **BPF_PROG_TYPE_LIRC_MODE2** 4325d999994SJoe Stringer * 4335d999994SJoe Stringer * LIRC device path (eg /dev/lircN). Requires the kernel 4345d999994SJoe Stringer * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 4355d999994SJoe Stringer * 4365d999994SJoe Stringer * **BPF_PROG_QUERY** always fetches the number of programs 4375d999994SJoe Stringer * attached and the *attach_flags* which were used to attach those 4385d999994SJoe Stringer * programs. Additionally, if *prog_ids* is nonzero and the number 4395d999994SJoe Stringer * of attached programs is less than *prog_cnt*, populates 4405d999994SJoe Stringer * *prog_ids* with the eBPF program ids of the programs attached 4415d999994SJoe Stringer * at *target_fd*. 4425d999994SJoe Stringer * 4435d999994SJoe Stringer * The following flags may alter the result: 4445d999994SJoe Stringer * 4455d999994SJoe Stringer * **BPF_F_QUERY_EFFECTIVE** 4465d999994SJoe Stringer * Only return information regarding programs which are 4475d999994SJoe Stringer * currently effective at the specified *target_fd*. 4485d999994SJoe Stringer * 449f67c9cbfSJoe Stringer * Return 450f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 451f67c9cbfSJoe Stringer * is set appropriately. 452f67c9cbfSJoe Stringer * 453f67c9cbfSJoe Stringer * BPF_RAW_TRACEPOINT_OPEN 454f67c9cbfSJoe Stringer * Description 455f67c9cbfSJoe Stringer * Attach an eBPF program to a tracepoint *name* to access kernel 456f67c9cbfSJoe Stringer * internal arguments of the tracepoint in their raw form. 457f67c9cbfSJoe Stringer * 458f67c9cbfSJoe Stringer * The *prog_fd* must be a valid file descriptor associated with 459f67c9cbfSJoe Stringer * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 460f67c9cbfSJoe Stringer * 461f67c9cbfSJoe Stringer * No ABI guarantees are made about the content of tracepoint 462f67c9cbfSJoe Stringer * arguments exposed to the corresponding eBPF program. 463f67c9cbfSJoe Stringer * 464f67c9cbfSJoe Stringer * Applying **close**\ (2) to the file descriptor returned by 465f67c9cbfSJoe Stringer * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 466f67c9cbfSJoe Stringer * 467f67c9cbfSJoe Stringer * Return 468f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 469f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 470f67c9cbfSJoe Stringer * 471f67c9cbfSJoe Stringer * BPF_BTF_LOAD 472f67c9cbfSJoe Stringer * Description 473f67c9cbfSJoe Stringer * Verify and load BPF Type Format (BTF) metadata into the kernel, 474f67c9cbfSJoe Stringer * returning a new file descriptor associated with the metadata. 475f67c9cbfSJoe Stringer * BTF is described in more detail at 476f67c9cbfSJoe Stringer * https://www.kernel.org/doc/html/latest/bpf/btf.html. 477f67c9cbfSJoe Stringer * 478f67c9cbfSJoe Stringer * The *btf* parameter must point to valid memory providing 479f67c9cbfSJoe Stringer * *btf_size* bytes of BTF binary metadata. 480f67c9cbfSJoe Stringer * 481f67c9cbfSJoe Stringer * The returned file descriptor can be passed to other **bpf**\ () 482f67c9cbfSJoe Stringer * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 483f67c9cbfSJoe Stringer * associate the BTF with those objects. 484f67c9cbfSJoe Stringer * 485f67c9cbfSJoe Stringer * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 486f67c9cbfSJoe Stringer * parameters to specify a *btf_log_buf*, *btf_log_size* and 487f67c9cbfSJoe Stringer * *btf_log_level* which allow the kernel to return freeform log 488f67c9cbfSJoe Stringer * output regarding the BTF verification process. 489f67c9cbfSJoe Stringer * 490f67c9cbfSJoe Stringer * Return 491f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 492f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 493f67c9cbfSJoe Stringer * 494f67c9cbfSJoe Stringer * BPF_BTF_GET_FD_BY_ID 495f67c9cbfSJoe Stringer * Description 496f67c9cbfSJoe Stringer * Open a file descriptor for the BPF Type Format (BTF) 497f67c9cbfSJoe Stringer * corresponding to *btf_id*. 498f67c9cbfSJoe Stringer * 499f67c9cbfSJoe Stringer * Return 500f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 501f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 502f67c9cbfSJoe Stringer * 503f67c9cbfSJoe Stringer * BPF_TASK_FD_QUERY 504f67c9cbfSJoe Stringer * Description 505f67c9cbfSJoe Stringer * Obtain information about eBPF programs associated with the 506f67c9cbfSJoe Stringer * target process identified by *pid* and *fd*. 507f67c9cbfSJoe Stringer * 508f67c9cbfSJoe Stringer * If the *pid* and *fd* are associated with a tracepoint, kprobe 509f67c9cbfSJoe Stringer * or uprobe perf event, then the *prog_id* and *fd_type* will 510f67c9cbfSJoe Stringer * be populated with the eBPF program id and file descriptor type 511f67c9cbfSJoe Stringer * of type **bpf_task_fd_type**. If associated with a kprobe or 512f67c9cbfSJoe Stringer * uprobe, the *probe_offset* and *probe_addr* will also be 513f67c9cbfSJoe Stringer * populated. Optionally, if *buf* is provided, then up to 514f67c9cbfSJoe Stringer * *buf_len* bytes of *buf* will be populated with the name of 515f67c9cbfSJoe Stringer * the tracepoint, kprobe or uprobe. 516f67c9cbfSJoe Stringer * 517f67c9cbfSJoe Stringer * The resulting *prog_id* may be introspected in deeper detail 518f67c9cbfSJoe Stringer * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 519f67c9cbfSJoe Stringer * 520f67c9cbfSJoe Stringer * Return 521f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 522f67c9cbfSJoe Stringer * is set appropriately. 523f67c9cbfSJoe Stringer * 524f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_AND_DELETE_ELEM 525f67c9cbfSJoe Stringer * Description 526f67c9cbfSJoe Stringer * Look up an element with the given *key* in the map referred to 527f67c9cbfSJoe Stringer * by the file descriptor *fd*, and if found, delete the element. 528f67c9cbfSJoe Stringer * 5293e87f192SDenis Salopek * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 5303e87f192SDenis Salopek * types, the *flags* argument needs to be set to 0, but for other 5313e87f192SDenis Salopek * map types, it may be specified as: 5323e87f192SDenis Salopek * 5333e87f192SDenis Salopek * **BPF_F_LOCK** 5343e87f192SDenis Salopek * Look up and delete the value of a spin-locked map 5353e87f192SDenis Salopek * without returning the lock. This must be specified if 5363e87f192SDenis Salopek * the elements contain a spinlock. 5373e87f192SDenis Salopek * 538f67c9cbfSJoe Stringer * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 539f67c9cbfSJoe Stringer * implement this command as a "pop" operation, deleting the top 540f67c9cbfSJoe Stringer * element rather than one corresponding to *key*. 541f67c9cbfSJoe Stringer * The *key* and *key_len* parameters should be zeroed when 542f67c9cbfSJoe Stringer * issuing this operation for these map types. 543f67c9cbfSJoe Stringer * 544f67c9cbfSJoe Stringer * This command is only valid for the following map types: 545f67c9cbfSJoe Stringer * * **BPF_MAP_TYPE_QUEUE** 546f67c9cbfSJoe Stringer * * **BPF_MAP_TYPE_STACK** 5473e87f192SDenis Salopek * * **BPF_MAP_TYPE_HASH** 5483e87f192SDenis Salopek * * **BPF_MAP_TYPE_PERCPU_HASH** 5493e87f192SDenis Salopek * * **BPF_MAP_TYPE_LRU_HASH** 5503e87f192SDenis Salopek * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 551f67c9cbfSJoe Stringer * 552f67c9cbfSJoe Stringer * Return 553f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 554f67c9cbfSJoe Stringer * is set appropriately. 555f67c9cbfSJoe Stringer * 556f67c9cbfSJoe Stringer * BPF_MAP_FREEZE 557f67c9cbfSJoe Stringer * Description 558f67c9cbfSJoe Stringer * Freeze the permissions of the specified map. 559f67c9cbfSJoe Stringer * 560f67c9cbfSJoe Stringer * Write permissions may be frozen by passing zero *flags*. 561f67c9cbfSJoe Stringer * Upon success, no future syscall invocations may alter the 562f67c9cbfSJoe Stringer * map state of *map_fd*. Write operations from eBPF programs 563f67c9cbfSJoe Stringer * are still possible for a frozen map. 564f67c9cbfSJoe Stringer * 565f67c9cbfSJoe Stringer * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 566f67c9cbfSJoe Stringer * 567f67c9cbfSJoe Stringer * Return 568f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 569f67c9cbfSJoe Stringer * is set appropriately. 570f67c9cbfSJoe Stringer * 571f67c9cbfSJoe Stringer * BPF_BTF_GET_NEXT_ID 572f67c9cbfSJoe Stringer * Description 573f67c9cbfSJoe Stringer * Fetch the next BPF Type Format (BTF) object currently loaded 574f67c9cbfSJoe Stringer * into the kernel. 575f67c9cbfSJoe Stringer * 576f67c9cbfSJoe Stringer * Looks for the BTF object with an id greater than *start_id* 577f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other BTF objects 578f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 579f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 580f67c9cbfSJoe Stringer * 581f67c9cbfSJoe Stringer * Return 582f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 583f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 584f67c9cbfSJoe Stringer * 585f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_BATCH 586f67c9cbfSJoe Stringer * Description 587f67c9cbfSJoe Stringer * Iterate and fetch multiple elements in a map. 588f67c9cbfSJoe Stringer * 5890cb80454SJoe Stringer * Two opaque values are used to manage batch operations, 5900cb80454SJoe Stringer * *in_batch* and *out_batch*. Initially, *in_batch* must be set 5910cb80454SJoe Stringer * to NULL to begin the batched operation. After each subsequent 5920cb80454SJoe Stringer * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 5930cb80454SJoe Stringer * *out_batch* as the *in_batch* for the next operation to 5940cb80454SJoe Stringer * continue iteration from the current point. 5950cb80454SJoe Stringer * 5960cb80454SJoe Stringer * The *keys* and *values* are output parameters which must point 5970cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 5980cb80454SJoe Stringer * and value size of the map *map_fd*. The *keys* buffer must be 5990cb80454SJoe Stringer * of *key_size* * *count*. The *values* buffer must be of 6000cb80454SJoe Stringer * *value_size* * *count*. 6010cb80454SJoe Stringer * 6020cb80454SJoe Stringer * The *elem_flags* argument may be specified as one of the 6030cb80454SJoe Stringer * following: 6040cb80454SJoe Stringer * 6050cb80454SJoe Stringer * **BPF_F_LOCK** 6060cb80454SJoe Stringer * Look up the value of a spin-locked map without 6070cb80454SJoe Stringer * returning the lock. This must be specified if the 6080cb80454SJoe Stringer * elements contain a spinlock. 6090cb80454SJoe Stringer * 6100cb80454SJoe Stringer * On success, *count* elements from the map are copied into the 6110cb80454SJoe Stringer * user buffer, with the keys copied into *keys* and the values 6120cb80454SJoe Stringer * copied into the corresponding indices in *values*. 6130cb80454SJoe Stringer * 6140cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 6150cb80454SJoe Stringer * is set to the number of successfully processed elements. 6160cb80454SJoe Stringer * 617f67c9cbfSJoe Stringer * Return 618f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 619f67c9cbfSJoe Stringer * is set appropriately. 620f67c9cbfSJoe Stringer * 6210cb80454SJoe Stringer * May set *errno* to **ENOSPC** to indicate that *keys* or 6220cb80454SJoe Stringer * *values* is too small to dump an entire bucket during 6230cb80454SJoe Stringer * iteration of a hash-based map type. 6240cb80454SJoe Stringer * 625f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_AND_DELETE_BATCH 626f67c9cbfSJoe Stringer * Description 6270cb80454SJoe Stringer * Iterate and delete all elements in a map. 6280cb80454SJoe Stringer * 6290cb80454SJoe Stringer * This operation has the same behavior as 6300cb80454SJoe Stringer * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 6310cb80454SJoe Stringer * 6320cb80454SJoe Stringer * * Every element that is successfully returned is also deleted 6330cb80454SJoe Stringer * from the map. This is at least *count* elements. Note that 6340cb80454SJoe Stringer * *count* is both an input and an output parameter. 6350cb80454SJoe Stringer * * Upon returning with *errno* set to **EFAULT**, up to 6360cb80454SJoe Stringer * *count* elements may be deleted without returning the keys 6370cb80454SJoe Stringer * and values of the deleted elements. 638f67c9cbfSJoe Stringer * 639f67c9cbfSJoe Stringer * Return 640f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 641f67c9cbfSJoe Stringer * is set appropriately. 642f67c9cbfSJoe Stringer * 643f67c9cbfSJoe Stringer * BPF_MAP_UPDATE_BATCH 644f67c9cbfSJoe Stringer * Description 6450cb80454SJoe Stringer * Update multiple elements in a map by *key*. 6460cb80454SJoe Stringer * 6470cb80454SJoe Stringer * The *keys* and *values* are input parameters which must point 6480cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 6490cb80454SJoe Stringer * and value size of the map *map_fd*. The *keys* buffer must be 6500cb80454SJoe Stringer * of *key_size* * *count*. The *values* buffer must be of 6510cb80454SJoe Stringer * *value_size* * *count*. 6520cb80454SJoe Stringer * 6530cb80454SJoe Stringer * Each element specified in *keys* is sequentially updated to the 6540cb80454SJoe Stringer * value in the corresponding index in *values*. The *in_batch* 6550cb80454SJoe Stringer * and *out_batch* parameters are ignored and should be zeroed. 6560cb80454SJoe Stringer * 6570cb80454SJoe Stringer * The *elem_flags* argument should be specified as one of the 6580cb80454SJoe Stringer * following: 6590cb80454SJoe Stringer * 6600cb80454SJoe Stringer * **BPF_ANY** 6610cb80454SJoe Stringer * Create new elements or update a existing elements. 6620cb80454SJoe Stringer * **BPF_NOEXIST** 6630cb80454SJoe Stringer * Create new elements only if they do not exist. 6640cb80454SJoe Stringer * **BPF_EXIST** 6650cb80454SJoe Stringer * Update existing elements. 6660cb80454SJoe Stringer * **BPF_F_LOCK** 6670cb80454SJoe Stringer * Update spin_lock-ed map elements. This must be 6680cb80454SJoe Stringer * specified if the map value contains a spinlock. 6690cb80454SJoe Stringer * 6700cb80454SJoe Stringer * On success, *count* elements from the map are updated. 6710cb80454SJoe Stringer * 6720cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 6730cb80454SJoe Stringer * is set to the number of successfully processed elements. 674f67c9cbfSJoe Stringer * 675f67c9cbfSJoe Stringer * Return 676f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 677f67c9cbfSJoe Stringer * is set appropriately. 678f67c9cbfSJoe Stringer * 6790cb80454SJoe Stringer * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 6800cb80454SJoe Stringer * **E2BIG**. **E2BIG** indicates that the number of elements in 6810cb80454SJoe Stringer * the map reached the *max_entries* limit specified at map 6820cb80454SJoe Stringer * creation time. 6830cb80454SJoe Stringer * 6840cb80454SJoe Stringer * May set *errno* to one of the following error codes under 6850cb80454SJoe Stringer * specific circumstances: 6860cb80454SJoe Stringer * 6870cb80454SJoe Stringer * **EEXIST** 6880cb80454SJoe Stringer * If *flags* specifies **BPF_NOEXIST** and the element 6890cb80454SJoe Stringer * with *key* already exists in the map. 6900cb80454SJoe Stringer * **ENOENT** 6910cb80454SJoe Stringer * If *flags* specifies **BPF_EXIST** and the element with 6920cb80454SJoe Stringer * *key* does not exist in the map. 6930cb80454SJoe Stringer * 694f67c9cbfSJoe Stringer * BPF_MAP_DELETE_BATCH 695f67c9cbfSJoe Stringer * Description 6960cb80454SJoe Stringer * Delete multiple elements in a map by *key*. 6970cb80454SJoe Stringer * 6980cb80454SJoe Stringer * The *keys* parameter is an input parameter which must point 6990cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 7000cb80454SJoe Stringer * size of the map *map_fd*, that is, *key_size* * *count*. 7010cb80454SJoe Stringer * 7020cb80454SJoe Stringer * Each element specified in *keys* is sequentially deleted. The 7030cb80454SJoe Stringer * *in_batch*, *out_batch*, and *values* parameters are ignored 7040cb80454SJoe Stringer * and should be zeroed. 7050cb80454SJoe Stringer * 7060cb80454SJoe Stringer * The *elem_flags* argument may be specified as one of the 7070cb80454SJoe Stringer * following: 7080cb80454SJoe Stringer * 7090cb80454SJoe Stringer * **BPF_F_LOCK** 7100cb80454SJoe Stringer * Look up the value of a spin-locked map without 7110cb80454SJoe Stringer * returning the lock. This must be specified if the 7120cb80454SJoe Stringer * elements contain a spinlock. 7130cb80454SJoe Stringer * 7140cb80454SJoe Stringer * On success, *count* elements from the map are updated. 7150cb80454SJoe Stringer * 7160cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 7170cb80454SJoe Stringer * is set to the number of successfully processed elements. If 7180cb80454SJoe Stringer * *errno* is **EFAULT**, up to *count* elements may be been 7190cb80454SJoe Stringer * deleted. 720f67c9cbfSJoe Stringer * 721f67c9cbfSJoe Stringer * Return 722f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 723f67c9cbfSJoe Stringer * is set appropriately. 724f67c9cbfSJoe Stringer * 725f67c9cbfSJoe Stringer * BPF_LINK_CREATE 726f67c9cbfSJoe Stringer * Description 727f67c9cbfSJoe Stringer * Attach an eBPF program to a *target_fd* at the specified 728f67c9cbfSJoe Stringer * *attach_type* hook and return a file descriptor handle for 729f67c9cbfSJoe Stringer * managing the link. 730f67c9cbfSJoe Stringer * 731f67c9cbfSJoe Stringer * Return 732f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 733f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 734f67c9cbfSJoe Stringer * 735f67c9cbfSJoe Stringer * BPF_LINK_UPDATE 736f67c9cbfSJoe Stringer * Description 737f67c9cbfSJoe Stringer * Update the eBPF program in the specified *link_fd* to 738f67c9cbfSJoe Stringer * *new_prog_fd*. 739f67c9cbfSJoe Stringer * 740f67c9cbfSJoe Stringer * Return 741f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 742f67c9cbfSJoe Stringer * is set appropriately. 743f67c9cbfSJoe Stringer * 744f67c9cbfSJoe Stringer * BPF_LINK_GET_FD_BY_ID 745f67c9cbfSJoe Stringer * Description 746f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF Link corresponding to 747f67c9cbfSJoe Stringer * *link_id*. 748f67c9cbfSJoe Stringer * 749f67c9cbfSJoe Stringer * Return 750f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 751f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 752f67c9cbfSJoe Stringer * 753f67c9cbfSJoe Stringer * BPF_LINK_GET_NEXT_ID 754f67c9cbfSJoe Stringer * Description 755f67c9cbfSJoe Stringer * Fetch the next eBPF link currently loaded into the kernel. 756f67c9cbfSJoe Stringer * 757f67c9cbfSJoe Stringer * Looks for the eBPF link with an id greater than *start_id* 758f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF links 759f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 760f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 761f67c9cbfSJoe Stringer * 762f67c9cbfSJoe Stringer * Return 763f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 764f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 765f67c9cbfSJoe Stringer * 766f67c9cbfSJoe Stringer * BPF_ENABLE_STATS 767f67c9cbfSJoe Stringer * Description 768f67c9cbfSJoe Stringer * Enable eBPF runtime statistics gathering. 769f67c9cbfSJoe Stringer * 770f67c9cbfSJoe Stringer * Runtime statistics gathering for the eBPF runtime is disabled 771f67c9cbfSJoe Stringer * by default to minimize the corresponding performance overhead. 772f67c9cbfSJoe Stringer * This command enables statistics globally. 773f67c9cbfSJoe Stringer * 774f67c9cbfSJoe Stringer * Multiple programs may independently enable statistics. 775f67c9cbfSJoe Stringer * After gathering the desired statistics, eBPF runtime statistics 776f67c9cbfSJoe Stringer * may be disabled again by calling **close**\ (2) for the file 777f67c9cbfSJoe Stringer * descriptor returned by this function. Statistics will only be 778f67c9cbfSJoe Stringer * disabled system-wide when all outstanding file descriptors 779f67c9cbfSJoe Stringer * returned by prior calls for this subcommand are closed. 780f67c9cbfSJoe Stringer * 781f67c9cbfSJoe Stringer * Return 782f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 783f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 784f67c9cbfSJoe Stringer * 785f67c9cbfSJoe Stringer * BPF_ITER_CREATE 786f67c9cbfSJoe Stringer * Description 787f67c9cbfSJoe Stringer * Create an iterator on top of the specified *link_fd* (as 788f67c9cbfSJoe Stringer * previously created using **BPF_LINK_CREATE**) and return a 789f67c9cbfSJoe Stringer * file descriptor that can be used to trigger the iteration. 790f67c9cbfSJoe Stringer * 791f67c9cbfSJoe Stringer * If the resulting file descriptor is pinned to the filesystem 792f67c9cbfSJoe Stringer * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 793f67c9cbfSJoe Stringer * for that path will trigger the iterator to read kernel state 794f67c9cbfSJoe Stringer * using the eBPF program attached to *link_fd*. 795f67c9cbfSJoe Stringer * 796f67c9cbfSJoe Stringer * Return 797f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 798f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 799f67c9cbfSJoe Stringer * 800f67c9cbfSJoe Stringer * BPF_LINK_DETACH 801f67c9cbfSJoe Stringer * Description 802f67c9cbfSJoe Stringer * Forcefully detach the specified *link_fd* from its 803f67c9cbfSJoe Stringer * corresponding attachment point. 804f67c9cbfSJoe Stringer * 805f67c9cbfSJoe Stringer * Return 806f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 807f67c9cbfSJoe Stringer * is set appropriately. 808f67c9cbfSJoe Stringer * 809f67c9cbfSJoe Stringer * BPF_PROG_BIND_MAP 810f67c9cbfSJoe Stringer * Description 811f67c9cbfSJoe Stringer * Bind a map to the lifetime of an eBPF program. 812f67c9cbfSJoe Stringer * 813f67c9cbfSJoe Stringer * The map identified by *map_fd* is bound to the program 814f67c9cbfSJoe Stringer * identified by *prog_fd* and only released when *prog_fd* is 815f67c9cbfSJoe Stringer * released. This may be used in cases where metadata should be 816f67c9cbfSJoe Stringer * associated with a program which otherwise does not contain any 817f67c9cbfSJoe Stringer * references to the map (for example, embedded in the eBPF 818f67c9cbfSJoe Stringer * program instructions). 819f67c9cbfSJoe Stringer * 820f67c9cbfSJoe Stringer * Return 821f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 822f67c9cbfSJoe Stringer * is set appropriately. 823f67c9cbfSJoe Stringer * 8247799e4d9SJoe Stringer * NOTES 8257799e4d9SJoe Stringer * eBPF objects (maps and programs) can be shared between processes. 8268aacb3c8SJoe Stringer * 8278aacb3c8SJoe Stringer * * After **fork**\ (2), the child inherits file descriptors 8288aacb3c8SJoe Stringer * referring to the same eBPF objects. 8298aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be transferred over 8308aacb3c8SJoe Stringer * **unix**\ (7) domain sockets. 8318aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be duplicated in the 8328aacb3c8SJoe Stringer * usual way, using **dup**\ (2) and similar calls. 8338aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be pinned to the 8348aacb3c8SJoe Stringer * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 8358aacb3c8SJoe Stringer * 8368aacb3c8SJoe Stringer * An eBPF object is deallocated only after all file descriptors referring 8378aacb3c8SJoe Stringer * to the object have been closed and no references remain pinned to the 8388aacb3c8SJoe Stringer * filesystem or attached (for example, bound to a program or device). 8397799e4d9SJoe Stringer */ 84099c55f7dSAlexei Starovoitov enum bpf_cmd { 84199c55f7dSAlexei Starovoitov BPF_MAP_CREATE, 842db20fd2bSAlexei Starovoitov BPF_MAP_LOOKUP_ELEM, 843db20fd2bSAlexei Starovoitov BPF_MAP_UPDATE_ELEM, 844db20fd2bSAlexei Starovoitov BPF_MAP_DELETE_ELEM, 845db20fd2bSAlexei Starovoitov BPF_MAP_GET_NEXT_KEY, 84609756af4SAlexei Starovoitov BPF_PROG_LOAD, 847b2197755SDaniel Borkmann BPF_OBJ_PIN, 848b2197755SDaniel Borkmann BPF_OBJ_GET, 849f4324551SDaniel Mack BPF_PROG_ATTACH, 850f4324551SDaniel Mack BPF_PROG_DETACH, 8511cf1cae9SAlexei Starovoitov BPF_PROG_TEST_RUN, 8525d67f349SAlexei Starovoitov BPF_PROG_RUN = BPF_PROG_TEST_RUN, 85334ad5580SMartin KaFai Lau BPF_PROG_GET_NEXT_ID, 85434ad5580SMartin KaFai Lau BPF_MAP_GET_NEXT_ID, 855b16d9aa4SMartin KaFai Lau BPF_PROG_GET_FD_BY_ID, 856bd5f5f4eSMartin KaFai Lau BPF_MAP_GET_FD_BY_ID, 8571e270976SMartin KaFai Lau BPF_OBJ_GET_INFO_BY_FD, 858468e2f64SAlexei Starovoitov BPF_PROG_QUERY, 859c4f6699dSAlexei Starovoitov BPF_RAW_TRACEPOINT_OPEN, 860f56a653cSMartin KaFai Lau BPF_BTF_LOAD, 86178958fcaSMartin KaFai Lau BPF_BTF_GET_FD_BY_ID, 86241bdc4b4SYonghong Song BPF_TASK_FD_QUERY, 863bd513cd0SMauricio Vasquez B BPF_MAP_LOOKUP_AND_DELETE_ELEM, 86487df15deSDaniel Borkmann BPF_MAP_FREEZE, 8651b9ed84eSQuentin Monnet BPF_BTF_GET_NEXT_ID, 866cb4d03abSBrian Vazquez BPF_MAP_LOOKUP_BATCH, 86705799638SYonghong Song BPF_MAP_LOOKUP_AND_DELETE_BATCH, 868aa2e93b8SBrian Vazquez BPF_MAP_UPDATE_BATCH, 869aa2e93b8SBrian Vazquez BPF_MAP_DELETE_BATCH, 870af6eea57SAndrii Nakryiko BPF_LINK_CREATE, 8710c991ebcSAndrii Nakryiko BPF_LINK_UPDATE, 8722d602c8cSAndrii Nakryiko BPF_LINK_GET_FD_BY_ID, 8732d602c8cSAndrii Nakryiko BPF_LINK_GET_NEXT_ID, 874d46edd67SSong Liu BPF_ENABLE_STATS, 875ac51d99bSYonghong Song BPF_ITER_CREATE, 87673b11c2aSAndrii Nakryiko BPF_LINK_DETACH, 877ef15314aSYiFei Zhu BPF_PROG_BIND_MAP, 87899c55f7dSAlexei Starovoitov }; 87999c55f7dSAlexei Starovoitov 88099c55f7dSAlexei Starovoitov enum bpf_map_type { 88199c55f7dSAlexei Starovoitov BPF_MAP_TYPE_UNSPEC, 8820f8e4bd8SAlexei Starovoitov BPF_MAP_TYPE_HASH, 88328fbcfa0SAlexei Starovoitov BPF_MAP_TYPE_ARRAY, 88404fd61abSAlexei Starovoitov BPF_MAP_TYPE_PROG_ARRAY, 885ea317b26SKaixu Xia BPF_MAP_TYPE_PERF_EVENT_ARRAY, 886824bd0ceSAlexei Starovoitov BPF_MAP_TYPE_PERCPU_HASH, 887a10423b8SAlexei Starovoitov BPF_MAP_TYPE_PERCPU_ARRAY, 888d5a3b1f6SAlexei Starovoitov BPF_MAP_TYPE_STACK_TRACE, 8894ed8ec52SMartin KaFai Lau BPF_MAP_TYPE_CGROUP_ARRAY, 89029ba732aSMartin KaFai Lau BPF_MAP_TYPE_LRU_HASH, 8918f844938SMartin KaFai Lau BPF_MAP_TYPE_LRU_PERCPU_HASH, 892b95a5c4dSDaniel Mack BPF_MAP_TYPE_LPM_TRIE, 89356f668dfSMartin KaFai Lau BPF_MAP_TYPE_ARRAY_OF_MAPS, 894bcc6b1b7SMartin KaFai Lau BPF_MAP_TYPE_HASH_OF_MAPS, 895546ac1ffSJohn Fastabend BPF_MAP_TYPE_DEVMAP, 896174a79ffSJohn Fastabend BPF_MAP_TYPE_SOCKMAP, 8976710e112SJesper Dangaard Brouer BPF_MAP_TYPE_CPUMAP, 898fbfc504aSBjörn Töpel BPF_MAP_TYPE_XSKMAP, 89981110384SJohn Fastabend BPF_MAP_TYPE_SOCKHASH, 900de9cbbaaSRoman Gushchin BPF_MAP_TYPE_CGROUP_STORAGE, 9015dc4c4b7SMartin KaFai Lau BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 902b741f163SRoman Gushchin BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 903f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_QUEUE, 904f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_STACK, 9056ac99e8fSMartin KaFai Lau BPF_MAP_TYPE_SK_STORAGE, 9066f9d451aSToke Høiland-Jørgensen BPF_MAP_TYPE_DEVMAP_HASH, 90785d33df3SMartin KaFai Lau BPF_MAP_TYPE_STRUCT_OPS, 908457f4436SAndrii Nakryiko BPF_MAP_TYPE_RINGBUF, 9098ea63684SKP Singh BPF_MAP_TYPE_INODE_STORAGE, 9104cf1bc1fSKP Singh BPF_MAP_TYPE_TASK_STORAGE, 9119330986cSJoanne Koong BPF_MAP_TYPE_BLOOM_FILTER, 91299c55f7dSAlexei Starovoitov }; 91399c55f7dSAlexei Starovoitov 9146c4fc209SDaniel Borkmann /* Note that tracing related programs such as 9156c4fc209SDaniel Borkmann * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 9166c4fc209SDaniel Borkmann * are not subject to a stable API since kernel internal data 9176c4fc209SDaniel Borkmann * structures can change from release to release and may 9186c4fc209SDaniel Borkmann * therefore break existing tracing BPF programs. Tracing BPF 9196c4fc209SDaniel Borkmann * programs correspond to /a/ specific kernel which is to be 9206c4fc209SDaniel Borkmann * analyzed, and not /a/ specific kernel /and/ all future ones. 9216c4fc209SDaniel Borkmann */ 92209756af4SAlexei Starovoitov enum bpf_prog_type { 92309756af4SAlexei Starovoitov BPF_PROG_TYPE_UNSPEC, 924ddd872bcSAlexei Starovoitov BPF_PROG_TYPE_SOCKET_FILTER, 9252541517cSAlexei Starovoitov BPF_PROG_TYPE_KPROBE, 92696be4325SDaniel Borkmann BPF_PROG_TYPE_SCHED_CLS, 92794caee8cSDaniel Borkmann BPF_PROG_TYPE_SCHED_ACT, 92898b5c2c6SAlexei Starovoitov BPF_PROG_TYPE_TRACEPOINT, 9296a773a15SBrenden Blanco BPF_PROG_TYPE_XDP, 9300515e599SAlexei Starovoitov BPF_PROG_TYPE_PERF_EVENT, 9310e33661dSDaniel Mack BPF_PROG_TYPE_CGROUP_SKB, 93261023658SDavid Ahern BPF_PROG_TYPE_CGROUP_SOCK, 9333a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_IN, 9343a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_OUT, 9353a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_XMIT, 93640304b2aSLawrence Brakmo BPF_PROG_TYPE_SOCK_OPS, 937b005fd18SJohn Fastabend BPF_PROG_TYPE_SK_SKB, 938ebc614f6SRoman Gushchin BPF_PROG_TYPE_CGROUP_DEVICE, 9394f738adbSJohn Fastabend BPF_PROG_TYPE_SK_MSG, 940c4f6699dSAlexei Starovoitov BPF_PROG_TYPE_RAW_TRACEPOINT, 9414fbac77dSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 942004d4b27SMathieu Xhonneux BPF_PROG_TYPE_LWT_SEG6LOCAL, 943f4364dcfSSean Young BPF_PROG_TYPE_LIRC_MODE2, 9442dbb9b9eSMartin KaFai Lau BPF_PROG_TYPE_SK_REUSEPORT, 945d58e468bSPetar Penkov BPF_PROG_TYPE_FLOW_DISSECTOR, 9467b146cebSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SYSCTL, 9479df1c28bSMatt Mullins BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 9480d01da6aSStanislav Fomichev BPF_PROG_TYPE_CGROUP_SOCKOPT, 949f1b9509cSAlexei Starovoitov BPF_PROG_TYPE_TRACING, 95027ae7997SMartin KaFai Lau BPF_PROG_TYPE_STRUCT_OPS, 951be8704ffSAlexei Starovoitov BPF_PROG_TYPE_EXT, 952fc611f47SKP Singh BPF_PROG_TYPE_LSM, 953e9ddbb77SJakub Sitnicki BPF_PROG_TYPE_SK_LOOKUP, 95479a7f8bdSAlexei Starovoitov BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 95509756af4SAlexei Starovoitov }; 95609756af4SAlexei Starovoitov 9570e33661dSDaniel Mack enum bpf_attach_type { 9580e33661dSDaniel Mack BPF_CGROUP_INET_INGRESS, 9590e33661dSDaniel Mack BPF_CGROUP_INET_EGRESS, 96061023658SDavid Ahern BPF_CGROUP_INET_SOCK_CREATE, 96140304b2aSLawrence Brakmo BPF_CGROUP_SOCK_OPS, 962464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_PARSER, 963464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_VERDICT, 964ebc614f6SRoman Gushchin BPF_CGROUP_DEVICE, 9654f738adbSJohn Fastabend BPF_SK_MSG_VERDICT, 9664fbac77dSAndrey Ignatov BPF_CGROUP_INET4_BIND, 9674fbac77dSAndrey Ignatov BPF_CGROUP_INET6_BIND, 968d74bad4eSAndrey Ignatov BPF_CGROUP_INET4_CONNECT, 969d74bad4eSAndrey Ignatov BPF_CGROUP_INET6_CONNECT, 970aac3fc32SAndrey Ignatov BPF_CGROUP_INET4_POST_BIND, 971aac3fc32SAndrey Ignatov BPF_CGROUP_INET6_POST_BIND, 9721cedee13SAndrey Ignatov BPF_CGROUP_UDP4_SENDMSG, 9731cedee13SAndrey Ignatov BPF_CGROUP_UDP6_SENDMSG, 974f4364dcfSSean Young BPF_LIRC_MODE2, 975d58e468bSPetar Penkov BPF_FLOW_DISSECTOR, 9767b146cebSAndrey Ignatov BPF_CGROUP_SYSCTL, 977983695faSDaniel Borkmann BPF_CGROUP_UDP4_RECVMSG, 978983695faSDaniel Borkmann BPF_CGROUP_UDP6_RECVMSG, 9790d01da6aSStanislav Fomichev BPF_CGROUP_GETSOCKOPT, 9800d01da6aSStanislav Fomichev BPF_CGROUP_SETSOCKOPT, 981f1b9509cSAlexei Starovoitov BPF_TRACE_RAW_TP, 982fec56f58SAlexei Starovoitov BPF_TRACE_FENTRY, 983fec56f58SAlexei Starovoitov BPF_TRACE_FEXIT, 984ae240823SKP Singh BPF_MODIFY_RETURN, 985fc611f47SKP Singh BPF_LSM_MAC, 98615d83c4dSYonghong Song BPF_TRACE_ITER, 9871b66d253SDaniel Borkmann BPF_CGROUP_INET4_GETPEERNAME, 9881b66d253SDaniel Borkmann BPF_CGROUP_INET6_GETPEERNAME, 9891b66d253SDaniel Borkmann BPF_CGROUP_INET4_GETSOCKNAME, 9901b66d253SDaniel Borkmann BPF_CGROUP_INET6_GETSOCKNAME, 991fbee97feSDavid Ahern BPF_XDP_DEVMAP, 992f5836749SStanislav Fomichev BPF_CGROUP_INET_SOCK_RELEASE, 99392164774SLorenzo Bianconi BPF_XDP_CPUMAP, 994e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP, 995aa8d3a71SAndrii Nakryiko BPF_XDP, 996a7ba4558SCong Wang BPF_SK_SKB_VERDICT, 997d5e4ddaeSKuniyuki Iwashima BPF_SK_REUSEPORT_SELECT, 998d5e4ddaeSKuniyuki Iwashima BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 999b89fbfbbSAndrii Nakryiko BPF_PERF_EVENT, 10000dcac272SJiri Olsa BPF_TRACE_KPROBE_MULTI, 100169fd337aSStanislav Fomichev BPF_LSM_CGROUP, 10020e33661dSDaniel Mack __MAX_BPF_ATTACH_TYPE 10030e33661dSDaniel Mack }; 10040e33661dSDaniel Mack 10050e33661dSDaniel Mack #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 10060e33661dSDaniel Mack 1007f2e10bffSAndrii Nakryiko enum bpf_link_type { 1008f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_UNSPEC = 0, 1009f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1010f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_TRACING = 2, 1011f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_CGROUP = 3, 1012de4e05caSYonghong Song BPF_LINK_TYPE_ITER = 4, 10137f045a49SJakub Sitnicki BPF_LINK_TYPE_NETNS = 5, 1014aa8d3a71SAndrii Nakryiko BPF_LINK_TYPE_XDP = 6, 1015b89fbfbbSAndrii Nakryiko BPF_LINK_TYPE_PERF_EVENT = 7, 10160dcac272SJiri Olsa BPF_LINK_TYPE_KPROBE_MULTI = 8, 1017f7e0beafSKui-Feng Lee BPF_LINK_TYPE_STRUCT_OPS = 9, 1018f2e10bffSAndrii Nakryiko 1019f2e10bffSAndrii Nakryiko MAX_BPF_LINK_TYPE, 1020f2e10bffSAndrii Nakryiko }; 1021f2e10bffSAndrii Nakryiko 1022324bda9eSAlexei Starovoitov /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1023324bda9eSAlexei Starovoitov * 1024324bda9eSAlexei Starovoitov * NONE(default): No further bpf programs allowed in the subtree. 1025324bda9eSAlexei Starovoitov * 1026324bda9eSAlexei Starovoitov * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1027324bda9eSAlexei Starovoitov * the program in this cgroup yields to sub-cgroup program. 1028324bda9eSAlexei Starovoitov * 1029324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1030324bda9eSAlexei Starovoitov * that cgroup program gets run in addition to the program in this cgroup. 1031324bda9eSAlexei Starovoitov * 1032324bda9eSAlexei Starovoitov * Only one program is allowed to be attached to a cgroup with 1033324bda9eSAlexei Starovoitov * NONE or BPF_F_ALLOW_OVERRIDE flag. 1034324bda9eSAlexei Starovoitov * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1035324bda9eSAlexei Starovoitov * release old program and attach the new one. Attach flags has to match. 1036324bda9eSAlexei Starovoitov * 1037324bda9eSAlexei Starovoitov * Multiple programs are allowed to be attached to a cgroup with 1038324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1039324bda9eSAlexei Starovoitov * (those that were attached first, run first) 1040324bda9eSAlexei Starovoitov * The programs of sub-cgroup are executed first, then programs of 1041324bda9eSAlexei Starovoitov * this cgroup and then programs of parent cgroup. 1042324bda9eSAlexei Starovoitov * When children program makes decision (like picking TCP CA or sock bind) 1043324bda9eSAlexei Starovoitov * parent program has a chance to override it. 1044324bda9eSAlexei Starovoitov * 10457dd68b32SAndrey Ignatov * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 10467dd68b32SAndrey Ignatov * programs for a cgroup. Though it's possible to replace an old program at 10477dd68b32SAndrey Ignatov * any position by also specifying BPF_F_REPLACE flag and position itself in 10487dd68b32SAndrey Ignatov * replace_bpf_fd attribute. Old program at this position will be released. 10497dd68b32SAndrey Ignatov * 1050324bda9eSAlexei Starovoitov * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1051324bda9eSAlexei Starovoitov * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1052324bda9eSAlexei Starovoitov * Ex1: 1053324bda9eSAlexei Starovoitov * cgrp1 (MULTI progs A, B) -> 1054324bda9eSAlexei Starovoitov * cgrp2 (OVERRIDE prog C) -> 1055324bda9eSAlexei Starovoitov * cgrp3 (MULTI prog D) -> 1056324bda9eSAlexei Starovoitov * cgrp4 (OVERRIDE prog E) -> 1057324bda9eSAlexei Starovoitov * cgrp5 (NONE prog F) 1058324bda9eSAlexei Starovoitov * the event in cgrp5 triggers execution of F,D,A,B in that order. 1059324bda9eSAlexei Starovoitov * if prog F is detached, the execution is E,D,A,B 1060324bda9eSAlexei Starovoitov * if prog F and D are detached, the execution is E,A,B 1061324bda9eSAlexei Starovoitov * if prog F, E and D are detached, the execution is C,A,B 1062324bda9eSAlexei Starovoitov * 1063324bda9eSAlexei Starovoitov * All eligible programs are executed regardless of return code from 1064324bda9eSAlexei Starovoitov * earlier programs. 10657f677633SAlexei Starovoitov */ 10667f677633SAlexei Starovoitov #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1067324bda9eSAlexei Starovoitov #define BPF_F_ALLOW_MULTI (1U << 1) 10687dd68b32SAndrey Ignatov #define BPF_F_REPLACE (1U << 2) 10697f677633SAlexei Starovoitov 1070e07b98d9SDavid S. Miller /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1071e07b98d9SDavid S. Miller * verifier will perform strict alignment checking as if the kernel 1072e07b98d9SDavid S. Miller * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1073e07b98d9SDavid S. Miller * and NET_IP_ALIGN defined to 2. 1074e07b98d9SDavid S. Miller */ 1075e07b98d9SDavid S. Miller #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1076e07b98d9SDavid S. Miller 1077e9ee9efcSDavid Miller /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 1078e9ee9efcSDavid Miller * verifier will allow any alignment whatsoever. On platforms 1079e9ee9efcSDavid Miller * with strict alignment requirements for loads ands stores (such 1080e9ee9efcSDavid Miller * as sparc and mips) the verifier validates that all loads and 1081e9ee9efcSDavid Miller * stores provably follow this requirement. This flag turns that 1082e9ee9efcSDavid Miller * checking and enforcement off. 1083e9ee9efcSDavid Miller * 1084e9ee9efcSDavid Miller * It is mostly used for testing when we want to validate the 1085e9ee9efcSDavid Miller * context and memory access aspects of the verifier, but because 1086e9ee9efcSDavid Miller * of an unaligned access the alignment check would trigger before 1087e9ee9efcSDavid Miller * the one we are interested in. 1088e9ee9efcSDavid Miller */ 1089e9ee9efcSDavid Miller #define BPF_F_ANY_ALIGNMENT (1U << 1) 1090e9ee9efcSDavid Miller 1091c240eff6SJiong Wang /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1092c240eff6SJiong Wang * Verifier does sub-register def/use analysis and identifies instructions whose 1093c240eff6SJiong Wang * def only matters for low 32-bit, high 32-bit is never referenced later 1094c240eff6SJiong Wang * through implicit zero extension. Therefore verifier notifies JIT back-ends 1095c240eff6SJiong Wang * that it is safe to ignore clearing high 32-bit for these instructions. This 1096c240eff6SJiong Wang * saves some back-ends a lot of code-gen. However such optimization is not 1097c240eff6SJiong Wang * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1098c240eff6SJiong Wang * hence hasn't used verifier's analysis result. But, we really want to have a 1099c240eff6SJiong Wang * way to be able to verify the correctness of the described optimization on 1100c240eff6SJiong Wang * x86_64 on which testsuites are frequently exercised. 1101c240eff6SJiong Wang * 1102c240eff6SJiong Wang * So, this flag is introduced. Once it is set, verifier will randomize high 1103c240eff6SJiong Wang * 32-bit for those instructions who has been identified as safe to ignore them. 1104c240eff6SJiong Wang * Then, if verifier is not doing correct analysis, such randomization will 1105c240eff6SJiong Wang * regress tests to expose bugs. 1106c240eff6SJiong Wang */ 1107c240eff6SJiong Wang #define BPF_F_TEST_RND_HI32 (1U << 2) 1108c240eff6SJiong Wang 110910d274e8SAlexei Starovoitov /* The verifier internal test flag. Behavior is undefined */ 111010d274e8SAlexei Starovoitov #define BPF_F_TEST_STATE_FREQ (1U << 3) 111110d274e8SAlexei Starovoitov 11121e6c62a8SAlexei Starovoitov /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 11131e6c62a8SAlexei Starovoitov * restrict map and helper usage for such programs. Sleepable BPF programs can 11141e6c62a8SAlexei Starovoitov * only be attached to hooks where kernel execution context allows sleeping. 11151e6c62a8SAlexei Starovoitov * Such programs are allowed to use helpers that may sleep like 11161e6c62a8SAlexei Starovoitov * bpf_copy_from_user(). 11171e6c62a8SAlexei Starovoitov */ 11181e6c62a8SAlexei Starovoitov #define BPF_F_SLEEPABLE (1U << 4) 11191e6c62a8SAlexei Starovoitov 1120c2f2cdbeSLorenzo Bianconi /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1121c2f2cdbeSLorenzo Bianconi * fully support xdp frags. 1122c2f2cdbeSLorenzo Bianconi */ 1123c2f2cdbeSLorenzo Bianconi #define BPF_F_XDP_HAS_FRAGS (1U << 5) 1124c2f2cdbeSLorenzo Bianconi 11250dcac272SJiri Olsa /* link_create.kprobe_multi.flags used in LINK_CREATE command for 11260dcac272SJiri Olsa * BPF_TRACE_KPROBE_MULTI attach type to create return probe. 11270dcac272SJiri Olsa */ 11280dcac272SJiri Olsa #define BPF_F_KPROBE_MULTI_RETURN (1U << 0) 11290dcac272SJiri Olsa 1130d8eca5bbSDaniel Borkmann /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 11314976b718SHao Luo * the following extensions: 1132d8eca5bbSDaniel Borkmann * 1133387544bfSAlexei Starovoitov * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1134387544bfSAlexei Starovoitov * insn[0].imm: map fd or fd_idx 11354976b718SHao Luo * insn[1].imm: 0 11364976b718SHao Luo * insn[0].off: 0 11374976b718SHao Luo * insn[1].off: 0 11384976b718SHao Luo * ldimm64 rewrite: address of map 11394976b718SHao Luo * verifier type: CONST_PTR_TO_MAP 1140d8eca5bbSDaniel Borkmann */ 1141f1a66f85SDaniel Borkmann #define BPF_PSEUDO_MAP_FD 1 1142387544bfSAlexei Starovoitov #define BPF_PSEUDO_MAP_IDX 5 1143387544bfSAlexei Starovoitov 1144387544bfSAlexei Starovoitov /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1145387544bfSAlexei Starovoitov * insn[0].imm: map fd or fd_idx 11464976b718SHao Luo * insn[1].imm: offset into value 11474976b718SHao Luo * insn[0].off: 0 11484976b718SHao Luo * insn[1].off: 0 11494976b718SHao Luo * ldimm64 rewrite: address of map[0]+offset 11504976b718SHao Luo * verifier type: PTR_TO_MAP_VALUE 11514976b718SHao Luo */ 1152d8eca5bbSDaniel Borkmann #define BPF_PSEUDO_MAP_VALUE 2 1153387544bfSAlexei Starovoitov #define BPF_PSEUDO_MAP_IDX_VALUE 6 1154387544bfSAlexei Starovoitov 11554976b718SHao Luo /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 11564976b718SHao Luo * insn[0].imm: kernel btd id of VAR 11574976b718SHao Luo * insn[1].imm: 0 11584976b718SHao Luo * insn[0].off: 0 11594976b718SHao Luo * insn[1].off: 0 11604976b718SHao Luo * ldimm64 rewrite: address of the kernel variable 11614976b718SHao Luo * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 11624976b718SHao Luo * is struct/union. 11634976b718SHao Luo */ 11644976b718SHao Luo #define BPF_PSEUDO_BTF_ID 3 116569c087baSYonghong Song /* insn[0].src_reg: BPF_PSEUDO_FUNC 116669c087baSYonghong Song * insn[0].imm: insn offset to the func 116769c087baSYonghong Song * insn[1].imm: 0 116869c087baSYonghong Song * insn[0].off: 0 116969c087baSYonghong Song * insn[1].off: 0 117069c087baSYonghong Song * ldimm64 rewrite: address of the function 117169c087baSYonghong Song * verifier type: PTR_TO_FUNC. 117269c087baSYonghong Song */ 117369c087baSYonghong Song #define BPF_PSEUDO_FUNC 4 1174f1a66f85SDaniel Borkmann 1175cc8b0b92SAlexei Starovoitov /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1176cc8b0b92SAlexei Starovoitov * offset to another bpf function 1177cc8b0b92SAlexei Starovoitov */ 1178cc8b0b92SAlexei Starovoitov #define BPF_PSEUDO_CALL 1 1179e6ac2450SMartin KaFai Lau /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1180e6ac2450SMartin KaFai Lau * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1181e6ac2450SMartin KaFai Lau */ 1182e6ac2450SMartin KaFai Lau #define BPF_PSEUDO_KFUNC_CALL 2 1183cc8b0b92SAlexei Starovoitov 11843274f520SAlexei Starovoitov /* flags for BPF_MAP_UPDATE_ELEM command */ 11851aae4bddSAndrii Nakryiko enum { 11861aae4bddSAndrii Nakryiko BPF_ANY = 0, /* create new element or update existing */ 11871aae4bddSAndrii Nakryiko BPF_NOEXIST = 1, /* create new element if it didn't exist */ 11881aae4bddSAndrii Nakryiko BPF_EXIST = 2, /* update existing element */ 11891aae4bddSAndrii Nakryiko BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 11901aae4bddSAndrii Nakryiko }; 11913274f520SAlexei Starovoitov 119296eabe7aSMartin KaFai Lau /* flags for BPF_MAP_CREATE command */ 11931aae4bddSAndrii Nakryiko enum { 11941aae4bddSAndrii Nakryiko BPF_F_NO_PREALLOC = (1U << 0), 119529ba732aSMartin KaFai Lau /* Instead of having one common LRU list in the 11968f844938SMartin KaFai Lau * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 119729ba732aSMartin KaFai Lau * which can scale and perform better. 119829ba732aSMartin KaFai Lau * Note, the LRU nodes (including free nodes) cannot be moved 119929ba732aSMartin KaFai Lau * across different LRU lists. 120029ba732aSMartin KaFai Lau */ 12011aae4bddSAndrii Nakryiko BPF_F_NO_COMMON_LRU = (1U << 1), 120296eabe7aSMartin KaFai Lau /* Specify numa node during map creation */ 12031aae4bddSAndrii Nakryiko BPF_F_NUMA_NODE = (1U << 2), 1204cb4d2b3fSMartin KaFai Lau 1205591fe988SDaniel Borkmann /* Flags for accessing BPF object from syscall side. */ 12061aae4bddSAndrii Nakryiko BPF_F_RDONLY = (1U << 3), 12071aae4bddSAndrii Nakryiko BPF_F_WRONLY = (1U << 4), 12086e71b04aSChenbo Feng 1209615755a7SSong Liu /* Flag for stack_map, store build_id+offset instead of pointer */ 12101aae4bddSAndrii Nakryiko BPF_F_STACK_BUILD_ID = (1U << 5), 1211615755a7SSong Liu 121296b3b6c9SLorenz Bauer /* Zero-initialize hash function seed. This should only be used for testing. */ 12131aae4bddSAndrii Nakryiko BPF_F_ZERO_SEED = (1U << 6), 121496b3b6c9SLorenz Bauer 1215591fe988SDaniel Borkmann /* Flags for accessing BPF object from program side. */ 12161aae4bddSAndrii Nakryiko BPF_F_RDONLY_PROG = (1U << 7), 12171aae4bddSAndrii Nakryiko BPF_F_WRONLY_PROG = (1U << 8), 1218591fe988SDaniel Borkmann 12198f51dfc7SStanislav Fomichev /* Clone map from listener for newly accepted socket */ 12201aae4bddSAndrii Nakryiko BPF_F_CLONE = (1U << 9), 12218f51dfc7SStanislav Fomichev 1222fc970227SAndrii Nakryiko /* Enable memory-mapping BPF map */ 12231aae4bddSAndrii Nakryiko BPF_F_MMAPABLE = (1U << 10), 1224792cacccSSong Liu 1225792cacccSSong Liu /* Share perf_event among processes */ 1226792cacccSSong Liu BPF_F_PRESERVE_ELEMS = (1U << 11), 12274a8f87e6SDaniel Borkmann 12284a8f87e6SDaniel Borkmann /* Create a map that is suitable to be an inner map with dynamic max entries */ 12294a8f87e6SDaniel Borkmann BPF_F_INNER_MAP = (1U << 12), 12301aae4bddSAndrii Nakryiko }; 1231fc970227SAndrii Nakryiko 1232f5bfcd95SAndrey Ignatov /* Flags for BPF_PROG_QUERY. */ 1233f5bfcd95SAndrey Ignatov 1234f5bfcd95SAndrey Ignatov /* Query effective (directly attached + inherited from ancestor cgroups) 1235f5bfcd95SAndrey Ignatov * programs that will be executed for events within a cgroup. 1236f5bfcd95SAndrey Ignatov * attach_flags with this flag are returned only for directly attached programs. 1237f5bfcd95SAndrey Ignatov */ 12382f183360SLorenz Bauer #define BPF_F_QUERY_EFFECTIVE (1U << 0) 12392f183360SLorenz Bauer 12401b4d60ecSSong Liu /* Flags for BPF_PROG_TEST_RUN */ 12411b4d60ecSSong Liu 12421b4d60ecSSong Liu /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 12431b4d60ecSSong Liu #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1244b530e9e1SToke Høiland-Jørgensen /* If set, XDP frames will be transmitted after processing */ 1245b530e9e1SToke Høiland-Jørgensen #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 12461b4d60ecSSong Liu 1247d46edd67SSong Liu /* type for BPF_ENABLE_STATS */ 1248d46edd67SSong Liu enum bpf_stats_type { 1249d46edd67SSong Liu /* enabled run_time_ns and run_cnt */ 1250d46edd67SSong Liu BPF_STATS_RUN_TIME = 0, 1251d46edd67SSong Liu }; 1252d46edd67SSong Liu 1253615755a7SSong Liu enum bpf_stack_build_id_status { 1254615755a7SSong Liu /* user space need an empty entry to identify end of a trace */ 1255615755a7SSong Liu BPF_STACK_BUILD_ID_EMPTY = 0, 1256615755a7SSong Liu /* with valid build_id and offset */ 1257615755a7SSong Liu BPF_STACK_BUILD_ID_VALID = 1, 1258615755a7SSong Liu /* couldn't get build_id, fallback to ip */ 1259615755a7SSong Liu BPF_STACK_BUILD_ID_IP = 2, 1260615755a7SSong Liu }; 1261615755a7SSong Liu 1262615755a7SSong Liu #define BPF_BUILD_ID_SIZE 20 1263615755a7SSong Liu struct bpf_stack_build_id { 1264615755a7SSong Liu __s32 status; 1265615755a7SSong Liu unsigned char build_id[BPF_BUILD_ID_SIZE]; 1266615755a7SSong Liu union { 1267615755a7SSong Liu __u64 offset; 1268615755a7SSong Liu __u64 ip; 1269615755a7SSong Liu }; 1270615755a7SSong Liu }; 1271615755a7SSong Liu 12721aae4bddSAndrii Nakryiko #define BPF_OBJ_NAME_LEN 16U 12731aae4bddSAndrii Nakryiko 127499c55f7dSAlexei Starovoitov union bpf_attr { 127599c55f7dSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_CREATE command */ 127699c55f7dSAlexei Starovoitov __u32 map_type; /* one of enum bpf_map_type */ 127799c55f7dSAlexei Starovoitov __u32 key_size; /* size of key in bytes */ 127899c55f7dSAlexei Starovoitov __u32 value_size; /* size of value in bytes */ 127999c55f7dSAlexei Starovoitov __u32 max_entries; /* max number of entries in a map */ 128096eabe7aSMartin KaFai Lau __u32 map_flags; /* BPF_MAP_CREATE related 128196eabe7aSMartin KaFai Lau * flags defined above. 128296eabe7aSMartin KaFai Lau */ 128356f668dfSMartin KaFai Lau __u32 inner_map_fd; /* fd pointing to the inner map */ 128496eabe7aSMartin KaFai Lau __u32 numa_node; /* numa node (effective only if 128596eabe7aSMartin KaFai Lau * BPF_F_NUMA_NODE is set). 128696eabe7aSMartin KaFai Lau */ 1287067cae47SMartin KaFai Lau char map_name[BPF_OBJ_NAME_LEN]; 1288a3884572SJakub Kicinski __u32 map_ifindex; /* ifindex of netdev to create on */ 1289a26ca7c9SMartin KaFai Lau __u32 btf_fd; /* fd pointing to a BTF type data */ 12909b2cf328SMartin KaFai Lau __u32 btf_key_type_id; /* BTF type_id of the key */ 12919b2cf328SMartin KaFai Lau __u32 btf_value_type_id; /* BTF type_id of the value */ 129285d33df3SMartin KaFai Lau __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 129385d33df3SMartin KaFai Lau * struct stored as the 129485d33df3SMartin KaFai Lau * map value 129585d33df3SMartin KaFai Lau */ 12969330986cSJoanne Koong /* Any per-map-type extra fields 12979330986cSJoanne Koong * 12989330986cSJoanne Koong * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 12999330986cSJoanne Koong * number of hash functions (if 0, the bloom filter will default 13009330986cSJoanne Koong * to using 5 hash functions). 13019330986cSJoanne Koong */ 13029330986cSJoanne Koong __u64 map_extra; 130399c55f7dSAlexei Starovoitov }; 1304db20fd2bSAlexei Starovoitov 1305db20fd2bSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1306db20fd2bSAlexei Starovoitov __u32 map_fd; 1307db20fd2bSAlexei Starovoitov __aligned_u64 key; 1308db20fd2bSAlexei Starovoitov union { 1309db20fd2bSAlexei Starovoitov __aligned_u64 value; 1310db20fd2bSAlexei Starovoitov __aligned_u64 next_key; 1311db20fd2bSAlexei Starovoitov }; 13123274f520SAlexei Starovoitov __u64 flags; 1313db20fd2bSAlexei Starovoitov }; 131409756af4SAlexei Starovoitov 1315cb4d03abSBrian Vazquez struct { /* struct used by BPF_MAP_*_BATCH commands */ 1316cb4d03abSBrian Vazquez __aligned_u64 in_batch; /* start batch, 1317cb4d03abSBrian Vazquez * NULL to start from beginning 1318cb4d03abSBrian Vazquez */ 1319cb4d03abSBrian Vazquez __aligned_u64 out_batch; /* output: next start batch */ 1320cb4d03abSBrian Vazquez __aligned_u64 keys; 1321cb4d03abSBrian Vazquez __aligned_u64 values; 1322cb4d03abSBrian Vazquez __u32 count; /* input/output: 1323cb4d03abSBrian Vazquez * input: # of key/value 1324cb4d03abSBrian Vazquez * elements 1325cb4d03abSBrian Vazquez * output: # of filled elements 1326cb4d03abSBrian Vazquez */ 1327cb4d03abSBrian Vazquez __u32 map_fd; 1328cb4d03abSBrian Vazquez __u64 elem_flags; 1329cb4d03abSBrian Vazquez __u64 flags; 1330cb4d03abSBrian Vazquez } batch; 1331cb4d03abSBrian Vazquez 133209756af4SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_LOAD command */ 133309756af4SAlexei Starovoitov __u32 prog_type; /* one of enum bpf_prog_type */ 133409756af4SAlexei Starovoitov __u32 insn_cnt; 133509756af4SAlexei Starovoitov __aligned_u64 insns; 133609756af4SAlexei Starovoitov __aligned_u64 license; 1337cbd35700SAlexei Starovoitov __u32 log_level; /* verbosity level of verifier */ 1338cbd35700SAlexei Starovoitov __u32 log_size; /* size of user buffer */ 1339cbd35700SAlexei Starovoitov __aligned_u64 log_buf; /* user supplied buffer */ 13406c4fc209SDaniel Borkmann __u32 kern_version; /* not used */ 1341e07b98d9SDavid S. Miller __u32 prog_flags; 1342067cae47SMartin KaFai Lau char prog_name[BPF_OBJ_NAME_LEN]; 13431f6f4cb7SJakub Kicinski __u32 prog_ifindex; /* ifindex of netdev to prep for */ 13445e43f899SAndrey Ignatov /* For some prog types expected attach type must be known at 13455e43f899SAndrey Ignatov * load time to verify attach type specific parts of prog 13465e43f899SAndrey Ignatov * (context accesses, allowed helpers, etc). 13475e43f899SAndrey Ignatov */ 13485e43f899SAndrey Ignatov __u32 expected_attach_type; 1349838e9690SYonghong Song __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1350838e9690SYonghong Song __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1351838e9690SYonghong Song __aligned_u64 func_info; /* func info */ 1352838e9690SYonghong Song __u32 func_info_cnt; /* number of bpf_func_info records */ 1353c454a46bSMartin KaFai Lau __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1354c454a46bSMartin KaFai Lau __aligned_u64 line_info; /* line info */ 1355c454a46bSMartin KaFai Lau __u32 line_info_cnt; /* number of bpf_line_info records */ 1356ccfe29ebSAlexei Starovoitov __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1357290248a5SAndrii Nakryiko union { 1358290248a5SAndrii Nakryiko /* valid prog_fd to attach to bpf prog */ 1359290248a5SAndrii Nakryiko __u32 attach_prog_fd; 1360290248a5SAndrii Nakryiko /* or valid module BTF object fd or 0 to attach to vmlinux */ 1361290248a5SAndrii Nakryiko __u32 attach_btf_obj_fd; 1362290248a5SAndrii Nakryiko }; 1363fbd94c7aSAlexei Starovoitov __u32 core_relo_cnt; /* number of bpf_core_relo */ 1364387544bfSAlexei Starovoitov __aligned_u64 fd_array; /* array of FDs */ 1365fbd94c7aSAlexei Starovoitov __aligned_u64 core_relos; 1366fbd94c7aSAlexei Starovoitov __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 136709756af4SAlexei Starovoitov }; 1368b2197755SDaniel Borkmann 1369b2197755SDaniel Borkmann struct { /* anonymous struct used by BPF_OBJ_* commands */ 1370b2197755SDaniel Borkmann __aligned_u64 pathname; 1371b2197755SDaniel Borkmann __u32 bpf_fd; 13726e71b04aSChenbo Feng __u32 file_flags; 1373b2197755SDaniel Borkmann }; 1374f4324551SDaniel Mack 1375f4324551SDaniel Mack struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1376f4324551SDaniel Mack __u32 target_fd; /* container object to attach to */ 1377f4324551SDaniel Mack __u32 attach_bpf_fd; /* eBPF program to attach */ 1378f4324551SDaniel Mack __u32 attach_type; 13797f677633SAlexei Starovoitov __u32 attach_flags; 13807dd68b32SAndrey Ignatov __u32 replace_bpf_fd; /* previously attached eBPF 13817dd68b32SAndrey Ignatov * program to replace if 13827dd68b32SAndrey Ignatov * BPF_F_REPLACE is used 13837dd68b32SAndrey Ignatov */ 1384f4324551SDaniel Mack }; 13851cf1cae9SAlexei Starovoitov 13861cf1cae9SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 13871cf1cae9SAlexei Starovoitov __u32 prog_fd; 13881cf1cae9SAlexei Starovoitov __u32 retval; 1389b5a36b1eSLorenz Bauer __u32 data_size_in; /* input: len of data_in */ 1390b5a36b1eSLorenz Bauer __u32 data_size_out; /* input/output: len of data_out 1391b5a36b1eSLorenz Bauer * returns ENOSPC if data_out 1392b5a36b1eSLorenz Bauer * is too small. 1393b5a36b1eSLorenz Bauer */ 13941cf1cae9SAlexei Starovoitov __aligned_u64 data_in; 13951cf1cae9SAlexei Starovoitov __aligned_u64 data_out; 13961cf1cae9SAlexei Starovoitov __u32 repeat; 13971cf1cae9SAlexei Starovoitov __u32 duration; 1398b0b9395dSStanislav Fomichev __u32 ctx_size_in; /* input: len of ctx_in */ 1399b0b9395dSStanislav Fomichev __u32 ctx_size_out; /* input/output: len of ctx_out 1400b0b9395dSStanislav Fomichev * returns ENOSPC if ctx_out 1401b0b9395dSStanislav Fomichev * is too small. 1402b0b9395dSStanislav Fomichev */ 1403b0b9395dSStanislav Fomichev __aligned_u64 ctx_in; 1404b0b9395dSStanislav Fomichev __aligned_u64 ctx_out; 14051b4d60ecSSong Liu __u32 flags; 14061b4d60ecSSong Liu __u32 cpu; 1407b530e9e1SToke Høiland-Jørgensen __u32 batch_size; 14081cf1cae9SAlexei Starovoitov } test; 140934ad5580SMartin KaFai Lau 1410b16d9aa4SMartin KaFai Lau struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1411b16d9aa4SMartin KaFai Lau union { 141234ad5580SMartin KaFai Lau __u32 start_id; 1413b16d9aa4SMartin KaFai Lau __u32 prog_id; 1414bd5f5f4eSMartin KaFai Lau __u32 map_id; 141578958fcaSMartin KaFai Lau __u32 btf_id; 1416a3b80e10SAndrii Nakryiko __u32 link_id; 1417b16d9aa4SMartin KaFai Lau }; 141834ad5580SMartin KaFai Lau __u32 next_id; 14196e71b04aSChenbo Feng __u32 open_flags; 142034ad5580SMartin KaFai Lau }; 14211e270976SMartin KaFai Lau 14221e270976SMartin KaFai Lau struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 14231e270976SMartin KaFai Lau __u32 bpf_fd; 14241e270976SMartin KaFai Lau __u32 info_len; 14251e270976SMartin KaFai Lau __aligned_u64 info; 14261e270976SMartin KaFai Lau } info; 1427468e2f64SAlexei Starovoitov 1428468e2f64SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1429468e2f64SAlexei Starovoitov __u32 target_fd; /* container object to query */ 1430468e2f64SAlexei Starovoitov __u32 attach_type; 1431468e2f64SAlexei Starovoitov __u32 query_flags; 1432468e2f64SAlexei Starovoitov __u32 attach_flags; 1433468e2f64SAlexei Starovoitov __aligned_u64 prog_ids; 1434468e2f64SAlexei Starovoitov __u32 prog_cnt; 1435b79c9fc9SStanislav Fomichev __aligned_u64 prog_attach_flags; /* output: per-program attach_flags */ 1436468e2f64SAlexei Starovoitov } query; 1437c4f6699dSAlexei Starovoitov 1438af6eea57SAndrii Nakryiko struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1439c4f6699dSAlexei Starovoitov __u64 name; 1440c4f6699dSAlexei Starovoitov __u32 prog_fd; 1441c4f6699dSAlexei Starovoitov } raw_tracepoint; 1442f56a653cSMartin KaFai Lau 1443f56a653cSMartin KaFai Lau struct { /* anonymous struct for BPF_BTF_LOAD */ 1444f56a653cSMartin KaFai Lau __aligned_u64 btf; 1445f56a653cSMartin KaFai Lau __aligned_u64 btf_log_buf; 1446f56a653cSMartin KaFai Lau __u32 btf_size; 1447f56a653cSMartin KaFai Lau __u32 btf_log_size; 1448f56a653cSMartin KaFai Lau __u32 btf_log_level; 1449f56a653cSMartin KaFai Lau }; 145041bdc4b4SYonghong Song 145141bdc4b4SYonghong Song struct { 145241bdc4b4SYonghong Song __u32 pid; /* input: pid */ 145341bdc4b4SYonghong Song __u32 fd; /* input: fd */ 145441bdc4b4SYonghong Song __u32 flags; /* input: flags */ 145541bdc4b4SYonghong Song __u32 buf_len; /* input/output: buf len */ 145641bdc4b4SYonghong Song __aligned_u64 buf; /* input/output: 145741bdc4b4SYonghong Song * tp_name for tracepoint 145841bdc4b4SYonghong Song * symbol for kprobe 145941bdc4b4SYonghong Song * filename for uprobe 146041bdc4b4SYonghong Song */ 146141bdc4b4SYonghong Song __u32 prog_id; /* output: prod_id */ 146241bdc4b4SYonghong Song __u32 fd_type; /* output: BPF_FD_TYPE_* */ 146341bdc4b4SYonghong Song __u64 probe_offset; /* output: probe_offset */ 146441bdc4b4SYonghong Song __u64 probe_addr; /* output: probe_addr */ 146541bdc4b4SYonghong Song } task_fd_query; 1466af6eea57SAndrii Nakryiko 1467af6eea57SAndrii Nakryiko struct { /* struct used by BPF_LINK_CREATE command */ 1468af6eea57SAndrii Nakryiko __u32 prog_fd; /* eBPF program to attach */ 1469aa8d3a71SAndrii Nakryiko union { 1470af6eea57SAndrii Nakryiko __u32 target_fd; /* object to attach to */ 1471aa8d3a71SAndrii Nakryiko __u32 target_ifindex; /* target ifindex */ 1472aa8d3a71SAndrii Nakryiko }; 1473af6eea57SAndrii Nakryiko __u32 attach_type; /* attach type */ 1474af6eea57SAndrii Nakryiko __u32 flags; /* extra flags */ 14754a1e7c0cSToke Høiland-Jørgensen union { 14764a1e7c0cSToke Høiland-Jørgensen __u32 target_btf_id; /* btf_id of target to attach to */ 14774a1e7c0cSToke Høiland-Jørgensen struct { 14785e7b3020SYonghong Song __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 14795e7b3020SYonghong Song __u32 iter_info_len; /* iter_info length */ 14804a1e7c0cSToke Høiland-Jørgensen }; 148182e6b1eeSAndrii Nakryiko struct { 148282e6b1eeSAndrii Nakryiko /* black box user-provided value passed through 148382e6b1eeSAndrii Nakryiko * to BPF program at the execution time and 148482e6b1eeSAndrii Nakryiko * accessible through bpf_get_attach_cookie() BPF helper 148582e6b1eeSAndrii Nakryiko */ 148682e6b1eeSAndrii Nakryiko __u64 bpf_cookie; 148782e6b1eeSAndrii Nakryiko } perf_event; 14880dcac272SJiri Olsa struct { 14890dcac272SJiri Olsa __u32 flags; 14900dcac272SJiri Olsa __u32 cnt; 14910dcac272SJiri Olsa __aligned_u64 syms; 14920dcac272SJiri Olsa __aligned_u64 addrs; 1493ca74823cSJiri Olsa __aligned_u64 cookies; 14940dcac272SJiri Olsa } kprobe_multi; 14952fcc8241SKui-Feng Lee struct { 14962fcc8241SKui-Feng Lee /* this is overlaid with the target_btf_id above. */ 14972fcc8241SKui-Feng Lee __u32 target_btf_id; 14982fcc8241SKui-Feng Lee /* black box user-provided value passed through 14992fcc8241SKui-Feng Lee * to BPF program at the execution time and 15002fcc8241SKui-Feng Lee * accessible through bpf_get_attach_cookie() BPF helper 15012fcc8241SKui-Feng Lee */ 15022fcc8241SKui-Feng Lee __u64 cookie; 15032fcc8241SKui-Feng Lee } tracing; 15044a1e7c0cSToke Høiland-Jørgensen }; 1505af6eea57SAndrii Nakryiko } link_create; 15060c991ebcSAndrii Nakryiko 15070c991ebcSAndrii Nakryiko struct { /* struct used by BPF_LINK_UPDATE command */ 15080c991ebcSAndrii Nakryiko __u32 link_fd; /* link fd */ 15090c991ebcSAndrii Nakryiko /* new program fd to update link with */ 15100c991ebcSAndrii Nakryiko __u32 new_prog_fd; 15110c991ebcSAndrii Nakryiko __u32 flags; /* extra flags */ 15120c991ebcSAndrii Nakryiko /* expected link's program fd; is specified only if 15130c991ebcSAndrii Nakryiko * BPF_F_REPLACE flag is set in flags */ 15140c991ebcSAndrii Nakryiko __u32 old_prog_fd; 15150c991ebcSAndrii Nakryiko } link_update; 15160c991ebcSAndrii Nakryiko 151773b11c2aSAndrii Nakryiko struct { 151873b11c2aSAndrii Nakryiko __u32 link_fd; 151973b11c2aSAndrii Nakryiko } link_detach; 152073b11c2aSAndrii Nakryiko 1521d46edd67SSong Liu struct { /* struct used by BPF_ENABLE_STATS command */ 1522d46edd67SSong Liu __u32 type; 1523d46edd67SSong Liu } enable_stats; 1524d46edd67SSong Liu 1525ac51d99bSYonghong Song struct { /* struct used by BPF_ITER_CREATE command */ 1526ac51d99bSYonghong Song __u32 link_fd; 1527ac51d99bSYonghong Song __u32 flags; 1528ac51d99bSYonghong Song } iter_create; 1529ac51d99bSYonghong Song 1530ef15314aSYiFei Zhu struct { /* struct used by BPF_PROG_BIND_MAP command */ 1531ef15314aSYiFei Zhu __u32 prog_fd; 1532ef15314aSYiFei Zhu __u32 map_fd; 1533ef15314aSYiFei Zhu __u32 flags; /* extra flags */ 1534ef15314aSYiFei Zhu } prog_bind_map; 1535ef15314aSYiFei Zhu 153699c55f7dSAlexei Starovoitov } __attribute__((aligned(8))); 153799c55f7dSAlexei Starovoitov 153856a092c8SQuentin Monnet /* The description below is an attempt at providing documentation to eBPF 153956a092c8SQuentin Monnet * developers about the multiple available eBPF helper functions. It can be 154056a092c8SQuentin Monnet * parsed and used to produce a manual page. The workflow is the following, 154156a092c8SQuentin Monnet * and requires the rst2man utility: 1542ebb676daSThomas Graf * 1543923a932cSJoe Stringer * $ ./scripts/bpf_doc.py \ 154456a092c8SQuentin Monnet * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 154556a092c8SQuentin Monnet * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 154656a092c8SQuentin Monnet * $ man /tmp/bpf-helpers.7 1547ebb676daSThomas Graf * 154856a092c8SQuentin Monnet * Note that in order to produce this external documentation, some RST 154956a092c8SQuentin Monnet * formatting is used in the descriptions to get "bold" and "italics" in 155056a092c8SQuentin Monnet * manual pages. Also note that the few trailing white spaces are 155156a092c8SQuentin Monnet * intentional, removing them would break paragraphs for rst2man. 1552ebb676daSThomas Graf * 155356a092c8SQuentin Monnet * Start of BPF helper function descriptions: 1554ad4a5223SQuentin Monnet * 1555ad4a5223SQuentin Monnet * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1556ad4a5223SQuentin Monnet * Description 1557ad4a5223SQuentin Monnet * Perform a lookup in *map* for an entry associated to *key*. 1558ad4a5223SQuentin Monnet * Return 1559ad4a5223SQuentin Monnet * Map value associated to *key*, or **NULL** if no entry was 1560ad4a5223SQuentin Monnet * found. 1561ad4a5223SQuentin Monnet * 1562bdb7b79bSAndrii Nakryiko * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1563ad4a5223SQuentin Monnet * Description 1564ad4a5223SQuentin Monnet * Add or update the value of the entry associated to *key* in 1565ad4a5223SQuentin Monnet * *map* with *value*. *flags* is one of: 1566ad4a5223SQuentin Monnet * 1567ad4a5223SQuentin Monnet * **BPF_NOEXIST** 1568ad4a5223SQuentin Monnet * The entry for *key* must not exist in the map. 1569ad4a5223SQuentin Monnet * **BPF_EXIST** 1570ad4a5223SQuentin Monnet * The entry for *key* must already exist in the map. 1571ad4a5223SQuentin Monnet * **BPF_ANY** 1572ad4a5223SQuentin Monnet * No condition on the existence of the entry for *key*. 1573ad4a5223SQuentin Monnet * 1574ad4a5223SQuentin Monnet * Flag value **BPF_NOEXIST** cannot be used for maps of types 1575ad4a5223SQuentin Monnet * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1576ad4a5223SQuentin Monnet * elements always exist), the helper would return an error. 1577ad4a5223SQuentin Monnet * Return 1578ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1579ad4a5223SQuentin Monnet * 1580bdb7b79bSAndrii Nakryiko * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1581ad4a5223SQuentin Monnet * Description 1582ad4a5223SQuentin Monnet * Delete entry with *key* from *map*. 1583ad4a5223SQuentin Monnet * Return 1584ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1585ad4a5223SQuentin Monnet * 1586bdb7b79bSAndrii Nakryiko * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1587ad4a5223SQuentin Monnet * Description 1588ad4a5223SQuentin Monnet * For tracing programs, safely attempt to read *size* bytes from 15896ae08ae3SDaniel Borkmann * kernel space address *unsafe_ptr* and store the data in *dst*. 15906ae08ae3SDaniel Borkmann * 1591ab8d7809SQuentin Monnet * Generally, use **bpf_probe_read_user**\ () or 1592ab8d7809SQuentin Monnet * **bpf_probe_read_kernel**\ () instead. 1593ad4a5223SQuentin Monnet * Return 1594ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1595ad4a5223SQuentin Monnet * 1596ad4a5223SQuentin Monnet * u64 bpf_ktime_get_ns(void) 1597ad4a5223SQuentin Monnet * Description 1598ad4a5223SQuentin Monnet * Return the time elapsed since system boot, in nanoseconds. 159971d19214SMaciej Żenczykowski * Does not include time the system was suspended. 1600ab8d7809SQuentin Monnet * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1601ad4a5223SQuentin Monnet * Return 1602ad4a5223SQuentin Monnet * Current *ktime*. 1603ad4a5223SQuentin Monnet * 1604bdb7b79bSAndrii Nakryiko * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1605ad4a5223SQuentin Monnet * Description 1606ad4a5223SQuentin Monnet * This helper is a "printk()-like" facility for debugging. It 1607ad4a5223SQuentin Monnet * prints a message defined by format *fmt* (of size *fmt_size*) 1608ad4a5223SQuentin Monnet * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 1609ad4a5223SQuentin Monnet * available. It can take up to three additional **u64** 1610ad4a5223SQuentin Monnet * arguments (as an eBPF helpers, the total number of arguments is 1611ad4a5223SQuentin Monnet * limited to five). 1612ad4a5223SQuentin Monnet * 1613ad4a5223SQuentin Monnet * Each time the helper is called, it appends a line to the trace. 161455c33dfbSPeter Wu * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 161555c33dfbSPeter Wu * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 1616ad4a5223SQuentin Monnet * The format of the trace is customizable, and the exact output 1617ad4a5223SQuentin Monnet * one will get depends on the options set in 1618ad4a5223SQuentin Monnet * *\/sys/kernel/debug/tracing/trace_options* (see also the 1619ad4a5223SQuentin Monnet * *README* file under the same directory). However, it usually 1620ad4a5223SQuentin Monnet * defaults to something like: 1621ad4a5223SQuentin Monnet * 1622ad4a5223SQuentin Monnet * :: 1623ad4a5223SQuentin Monnet * 1624ad4a5223SQuentin Monnet * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1625ad4a5223SQuentin Monnet * 1626ad4a5223SQuentin Monnet * In the above: 1627ad4a5223SQuentin Monnet * 1628ad4a5223SQuentin Monnet * * ``telnet`` is the name of the current task. 1629ad4a5223SQuentin Monnet * * ``470`` is the PID of the current task. 1630ad4a5223SQuentin Monnet * * ``001`` is the CPU number on which the task is 1631ad4a5223SQuentin Monnet * running. 1632ad4a5223SQuentin Monnet * * In ``.N..``, each character refers to a set of 1633ad4a5223SQuentin Monnet * options (whether irqs are enabled, scheduling 1634ad4a5223SQuentin Monnet * options, whether hard/softirqs are running, level of 1635ad4a5223SQuentin Monnet * preempt_disabled respectively). **N** means that 1636ad4a5223SQuentin Monnet * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1637ad4a5223SQuentin Monnet * are set. 1638ad4a5223SQuentin Monnet * * ``419421.045894`` is a timestamp. 1639ad4a5223SQuentin Monnet * * ``0x00000001`` is a fake value used by BPF for the 1640ad4a5223SQuentin Monnet * instruction pointer register. 1641ad4a5223SQuentin Monnet * * ``<formatted msg>`` is the message formatted with 1642ad4a5223SQuentin Monnet * *fmt*. 1643ad4a5223SQuentin Monnet * 1644ad4a5223SQuentin Monnet * The conversion specifiers supported by *fmt* are similar, but 1645ad4a5223SQuentin Monnet * more limited than for printk(). They are **%d**, **%i**, 1646ad4a5223SQuentin Monnet * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1647ad4a5223SQuentin Monnet * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1648ad4a5223SQuentin Monnet * of field, padding with zeroes, etc.) is available, and the 1649ad4a5223SQuentin Monnet * helper will return **-EINVAL** (but print nothing) if it 1650ad4a5223SQuentin Monnet * encounters an unknown specifier. 1651ad4a5223SQuentin Monnet * 1652ad4a5223SQuentin Monnet * Also, note that **bpf_trace_printk**\ () is slow, and should 1653ad4a5223SQuentin Monnet * only be used for debugging purposes. For this reason, a notice 1654b16fc097STobias Klauser * block (spanning several lines) is printed to kernel logs and 1655ad4a5223SQuentin Monnet * states that the helper should not be used "for production use" 1656ad4a5223SQuentin Monnet * the first time this helper is used (or more precisely, when 1657ad4a5223SQuentin Monnet * **trace_printk**\ () buffers are allocated). For passing values 1658ad4a5223SQuentin Monnet * to user space, perf events should be preferred. 1659ad4a5223SQuentin Monnet * Return 1660ad4a5223SQuentin Monnet * The number of bytes written to the buffer, or a negative error 1661ad4a5223SQuentin Monnet * in case of failure. 1662ad4a5223SQuentin Monnet * 16631fdd08beSQuentin Monnet * u32 bpf_get_prandom_u32(void) 16641fdd08beSQuentin Monnet * Description 16651fdd08beSQuentin Monnet * Get a pseudo-random number. 16661fdd08beSQuentin Monnet * 16671fdd08beSQuentin Monnet * From a security point of view, this helper uses its own 16681fdd08beSQuentin Monnet * pseudo-random internal state, and cannot be used to infer the 16691fdd08beSQuentin Monnet * seed of other random functions in the kernel. However, it is 16701fdd08beSQuentin Monnet * essential to note that the generator used by the helper is not 16711fdd08beSQuentin Monnet * cryptographically secure. 16721fdd08beSQuentin Monnet * Return 16731fdd08beSQuentin Monnet * A random 32-bit unsigned value. 16741fdd08beSQuentin Monnet * 16751fdd08beSQuentin Monnet * u32 bpf_get_smp_processor_id(void) 16761fdd08beSQuentin Monnet * Description 16771fdd08beSQuentin Monnet * Get the SMP (symmetric multiprocessing) processor id. Note that 167833656275SMatteo Croce * all programs run with migration disabled, which means that the 16791fdd08beSQuentin Monnet * SMP processor id is stable during all the execution of the 16801fdd08beSQuentin Monnet * program. 16811fdd08beSQuentin Monnet * Return 16821fdd08beSQuentin Monnet * The SMP id of the processor running the program. 16831fdd08beSQuentin Monnet * 1684bdb7b79bSAndrii Nakryiko * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1685ad4a5223SQuentin Monnet * Description 1686ad4a5223SQuentin Monnet * Store *len* bytes from address *from* into the packet 1687ad4a5223SQuentin Monnet * associated to *skb*, at *offset*. *flags* are a combination of 1688ad4a5223SQuentin Monnet * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1689ad4a5223SQuentin Monnet * checksum for the packet after storing the bytes) and 1690ad4a5223SQuentin Monnet * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1691ad4a5223SQuentin Monnet * **->swhash** and *skb*\ **->l4hash** to 0). 1692ad4a5223SQuentin Monnet * 169332e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1694ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1695ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1696ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1697ad4a5223SQuentin Monnet * direct packet access. 1698ad4a5223SQuentin Monnet * Return 1699ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1700ad4a5223SQuentin Monnet * 1701bdb7b79bSAndrii Nakryiko * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1702ad4a5223SQuentin Monnet * Description 1703ad4a5223SQuentin Monnet * Recompute the layer 3 (e.g. IP) checksum for the packet 1704ad4a5223SQuentin Monnet * associated to *skb*. Computation is incremental, so the helper 1705ad4a5223SQuentin Monnet * must know the former value of the header field that was 1706ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 1707ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored in *size*. 1708ad4a5223SQuentin Monnet * Alternatively, it is possible to store the difference between 1709ad4a5223SQuentin Monnet * the previous and the new values of the header field in *to*, by 1710ad4a5223SQuentin Monnet * setting *from* and *size* to 0. For both methods, *offset* 1711ad4a5223SQuentin Monnet * indicates the location of the IP checksum within the packet. 1712ad4a5223SQuentin Monnet * 1713ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 1714ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 1715ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 1716ad4a5223SQuentin Monnet * checksum to update. 1717ad4a5223SQuentin Monnet * 171832e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1719ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1720ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1721ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1722ad4a5223SQuentin Monnet * direct packet access. 1723ad4a5223SQuentin Monnet * Return 1724ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1725ad4a5223SQuentin Monnet * 1726bdb7b79bSAndrii Nakryiko * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1727ad4a5223SQuentin Monnet * Description 1728ad4a5223SQuentin Monnet * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1729ad4a5223SQuentin Monnet * packet associated to *skb*. Computation is incremental, so the 1730ad4a5223SQuentin Monnet * helper must know the former value of the header field that was 1731ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 1732ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored on the lowest 1733ad4a5223SQuentin Monnet * four bits of *flags*. Alternatively, it is possible to store 1734ad4a5223SQuentin Monnet * the difference between the previous and the new values of the 1735ad4a5223SQuentin Monnet * header field in *to*, by setting *from* and the four lowest 1736ad4a5223SQuentin Monnet * bits of *flags* to 0. For both methods, *offset* indicates the 1737ad4a5223SQuentin Monnet * location of the IP checksum within the packet. In addition to 1738ad4a5223SQuentin Monnet * the size of the field, *flags* can be added (bitwise OR) actual 1739ad4a5223SQuentin Monnet * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1740ad4a5223SQuentin Monnet * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1741ad4a5223SQuentin Monnet * for updates resulting in a null checksum the value is set to 1742ad4a5223SQuentin Monnet * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1743ad4a5223SQuentin Monnet * the checksum is to be computed against a pseudo-header. 1744ad4a5223SQuentin Monnet * 1745ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 1746ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 1747ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 1748ad4a5223SQuentin Monnet * checksum to update. 1749ad4a5223SQuentin Monnet * 175032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1751ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1752ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1753ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1754ad4a5223SQuentin Monnet * direct packet access. 1755ad4a5223SQuentin Monnet * Return 1756ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1757ad4a5223SQuentin Monnet * 1758bdb7b79bSAndrii Nakryiko * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 1759ad4a5223SQuentin Monnet * Description 1760ad4a5223SQuentin Monnet * This special helper is used to trigger a "tail call", or in 1761ad4a5223SQuentin Monnet * other words, to jump into another eBPF program. The same stack 1762ad4a5223SQuentin Monnet * frame is used (but values on stack and in registers for the 1763ad4a5223SQuentin Monnet * caller are not accessible to the callee). This mechanism allows 1764ad4a5223SQuentin Monnet * for program chaining, either for raising the maximum number of 1765ad4a5223SQuentin Monnet * available eBPF instructions, or to execute given programs in 1766ad4a5223SQuentin Monnet * conditional blocks. For security reasons, there is an upper 1767ad4a5223SQuentin Monnet * limit to the number of successive tail calls that can be 1768ad4a5223SQuentin Monnet * performed. 1769ad4a5223SQuentin Monnet * 1770ad4a5223SQuentin Monnet * Upon call of this helper, the program attempts to jump into a 1771ad4a5223SQuentin Monnet * program referenced at index *index* in *prog_array_map*, a 1772ad4a5223SQuentin Monnet * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 1773ad4a5223SQuentin Monnet * *ctx*, a pointer to the context. 1774ad4a5223SQuentin Monnet * 1775ad4a5223SQuentin Monnet * If the call succeeds, the kernel immediately runs the first 1776ad4a5223SQuentin Monnet * instruction of the new program. This is not a function call, 1777ad4a5223SQuentin Monnet * and it never returns to the previous program. If the call 1778ad4a5223SQuentin Monnet * fails, then the helper has no effect, and the caller continues 1779ad4a5223SQuentin Monnet * to run its subsequent instructions. A call can fail if the 1780ad4a5223SQuentin Monnet * destination program for the jump does not exist (i.e. *index* 1781ad4a5223SQuentin Monnet * is superior to the number of entries in *prog_array_map*), or 1782ad4a5223SQuentin Monnet * if the maximum number of tail calls has been reached for this 1783ad4a5223SQuentin Monnet * chain of programs. This limit is defined in the kernel by the 1784ad4a5223SQuentin Monnet * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 1785ebf7f6f0STiezhu Yang * which is currently set to 33. 1786ad4a5223SQuentin Monnet * Return 1787ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1788ad4a5223SQuentin Monnet * 1789bdb7b79bSAndrii Nakryiko * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 1790ad4a5223SQuentin Monnet * Description 1791ad4a5223SQuentin Monnet * Clone and redirect the packet associated to *skb* to another 1792ad4a5223SQuentin Monnet * net device of index *ifindex*. Both ingress and egress 1793ad4a5223SQuentin Monnet * interfaces can be used for redirection. The **BPF_F_INGRESS** 1794ad4a5223SQuentin Monnet * value in *flags* is used to make the distinction (ingress path 1795ad4a5223SQuentin Monnet * is selected if the flag is present, egress path otherwise). 1796ad4a5223SQuentin Monnet * This is the only flag supported for now. 1797ad4a5223SQuentin Monnet * 1798ad4a5223SQuentin Monnet * In comparison with **bpf_redirect**\ () helper, 1799ad4a5223SQuentin Monnet * **bpf_clone_redirect**\ () has the associated cost of 1800ad4a5223SQuentin Monnet * duplicating the packet buffer, but this can be executed out of 1801ad4a5223SQuentin Monnet * the eBPF program. Conversely, **bpf_redirect**\ () is more 1802ad4a5223SQuentin Monnet * efficient, but it is handled through an action code where the 1803ad4a5223SQuentin Monnet * redirection happens only after the eBPF program has returned. 1804ad4a5223SQuentin Monnet * 180532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1806ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1807ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1808ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1809ad4a5223SQuentin Monnet * direct packet access. 1810ad4a5223SQuentin Monnet * Return 1811ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1812c456dec4SQuentin Monnet * 1813c456dec4SQuentin Monnet * u64 bpf_get_current_pid_tgid(void) 1814e40fbbf0SUsama Arif * Description 1815e40fbbf0SUsama Arif * Get the current pid and tgid. 1816c456dec4SQuentin Monnet * Return 1817c456dec4SQuentin Monnet * A 64-bit integer containing the current tgid and pid, and 1818c456dec4SQuentin Monnet * created as such: 1819c456dec4SQuentin Monnet * *current_task*\ **->tgid << 32 \|** 1820c456dec4SQuentin Monnet * *current_task*\ **->pid**. 1821c456dec4SQuentin Monnet * 1822c456dec4SQuentin Monnet * u64 bpf_get_current_uid_gid(void) 1823e40fbbf0SUsama Arif * Description 1824e40fbbf0SUsama Arif * Get the current uid and gid. 1825c456dec4SQuentin Monnet * Return 1826c456dec4SQuentin Monnet * A 64-bit integer containing the current GID and UID, and 1827c456dec4SQuentin Monnet * created as such: *current_gid* **<< 32 \|** *current_uid*. 1828c456dec4SQuentin Monnet * 1829bdb7b79bSAndrii Nakryiko * long bpf_get_current_comm(void *buf, u32 size_of_buf) 1830c456dec4SQuentin Monnet * Description 1831c456dec4SQuentin Monnet * Copy the **comm** attribute of the current task into *buf* of 1832c456dec4SQuentin Monnet * *size_of_buf*. The **comm** attribute contains the name of 1833c456dec4SQuentin Monnet * the executable (excluding the path) for the current task. The 1834c456dec4SQuentin Monnet * *size_of_buf* must be strictly positive. On success, the 1835c456dec4SQuentin Monnet * helper makes sure that the *buf* is NUL-terminated. On failure, 1836c456dec4SQuentin Monnet * it is filled with zeroes. 1837c456dec4SQuentin Monnet * Return 1838c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1839c456dec4SQuentin Monnet * 18401fdd08beSQuentin Monnet * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 18411fdd08beSQuentin Monnet * Description 18421fdd08beSQuentin Monnet * Retrieve the classid for the current task, i.e. for the net_cls 18431fdd08beSQuentin Monnet * cgroup to which *skb* belongs. 18441fdd08beSQuentin Monnet * 18451fdd08beSQuentin Monnet * This helper can be used on TC egress path, but not on ingress. 18461fdd08beSQuentin Monnet * 18471fdd08beSQuentin Monnet * The net_cls cgroup provides an interface to tag network packets 18481fdd08beSQuentin Monnet * based on a user-provided identifier for all traffic coming from 18491fdd08beSQuentin Monnet * the tasks belonging to the related cgroup. See also the related 18501fdd08beSQuentin Monnet * kernel documentation, available from the Linux sources in file 1851da82c92fSMauro Carvalho Chehab * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 18521fdd08beSQuentin Monnet * 18531fdd08beSQuentin Monnet * The Linux kernel has two versions for cgroups: there are 18541fdd08beSQuentin Monnet * cgroups v1 and cgroups v2. Both are available to users, who can 18551fdd08beSQuentin Monnet * use a mixture of them, but note that the net_cls cgroup is for 18561fdd08beSQuentin Monnet * cgroup v1 only. This makes it incompatible with BPF programs 18571fdd08beSQuentin Monnet * run on cgroups, which is a cgroup-v2-only feature (a socket can 18581fdd08beSQuentin Monnet * only hold data for one version of cgroups at a time). 18591fdd08beSQuentin Monnet * 18601fdd08beSQuentin Monnet * This helper is only available is the kernel was compiled with 18611fdd08beSQuentin Monnet * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 18621fdd08beSQuentin Monnet * "**y**" or to "**m**". 18631fdd08beSQuentin Monnet * Return 18641fdd08beSQuentin Monnet * The classid, or 0 for the default unconfigured classid. 18651fdd08beSQuentin Monnet * 1866bdb7b79bSAndrii Nakryiko * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1867c456dec4SQuentin Monnet * Description 1868c456dec4SQuentin Monnet * Push a *vlan_tci* (VLAN tag control information) of protocol 1869c456dec4SQuentin Monnet * *vlan_proto* to the packet associated to *skb*, then update 1870c456dec4SQuentin Monnet * the checksum. Note that if *vlan_proto* is different from 1871c456dec4SQuentin Monnet * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1872c456dec4SQuentin Monnet * be **ETH_P_8021Q**. 1873c456dec4SQuentin Monnet * 187432e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1875c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1876c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 1877c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 1878c456dec4SQuentin Monnet * direct packet access. 1879c456dec4SQuentin Monnet * Return 1880c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1881c456dec4SQuentin Monnet * 1882bdb7b79bSAndrii Nakryiko * long bpf_skb_vlan_pop(struct sk_buff *skb) 1883c456dec4SQuentin Monnet * Description 1884c456dec4SQuentin Monnet * Pop a VLAN header from the packet associated to *skb*. 1885c456dec4SQuentin Monnet * 188632e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1887c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1888c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 1889c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 1890c456dec4SQuentin Monnet * direct packet access. 1891c456dec4SQuentin Monnet * Return 1892c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1893c456dec4SQuentin Monnet * 1894bdb7b79bSAndrii Nakryiko * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1895c456dec4SQuentin Monnet * Description 1896c456dec4SQuentin Monnet * Get tunnel metadata. This helper takes a pointer *key* to an 1897c456dec4SQuentin Monnet * empty **struct bpf_tunnel_key** of **size**, that will be 1898c456dec4SQuentin Monnet * filled with tunnel metadata for the packet associated to *skb*. 1899c456dec4SQuentin Monnet * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1900c456dec4SQuentin Monnet * indicates that the tunnel is based on IPv6 protocol instead of 1901c456dec4SQuentin Monnet * IPv4. 1902c456dec4SQuentin Monnet * 1903c456dec4SQuentin Monnet * The **struct bpf_tunnel_key** is an object that generalizes the 1904c456dec4SQuentin Monnet * principal parameters used by various tunneling protocols into a 1905c456dec4SQuentin Monnet * single struct. This way, it can be used to easily make a 1906c456dec4SQuentin Monnet * decision based on the contents of the encapsulation header, 1907c456dec4SQuentin Monnet * "summarized" in this struct. In particular, it holds the IP 1908c456dec4SQuentin Monnet * address of the remote end (IPv4 or IPv6, depending on the case) 1909c456dec4SQuentin Monnet * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1910c456dec4SQuentin Monnet * this struct exposes the *key*\ **->tunnel_id**, which is 1911c456dec4SQuentin Monnet * generally mapped to a VNI (Virtual Network Identifier), making 1912c456dec4SQuentin Monnet * it programmable together with the **bpf_skb_set_tunnel_key**\ 1913c456dec4SQuentin Monnet * () helper. 1914c456dec4SQuentin Monnet * 1915c456dec4SQuentin Monnet * Let's imagine that the following code is part of a program 1916c456dec4SQuentin Monnet * attached to the TC ingress interface, on one end of a GRE 1917c456dec4SQuentin Monnet * tunnel, and is supposed to filter out all messages coming from 1918c456dec4SQuentin Monnet * remote ends with IPv4 address other than 10.0.0.1: 1919c456dec4SQuentin Monnet * 1920c456dec4SQuentin Monnet * :: 1921c456dec4SQuentin Monnet * 1922c456dec4SQuentin Monnet * int ret; 1923c456dec4SQuentin Monnet * struct bpf_tunnel_key key = {}; 1924c456dec4SQuentin Monnet * 1925c456dec4SQuentin Monnet * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1926c456dec4SQuentin Monnet * if (ret < 0) 1927c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 1928c456dec4SQuentin Monnet * 1929c456dec4SQuentin Monnet * if (key.remote_ipv4 != 0x0a000001) 1930c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 1931c456dec4SQuentin Monnet * 1932c456dec4SQuentin Monnet * return TC_ACT_OK; // accept packet 1933c456dec4SQuentin Monnet * 1934c456dec4SQuentin Monnet * This interface can also be used with all encapsulation devices 1935c456dec4SQuentin Monnet * that can operate in "collect metadata" mode: instead of having 1936c456dec4SQuentin Monnet * one network device per specific configuration, the "collect 1937c456dec4SQuentin Monnet * metadata" mode only requires a single device where the 1938c456dec4SQuentin Monnet * configuration can be extracted from this helper. 1939c456dec4SQuentin Monnet * 1940c456dec4SQuentin Monnet * This can be used together with various tunnels such as VXLan, 1941c456dec4SQuentin Monnet * Geneve, GRE or IP in IP (IPIP). 1942c456dec4SQuentin Monnet * Return 1943c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1944c456dec4SQuentin Monnet * 1945bdb7b79bSAndrii Nakryiko * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1946c456dec4SQuentin Monnet * Description 1947c456dec4SQuentin Monnet * Populate tunnel metadata for packet associated to *skb.* The 1948c456dec4SQuentin Monnet * tunnel metadata is set to the contents of *key*, of *size*. The 1949c456dec4SQuentin Monnet * *flags* can be set to a combination of the following values: 1950c456dec4SQuentin Monnet * 1951c456dec4SQuentin Monnet * **BPF_F_TUNINFO_IPV6** 1952c456dec4SQuentin Monnet * Indicate that the tunnel is based on IPv6 protocol 1953c456dec4SQuentin Monnet * instead of IPv4. 1954c456dec4SQuentin Monnet * **BPF_F_ZERO_CSUM_TX** 1955c456dec4SQuentin Monnet * For IPv4 packets, add a flag to tunnel metadata 1956c456dec4SQuentin Monnet * indicating that checksum computation should be skipped 1957c456dec4SQuentin Monnet * and checksum set to zeroes. 1958c456dec4SQuentin Monnet * **BPF_F_DONT_FRAGMENT** 1959c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that the 1960c456dec4SQuentin Monnet * packet should not be fragmented. 1961c456dec4SQuentin Monnet * **BPF_F_SEQ_NUMBER** 1962c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that a 1963c456dec4SQuentin Monnet * sequence number should be added to tunnel header before 1964c456dec4SQuentin Monnet * sending the packet. This flag was added for GRE 1965c456dec4SQuentin Monnet * encapsulation, but might be used with other protocols 1966c456dec4SQuentin Monnet * as well in the future. 1967c456dec4SQuentin Monnet * 1968c456dec4SQuentin Monnet * Here is a typical usage on the transmit path: 1969c456dec4SQuentin Monnet * 1970c456dec4SQuentin Monnet * :: 1971c456dec4SQuentin Monnet * 1972c456dec4SQuentin Monnet * struct bpf_tunnel_key key; 1973c456dec4SQuentin Monnet * populate key ... 1974c456dec4SQuentin Monnet * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1975c456dec4SQuentin Monnet * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1976c456dec4SQuentin Monnet * 1977c456dec4SQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_key**\ () 1978c456dec4SQuentin Monnet * helper for additional information. 1979c456dec4SQuentin Monnet * Return 1980c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1981c456dec4SQuentin Monnet * 1982c6b5fb86SQuentin Monnet * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1983c6b5fb86SQuentin Monnet * Description 1984c6b5fb86SQuentin Monnet * Read the value of a perf event counter. This helper relies on a 1985c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1986c6b5fb86SQuentin Monnet * the perf event counter is selected when *map* is updated with 1987c6b5fb86SQuentin Monnet * perf event file descriptors. The *map* is an array whose size 1988c6b5fb86SQuentin Monnet * is the number of available CPUs, and each cell contains a value 1989c6b5fb86SQuentin Monnet * relative to one CPU. The value to retrieve is indicated by 1990c6b5fb86SQuentin Monnet * *flags*, that contains the index of the CPU to look up, masked 1991c6b5fb86SQuentin Monnet * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1992c6b5fb86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 1993c6b5fb86SQuentin Monnet * current CPU should be retrieved. 1994c6b5fb86SQuentin Monnet * 1995c6b5fb86SQuentin Monnet * Note that before Linux 4.13, only hardware perf event can be 1996c6b5fb86SQuentin Monnet * retrieved. 1997c6b5fb86SQuentin Monnet * 1998c6b5fb86SQuentin Monnet * Also, be aware that the newer helper 1999c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () is recommended over 20003bd5a09bSQuentin Monnet * **bpf_perf_event_read**\ () in general. The latter has some ABI 2001c6b5fb86SQuentin Monnet * quirks where error and counter value are used as a return code 2002c6b5fb86SQuentin Monnet * (which is wrong to do since ranges may overlap). This issue is 20033bd5a09bSQuentin Monnet * fixed with **bpf_perf_event_read_value**\ (), which at the same 20043bd5a09bSQuentin Monnet * time provides more features over the **bpf_perf_event_read**\ 20053bd5a09bSQuentin Monnet * () interface. Please refer to the description of 2006c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () for details. 2007c6b5fb86SQuentin Monnet * Return 2008c6b5fb86SQuentin Monnet * The value of the perf event counter read from the map, or a 2009c6b5fb86SQuentin Monnet * negative error code in case of failure. 2010c6b5fb86SQuentin Monnet * 2011bdb7b79bSAndrii Nakryiko * long bpf_redirect(u32 ifindex, u64 flags) 2012c456dec4SQuentin Monnet * Description 2013c456dec4SQuentin Monnet * Redirect the packet to another net device of index *ifindex*. 2014c456dec4SQuentin Monnet * This helper is somewhat similar to **bpf_clone_redirect**\ 2015c456dec4SQuentin Monnet * (), except that the packet is not cloned, which provides 2016c456dec4SQuentin Monnet * increased performance. 2017c456dec4SQuentin Monnet * 2018c456dec4SQuentin Monnet * Except for XDP, both ingress and egress interfaces can be used 2019c456dec4SQuentin Monnet * for redirection. The **BPF_F_INGRESS** value in *flags* is used 2020c456dec4SQuentin Monnet * to make the distinction (ingress path is selected if the flag 2021c456dec4SQuentin Monnet * is present, egress path otherwise). Currently, XDP only 2022c456dec4SQuentin Monnet * supports redirection to the egress interface, and accepts no 2023c456dec4SQuentin Monnet * flag at all. 2024c456dec4SQuentin Monnet * 2025f25975f4SToke Høiland-Jørgensen * The same effect can also be attained with the more generic 2026f25975f4SToke Høiland-Jørgensen * **bpf_redirect_map**\ (), which uses a BPF map to store the 2027f25975f4SToke Høiland-Jørgensen * redirect target instead of providing it directly to the helper. 2028c456dec4SQuentin Monnet * Return 2029c456dec4SQuentin Monnet * For XDP, the helper returns **XDP_REDIRECT** on success or 2030c456dec4SQuentin Monnet * **XDP_ABORTED** on error. For other program types, the values 2031c456dec4SQuentin Monnet * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2032c456dec4SQuentin Monnet * error. 2033c456dec4SQuentin Monnet * 20341fdd08beSQuentin Monnet * u32 bpf_get_route_realm(struct sk_buff *skb) 20351fdd08beSQuentin Monnet * Description 20361fdd08beSQuentin Monnet * Retrieve the realm or the route, that is to say the 20371fdd08beSQuentin Monnet * **tclassid** field of the destination for the *skb*. The 2038b16fc097STobias Klauser * identifier retrieved is a user-provided tag, similar to the 20391fdd08beSQuentin Monnet * one used with the net_cls cgroup (see description for 20401fdd08beSQuentin Monnet * **bpf_get_cgroup_classid**\ () helper), but here this tag is 20411fdd08beSQuentin Monnet * held by a route (a destination entry), not by a task. 20421fdd08beSQuentin Monnet * 20431fdd08beSQuentin Monnet * Retrieving this identifier works with the clsact TC egress hook 20441fdd08beSQuentin Monnet * (see also **tc-bpf(8)**), or alternatively on conventional 20451fdd08beSQuentin Monnet * classful egress qdiscs, but not on TC ingress path. In case of 20461fdd08beSQuentin Monnet * clsact TC egress hook, this has the advantage that, internally, 20471fdd08beSQuentin Monnet * the destination entry has not been dropped yet in the transmit 20481fdd08beSQuentin Monnet * path. Therefore, the destination entry does not need to be 20491fdd08beSQuentin Monnet * artificially held via **netif_keep_dst**\ () for a classful 20501fdd08beSQuentin Monnet * qdisc until the *skb* is freed. 20511fdd08beSQuentin Monnet * 20521fdd08beSQuentin Monnet * This helper is available only if the kernel was compiled with 20531fdd08beSQuentin Monnet * **CONFIG_IP_ROUTE_CLASSID** configuration option. 20541fdd08beSQuentin Monnet * Return 20551fdd08beSQuentin Monnet * The realm of the route for the packet associated to *skb*, or 0 20561fdd08beSQuentin Monnet * if none was found. 20571fdd08beSQuentin Monnet * 2058bdb7b79bSAndrii Nakryiko * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2059c456dec4SQuentin Monnet * Description 2060c456dec4SQuentin Monnet * Write raw *data* blob into a special BPF perf event held by 2061c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2062c456dec4SQuentin Monnet * event must have the following attributes: **PERF_SAMPLE_RAW** 2063c456dec4SQuentin Monnet * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2064c456dec4SQuentin Monnet * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2065c456dec4SQuentin Monnet * 2066c456dec4SQuentin Monnet * The *flags* are used to indicate the index in *map* for which 2067c456dec4SQuentin Monnet * the value must be put, masked with **BPF_F_INDEX_MASK**. 2068c456dec4SQuentin Monnet * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2069c456dec4SQuentin Monnet * to indicate that the index of the current CPU core should be 2070c456dec4SQuentin Monnet * used. 2071c456dec4SQuentin Monnet * 2072c456dec4SQuentin Monnet * The value to write, of *size*, is passed through eBPF stack and 2073c456dec4SQuentin Monnet * pointed by *data*. 2074c456dec4SQuentin Monnet * 2075c456dec4SQuentin Monnet * The context of the program *ctx* needs also be passed to the 2076c456dec4SQuentin Monnet * helper. 2077c456dec4SQuentin Monnet * 2078c456dec4SQuentin Monnet * On user space, a program willing to read the values needs to 2079c456dec4SQuentin Monnet * call **perf_event_open**\ () on the perf event (either for 2080c456dec4SQuentin Monnet * one or for all CPUs) and to store the file descriptor into the 2081c456dec4SQuentin Monnet * *map*. This must be done before the eBPF program can send data 2082c456dec4SQuentin Monnet * into it. An example is available in file 2083c456dec4SQuentin Monnet * *samples/bpf/trace_output_user.c* in the Linux kernel source 2084c456dec4SQuentin Monnet * tree (the eBPF program counterpart is in 2085c456dec4SQuentin Monnet * *samples/bpf/trace_output_kern.c*). 2086c456dec4SQuentin Monnet * 2087c456dec4SQuentin Monnet * **bpf_perf_event_output**\ () achieves better performance 2088c456dec4SQuentin Monnet * than **bpf_trace_printk**\ () for sharing data with user 2089c456dec4SQuentin Monnet * space, and is much better suitable for streaming data from eBPF 2090c456dec4SQuentin Monnet * programs. 2091c456dec4SQuentin Monnet * 2092c456dec4SQuentin Monnet * Note that this helper is not restricted to tracing use cases 2093c456dec4SQuentin Monnet * and can be used with programs attached to TC or XDP as well, 2094c456dec4SQuentin Monnet * where it allows for passing data to user space listeners. Data 2095c456dec4SQuentin Monnet * can be: 2096c456dec4SQuentin Monnet * 2097c456dec4SQuentin Monnet * * Only custom structs, 2098c456dec4SQuentin Monnet * * Only the packet payload, or 2099c456dec4SQuentin Monnet * * A combination of both. 2100c456dec4SQuentin Monnet * Return 2101c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 2102c456dec4SQuentin Monnet * 2103bdb7b79bSAndrii Nakryiko * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 21041fdd08beSQuentin Monnet * Description 21051fdd08beSQuentin Monnet * This helper was provided as an easy way to load data from a 21061fdd08beSQuentin Monnet * packet. It can be used to load *len* bytes from *offset* from 21071fdd08beSQuentin Monnet * the packet associated to *skb*, into the buffer pointed by 21081fdd08beSQuentin Monnet * *to*. 21091fdd08beSQuentin Monnet * 21101fdd08beSQuentin Monnet * Since Linux 4.7, usage of this helper has mostly been replaced 21111fdd08beSQuentin Monnet * by "direct packet access", enabling packet data to be 21121fdd08beSQuentin Monnet * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 21131fdd08beSQuentin Monnet * pointing respectively to the first byte of packet data and to 21141fdd08beSQuentin Monnet * the byte after the last byte of packet data. However, it 21151fdd08beSQuentin Monnet * remains useful if one wishes to read large quantities of data 21161fdd08beSQuentin Monnet * at once from a packet into the eBPF stack. 21171fdd08beSQuentin Monnet * Return 21181fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 21191fdd08beSQuentin Monnet * 2120bdb7b79bSAndrii Nakryiko * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2121c456dec4SQuentin Monnet * Description 2122c456dec4SQuentin Monnet * Walk a user or a kernel stack and return its id. To achieve 2123c456dec4SQuentin Monnet * this, the helper needs *ctx*, which is a pointer to the context 2124c456dec4SQuentin Monnet * on which the tracing program is executed, and a pointer to a 2125c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2126c456dec4SQuentin Monnet * 2127c456dec4SQuentin Monnet * The last argument, *flags*, holds the number of stack frames to 2128c456dec4SQuentin Monnet * skip (from 0 to 255), masked with 2129c456dec4SQuentin Monnet * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2130c456dec4SQuentin Monnet * a combination of the following flags: 2131c456dec4SQuentin Monnet * 2132c456dec4SQuentin Monnet * **BPF_F_USER_STACK** 2133c456dec4SQuentin Monnet * Collect a user space stack instead of a kernel stack. 2134c456dec4SQuentin Monnet * **BPF_F_FAST_STACK_CMP** 2135c456dec4SQuentin Monnet * Compare stacks by hash only. 2136c456dec4SQuentin Monnet * **BPF_F_REUSE_STACKID** 2137c456dec4SQuentin Monnet * If two different stacks hash into the same *stackid*, 2138c456dec4SQuentin Monnet * discard the old one. 2139c456dec4SQuentin Monnet * 2140c456dec4SQuentin Monnet * The stack id retrieved is a 32 bit long integer handle which 2141c456dec4SQuentin Monnet * can be further combined with other data (including other stack 2142c456dec4SQuentin Monnet * ids) and used as a key into maps. This can be useful for 2143c456dec4SQuentin Monnet * generating a variety of graphs (such as flame graphs or off-cpu 2144c456dec4SQuentin Monnet * graphs). 2145c456dec4SQuentin Monnet * 2146c456dec4SQuentin Monnet * For walking a stack, this helper is an improvement over 2147c456dec4SQuentin Monnet * **bpf_probe_read**\ (), which can be used with unrolled loops 2148c456dec4SQuentin Monnet * but is not efficient and consumes a lot of eBPF instructions. 2149c456dec4SQuentin Monnet * Instead, **bpf_get_stackid**\ () can collect up to 2150c456dec4SQuentin Monnet * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2151c456dec4SQuentin Monnet * this limit can be controlled with the **sysctl** program, and 2152c456dec4SQuentin Monnet * that it should be manually increased in order to profile long 2153c456dec4SQuentin Monnet * user stacks (such as stacks for Java programs). To do so, use: 2154c456dec4SQuentin Monnet * 2155c456dec4SQuentin Monnet * :: 2156c456dec4SQuentin Monnet * 2157c456dec4SQuentin Monnet * # sysctl kernel.perf_event_max_stack=<new value> 2158c456dec4SQuentin Monnet * Return 2159c456dec4SQuentin Monnet * The positive or null stack id on success, or a negative error 2160c456dec4SQuentin Monnet * in case of failure. 2161c456dec4SQuentin Monnet * 21621fdd08beSQuentin Monnet * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 21631fdd08beSQuentin Monnet * Description 21641fdd08beSQuentin Monnet * Compute a checksum difference, from the raw buffer pointed by 21651fdd08beSQuentin Monnet * *from*, of length *from_size* (that must be a multiple of 4), 21661fdd08beSQuentin Monnet * towards the raw buffer pointed by *to*, of size *to_size* 21671fdd08beSQuentin Monnet * (same remark). An optional *seed* can be added to the value 21681fdd08beSQuentin Monnet * (this can be cascaded, the seed may come from a previous call 21691fdd08beSQuentin Monnet * to the helper). 21701fdd08beSQuentin Monnet * 21711fdd08beSQuentin Monnet * This is flexible enough to be used in several ways: 21721fdd08beSQuentin Monnet * 21731fdd08beSQuentin Monnet * * With *from_size* == 0, *to_size* > 0 and *seed* set to 21741fdd08beSQuentin Monnet * checksum, it can be used when pushing new data. 21751fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* == 0 and *seed* set to 21761fdd08beSQuentin Monnet * checksum, it can be used when removing data from a packet. 21771fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 21781fdd08beSQuentin Monnet * can be used to compute a diff. Note that *from_size* and 21791fdd08beSQuentin Monnet * *to_size* do not need to be equal. 21801fdd08beSQuentin Monnet * 21811fdd08beSQuentin Monnet * This helper can be used in combination with 21821fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 21831fdd08beSQuentin Monnet * which one can feed in the difference computed with 21841fdd08beSQuentin Monnet * **bpf_csum_diff**\ (). 21851fdd08beSQuentin Monnet * Return 21861fdd08beSQuentin Monnet * The checksum result, or a negative error code in case of 21871fdd08beSQuentin Monnet * failure. 21881fdd08beSQuentin Monnet * 2189bdb7b79bSAndrii Nakryiko * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 21901fdd08beSQuentin Monnet * Description 21911fdd08beSQuentin Monnet * Retrieve tunnel options metadata for the packet associated to 21921fdd08beSQuentin Monnet * *skb*, and store the raw tunnel option data to the buffer *opt* 21931fdd08beSQuentin Monnet * of *size*. 21941fdd08beSQuentin Monnet * 21951fdd08beSQuentin Monnet * This helper can be used with encapsulation devices that can 21961fdd08beSQuentin Monnet * operate in "collect metadata" mode (please refer to the related 21971fdd08beSQuentin Monnet * note in the description of **bpf_skb_get_tunnel_key**\ () for 21981fdd08beSQuentin Monnet * more details). A particular example where this can be used is 21991fdd08beSQuentin Monnet * in combination with the Geneve encapsulation protocol, where it 22001fdd08beSQuentin Monnet * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 22011fdd08beSQuentin Monnet * and retrieving arbitrary TLVs (Type-Length-Value headers) from 22021fdd08beSQuentin Monnet * the eBPF program. This allows for full customization of these 22031fdd08beSQuentin Monnet * headers. 22041fdd08beSQuentin Monnet * Return 22051fdd08beSQuentin Monnet * The size of the option data retrieved. 22061fdd08beSQuentin Monnet * 2207bdb7b79bSAndrii Nakryiko * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 22081fdd08beSQuentin Monnet * Description 22091fdd08beSQuentin Monnet * Set tunnel options metadata for the packet associated to *skb* 22101fdd08beSQuentin Monnet * to the option data contained in the raw buffer *opt* of *size*. 22111fdd08beSQuentin Monnet * 22121fdd08beSQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_opt**\ () 22131fdd08beSQuentin Monnet * helper for additional information. 22141fdd08beSQuentin Monnet * Return 22151fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 22161fdd08beSQuentin Monnet * 2217bdb7b79bSAndrii Nakryiko * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 22181fdd08beSQuentin Monnet * Description 22191fdd08beSQuentin Monnet * Change the protocol of the *skb* to *proto*. Currently 22201fdd08beSQuentin Monnet * supported are transition from IPv4 to IPv6, and from IPv6 to 22211fdd08beSQuentin Monnet * IPv4. The helper takes care of the groundwork for the 22221fdd08beSQuentin Monnet * transition, including resizing the socket buffer. The eBPF 22231fdd08beSQuentin Monnet * program is expected to fill the new headers, if any, via 22241fdd08beSQuentin Monnet * **skb_store_bytes**\ () and to recompute the checksums with 22251fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 22261fdd08beSQuentin Monnet * (). The main case for this helper is to perform NAT64 22271fdd08beSQuentin Monnet * operations out of an eBPF program. 22281fdd08beSQuentin Monnet * 22291fdd08beSQuentin Monnet * Internally, the GSO type is marked as dodgy so that headers are 22301fdd08beSQuentin Monnet * checked and segments are recalculated by the GSO/GRO engine. 22311fdd08beSQuentin Monnet * The size for GSO target is adapted as well. 22321fdd08beSQuentin Monnet * 22331fdd08beSQuentin Monnet * All values for *flags* are reserved for future usage, and must 22341fdd08beSQuentin Monnet * be left at zero. 22351fdd08beSQuentin Monnet * 223632e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 22371fdd08beSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 22381fdd08beSQuentin Monnet * previously done by the verifier are invalidated and must be 22391fdd08beSQuentin Monnet * performed again, if the helper is used in combination with 22401fdd08beSQuentin Monnet * direct packet access. 22411fdd08beSQuentin Monnet * Return 22421fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 22431fdd08beSQuentin Monnet * 2244bdb7b79bSAndrii Nakryiko * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 22451fdd08beSQuentin Monnet * Description 22461fdd08beSQuentin Monnet * Change the packet type for the packet associated to *skb*. This 22471fdd08beSQuentin Monnet * comes down to setting *skb*\ **->pkt_type** to *type*, except 22481fdd08beSQuentin Monnet * the eBPF program does not have a write access to *skb*\ 22491fdd08beSQuentin Monnet * **->pkt_type** beside this helper. Using a helper here allows 22501fdd08beSQuentin Monnet * for graceful handling of errors. 22511fdd08beSQuentin Monnet * 22521fdd08beSQuentin Monnet * The major use case is to change incoming *skb*s to 22531fdd08beSQuentin Monnet * **PACKET_HOST** in a programmatic way instead of having to 22541fdd08beSQuentin Monnet * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 22551fdd08beSQuentin Monnet * example. 22561fdd08beSQuentin Monnet * 22571fdd08beSQuentin Monnet * Note that *type* only allows certain values. At this time, they 22581fdd08beSQuentin Monnet * are: 22591fdd08beSQuentin Monnet * 22601fdd08beSQuentin Monnet * **PACKET_HOST** 22611fdd08beSQuentin Monnet * Packet is for us. 22621fdd08beSQuentin Monnet * **PACKET_BROADCAST** 22631fdd08beSQuentin Monnet * Send packet to all. 22641fdd08beSQuentin Monnet * **PACKET_MULTICAST** 22651fdd08beSQuentin Monnet * Send packet to group. 22661fdd08beSQuentin Monnet * **PACKET_OTHERHOST** 22671fdd08beSQuentin Monnet * Send packet to someone else. 22681fdd08beSQuentin Monnet * Return 22691fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 22701fdd08beSQuentin Monnet * 2271bdb7b79bSAndrii Nakryiko * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2272c6b5fb86SQuentin Monnet * Description 2273c6b5fb86SQuentin Monnet * Check whether *skb* is a descendant of the cgroup2 held by 2274c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2275c6b5fb86SQuentin Monnet * Return 2276c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 2277c6b5fb86SQuentin Monnet * 2278c6b5fb86SQuentin Monnet * * 0, if the *skb* failed the cgroup2 descendant test. 2279c6b5fb86SQuentin Monnet * * 1, if the *skb* succeeded the cgroup2 descendant test. 2280c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 2281c6b5fb86SQuentin Monnet * 2282fa15601aSQuentin Monnet * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2283fa15601aSQuentin Monnet * Description 2284fa15601aSQuentin Monnet * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2285fa15601aSQuentin Monnet * not set, in particular if the hash was cleared due to mangling, 2286fa15601aSQuentin Monnet * recompute this hash. Later accesses to the hash can be done 2287fa15601aSQuentin Monnet * directly with *skb*\ **->hash**. 2288fa15601aSQuentin Monnet * 2289fa15601aSQuentin Monnet * Calling **bpf_set_hash_invalid**\ (), changing a packet 2290fa15601aSQuentin Monnet * prototype with **bpf_skb_change_proto**\ (), or calling 2291fa15601aSQuentin Monnet * **bpf_skb_store_bytes**\ () with the 2292fa15601aSQuentin Monnet * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2293fa15601aSQuentin Monnet * the hash and to trigger a new computation for the next call to 2294fa15601aSQuentin Monnet * **bpf_get_hash_recalc**\ (). 2295fa15601aSQuentin Monnet * Return 2296fa15601aSQuentin Monnet * The 32-bit hash. 2297fa15601aSQuentin Monnet * 2298c456dec4SQuentin Monnet * u64 bpf_get_current_task(void) 2299e40fbbf0SUsama Arif * Description 2300e40fbbf0SUsama Arif * Get the current task. 2301c456dec4SQuentin Monnet * Return 2302c456dec4SQuentin Monnet * A pointer to the current task struct. 2303fa15601aSQuentin Monnet * 2304bdb7b79bSAndrii Nakryiko * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2305c6b5fb86SQuentin Monnet * Description 2306c6b5fb86SQuentin Monnet * Attempt in a safe way to write *len* bytes from the buffer 2307c6b5fb86SQuentin Monnet * *src* to *dst* in memory. It only works for threads that are in 2308c6b5fb86SQuentin Monnet * user context, and *dst* must be a valid user space address. 2309c6b5fb86SQuentin Monnet * 2310c6b5fb86SQuentin Monnet * This helper should not be used to implement any kind of 2311c6b5fb86SQuentin Monnet * security mechanism because of TOC-TOU attacks, but rather to 2312c6b5fb86SQuentin Monnet * debug, divert, and manipulate execution of semi-cooperative 2313c6b5fb86SQuentin Monnet * processes. 2314c6b5fb86SQuentin Monnet * 2315c6b5fb86SQuentin Monnet * Keep in mind that this feature is meant for experiments, and it 2316c6b5fb86SQuentin Monnet * has a risk of crashing the system and running programs. 2317c6b5fb86SQuentin Monnet * Therefore, when an eBPF program using this helper is attached, 2318c6b5fb86SQuentin Monnet * a warning including PID and process name is printed to kernel 2319c6b5fb86SQuentin Monnet * logs. 2320c6b5fb86SQuentin Monnet * Return 2321c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2322c6b5fb86SQuentin Monnet * 2323bdb7b79bSAndrii Nakryiko * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2324c6b5fb86SQuentin Monnet * Description 2325c6b5fb86SQuentin Monnet * Check whether the probe is being run is the context of a given 2326c6b5fb86SQuentin Monnet * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2327c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2328c6b5fb86SQuentin Monnet * Return 2329c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 2330c6b5fb86SQuentin Monnet * 233158617014SHengqi Chen * * 1, if current task belongs to the cgroup2. 233258617014SHengqi Chen * * 0, if current task does not belong to the cgroup2. 2333c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 2334c6b5fb86SQuentin Monnet * 2335bdb7b79bSAndrii Nakryiko * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2336fa15601aSQuentin Monnet * Description 2337fa15601aSQuentin Monnet * Resize (trim or grow) the packet associated to *skb* to the 2338fa15601aSQuentin Monnet * new *len*. The *flags* are reserved for future usage, and must 2339fa15601aSQuentin Monnet * be left at zero. 2340fa15601aSQuentin Monnet * 2341fa15601aSQuentin Monnet * The basic idea is that the helper performs the needed work to 2342fa15601aSQuentin Monnet * change the size of the packet, then the eBPF program rewrites 2343fa15601aSQuentin Monnet * the rest via helpers like **bpf_skb_store_bytes**\ (), 2344fa15601aSQuentin Monnet * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2345fa15601aSQuentin Monnet * and others. This helper is a slow path utility intended for 2346fa15601aSQuentin Monnet * replies with control messages. And because it is targeted for 2347fa15601aSQuentin Monnet * slow path, the helper itself can afford to be slow: it 2348fa15601aSQuentin Monnet * implicitly linearizes, unclones and drops offloads from the 2349fa15601aSQuentin Monnet * *skb*. 2350fa15601aSQuentin Monnet * 235132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2352fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2353fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2354fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2355fa15601aSQuentin Monnet * direct packet access. 2356fa15601aSQuentin Monnet * Return 2357fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2358fa15601aSQuentin Monnet * 2359bdb7b79bSAndrii Nakryiko * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2360fa15601aSQuentin Monnet * Description 2361fa15601aSQuentin Monnet * Pull in non-linear data in case the *skb* is non-linear and not 2362fa15601aSQuentin Monnet * all of *len* are part of the linear section. Make *len* bytes 2363fa15601aSQuentin Monnet * from *skb* readable and writable. If a zero value is passed for 2364bdb2bc75SJoanne Koong * *len*, then all bytes in the linear part of *skb* will be made 2365bdb2bc75SJoanne Koong * readable and writable. 2366fa15601aSQuentin Monnet * 2367fa15601aSQuentin Monnet * This helper is only needed for reading and writing with direct 2368fa15601aSQuentin Monnet * packet access. 2369fa15601aSQuentin Monnet * 2370fa15601aSQuentin Monnet * For direct packet access, testing that offsets to access 2371fa15601aSQuentin Monnet * are within packet boundaries (test on *skb*\ **->data_end**) is 2372fa15601aSQuentin Monnet * susceptible to fail if offsets are invalid, or if the requested 2373fa15601aSQuentin Monnet * data is in non-linear parts of the *skb*. On failure the 2374fa15601aSQuentin Monnet * program can just bail out, or in the case of a non-linear 2375fa15601aSQuentin Monnet * buffer, use a helper to make the data available. The 2376fa15601aSQuentin Monnet * **bpf_skb_load_bytes**\ () helper is a first solution to access 2377fa15601aSQuentin Monnet * the data. Another one consists in using **bpf_skb_pull_data** 2378fa15601aSQuentin Monnet * to pull in once the non-linear parts, then retesting and 2379fa15601aSQuentin Monnet * eventually access the data. 2380fa15601aSQuentin Monnet * 2381fa15601aSQuentin Monnet * At the same time, this also makes sure the *skb* is uncloned, 2382fa15601aSQuentin Monnet * which is a necessary condition for direct write. As this needs 2383fa15601aSQuentin Monnet * to be an invariant for the write part only, the verifier 2384fa15601aSQuentin Monnet * detects writes and adds a prologue that is calling 2385fa15601aSQuentin Monnet * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2386fa15601aSQuentin Monnet * the very beginning in case it is indeed cloned. 2387fa15601aSQuentin Monnet * 238832e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2389fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2390fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2391fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2392fa15601aSQuentin Monnet * direct packet access. 2393fa15601aSQuentin Monnet * Return 2394fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2395fa15601aSQuentin Monnet * 2396fa15601aSQuentin Monnet * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2397fa15601aSQuentin Monnet * Description 2398fa15601aSQuentin Monnet * Add the checksum *csum* into *skb*\ **->csum** in case the 2399fa15601aSQuentin Monnet * driver has supplied a checksum for the entire packet into that 2400fa15601aSQuentin Monnet * field. Return an error otherwise. This helper is intended to be 2401fa15601aSQuentin Monnet * used in combination with **bpf_csum_diff**\ (), in particular 2402fa15601aSQuentin Monnet * when the checksum needs to be updated after data has been 2403fa15601aSQuentin Monnet * written into the packet through direct packet access. 2404fa15601aSQuentin Monnet * Return 2405fa15601aSQuentin Monnet * The checksum on success, or a negative error code in case of 2406fa15601aSQuentin Monnet * failure. 2407fa15601aSQuentin Monnet * 2408fa15601aSQuentin Monnet * void bpf_set_hash_invalid(struct sk_buff *skb) 2409fa15601aSQuentin Monnet * Description 2410fa15601aSQuentin Monnet * Invalidate the current *skb*\ **->hash**. It can be used after 2411fa15601aSQuentin Monnet * mangling on headers through direct packet access, in order to 2412fa15601aSQuentin Monnet * indicate that the hash is outdated and to trigger a 2413fa15601aSQuentin Monnet * recalculation the next time the kernel tries to access this 2414fa15601aSQuentin Monnet * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2415e40fbbf0SUsama Arif * Return 2416e40fbbf0SUsama Arif * void. 2417fa15601aSQuentin Monnet * 2418bdb7b79bSAndrii Nakryiko * long bpf_get_numa_node_id(void) 2419fa15601aSQuentin Monnet * Description 2420fa15601aSQuentin Monnet * Return the id of the current NUMA node. The primary use case 2421fa15601aSQuentin Monnet * for this helper is the selection of sockets for the local NUMA 2422fa15601aSQuentin Monnet * node, when the program is attached to sockets using the 2423fa15601aSQuentin Monnet * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2424fa15601aSQuentin Monnet * but the helper is also available to other eBPF program types, 2425fa15601aSQuentin Monnet * similarly to **bpf_get_smp_processor_id**\ (). 2426fa15601aSQuentin Monnet * Return 2427fa15601aSQuentin Monnet * The id of current NUMA node. 2428fa15601aSQuentin Monnet * 2429bdb7b79bSAndrii Nakryiko * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2430c6b5fb86SQuentin Monnet * Description 2431c6b5fb86SQuentin Monnet * Grows headroom of packet associated to *skb* and adjusts the 2432c6b5fb86SQuentin Monnet * offset of the MAC header accordingly, adding *len* bytes of 2433c6b5fb86SQuentin Monnet * space. It automatically extends and reallocates memory as 2434c6b5fb86SQuentin Monnet * required. 2435c6b5fb86SQuentin Monnet * 2436c6b5fb86SQuentin Monnet * This helper can be used on a layer 3 *skb* to push a MAC header 2437c6b5fb86SQuentin Monnet * for redirection into a layer 2 device. 2438c6b5fb86SQuentin Monnet * 2439c6b5fb86SQuentin Monnet * All values for *flags* are reserved for future usage, and must 2440c6b5fb86SQuentin Monnet * be left at zero. 2441c6b5fb86SQuentin Monnet * 244232e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2443c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2444c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 2445c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 2446c6b5fb86SQuentin Monnet * direct packet access. 2447c6b5fb86SQuentin Monnet * Return 2448c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2449c6b5fb86SQuentin Monnet * 2450bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2451c6b5fb86SQuentin Monnet * Description 2452c6b5fb86SQuentin Monnet * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2453c6b5fb86SQuentin Monnet * it is possible to use a negative value for *delta*. This helper 2454c6b5fb86SQuentin Monnet * can be used to prepare the packet for pushing or popping 2455c6b5fb86SQuentin Monnet * headers. 2456c6b5fb86SQuentin Monnet * 245732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2458c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2459c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 2460c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 2461c6b5fb86SQuentin Monnet * direct packet access. 2462c6b5fb86SQuentin Monnet * Return 2463c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2464c6b5fb86SQuentin Monnet * 2465bdb7b79bSAndrii Nakryiko * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2466c6b5fb86SQuentin Monnet * Description 24676ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe kernel address 2468ab8d7809SQuentin Monnet * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 24696ae08ae3SDaniel Borkmann * more details. 2470c6b5fb86SQuentin Monnet * 2471ab8d7809SQuentin Monnet * Generally, use **bpf_probe_read_user_str**\ () or 2472ab8d7809SQuentin Monnet * **bpf_probe_read_kernel_str**\ () instead. 2473c6b5fb86SQuentin Monnet * Return 2474c6b5fb86SQuentin Monnet * On success, the strictly positive length of the string, 2475c6b5fb86SQuentin Monnet * including the trailing NUL character. On error, a negative 2476c6b5fb86SQuentin Monnet * value. 2477c6b5fb86SQuentin Monnet * 2478c6b5fb86SQuentin Monnet * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2479c6b5fb86SQuentin Monnet * Description 2480c6b5fb86SQuentin Monnet * If the **struct sk_buff** pointed by *skb* has a known socket, 2481c6b5fb86SQuentin Monnet * retrieve the cookie (generated by the kernel) of this socket. 2482c6b5fb86SQuentin Monnet * If no cookie has been set yet, generate a new cookie. Once 2483c6b5fb86SQuentin Monnet * generated, the socket cookie remains stable for the life of the 2484c6b5fb86SQuentin Monnet * socket. This helper can be useful for monitoring per socket 2485cd48bddaSDaniel Borkmann * networking traffic statistics as it provides a global socket 2486cd48bddaSDaniel Borkmann * identifier that can be assumed unique. 2487c6b5fb86SQuentin Monnet * Return 248807881ccbSFlorent Revest * A 8-byte long unique number on success, or 0 if the socket 248907881ccbSFlorent Revest * field is missing inside *skb*. 2490c6b5fb86SQuentin Monnet * 2491d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2492d692f113SAndrey Ignatov * Description 2493d692f113SAndrey Ignatov * Equivalent to bpf_get_socket_cookie() helper that accepts 249462369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_addr** context. 2495d692f113SAndrey Ignatov * Return 249607881ccbSFlorent Revest * A 8-byte long unique number. 2497d692f113SAndrey Ignatov * 2498d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2499d692f113SAndrey Ignatov * Description 2500ab8d7809SQuentin Monnet * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 250162369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_ops** context. 2502d692f113SAndrey Ignatov * Return 250307881ccbSFlorent Revest * A 8-byte long unique number. 2504d692f113SAndrey Ignatov * 2505c5dbb89fSFlorent Revest * u64 bpf_get_socket_cookie(struct sock *sk) 2506c5dbb89fSFlorent Revest * Description 2507c5dbb89fSFlorent Revest * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2508c5dbb89fSFlorent Revest * *sk*, but gets socket from a BTF **struct sock**. This helper 2509c5dbb89fSFlorent Revest * also works for sleepable programs. 2510c5dbb89fSFlorent Revest * Return 2511c5dbb89fSFlorent Revest * A 8-byte long unique number or 0 if *sk* is NULL. 2512c5dbb89fSFlorent Revest * 2513c6b5fb86SQuentin Monnet * u32 bpf_get_socket_uid(struct sk_buff *skb) 2514e40fbbf0SUsama Arif * Description 2515e40fbbf0SUsama Arif * Get the owner UID of the socked associated to *skb*. 2516c6b5fb86SQuentin Monnet * Return 2517c6b5fb86SQuentin Monnet * The owner UID of the socket associated to *skb*. If the socket 2518c6b5fb86SQuentin Monnet * is **NULL**, or if it is not a full socket (i.e. if it is a 2519c6b5fb86SQuentin Monnet * time-wait or a request socket instead), **overflowuid** value 2520c6b5fb86SQuentin Monnet * is returned (note that **overflowuid** might also be the actual 2521c6b5fb86SQuentin Monnet * UID value for the socket). 2522c6b5fb86SQuentin Monnet * 2523bdb7b79bSAndrii Nakryiko * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2524fa15601aSQuentin Monnet * Description 2525fa15601aSQuentin Monnet * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2526fa15601aSQuentin Monnet * to value *hash*. 2527fa15601aSQuentin Monnet * Return 2528fa15601aSQuentin Monnet * 0 2529fa15601aSQuentin Monnet * 2530bdb7b79bSAndrii Nakryiko * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 25317aa79a86SQuentin Monnet * Description 25327aa79a86SQuentin Monnet * Emulate a call to **setsockopt()** on the socket associated to 25337aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 25347aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 25357aa79a86SQuentin Monnet * must be specified, see **setsockopt(2)** for more information. 25367aa79a86SQuentin Monnet * The option value of length *optlen* is pointed by *optval*. 25377aa79a86SQuentin Monnet * 2538beecf11bSStanislav Fomichev * *bpf_socket* should be one of the following: 2539ab8d7809SQuentin Monnet * 2540beecf11bSStanislav Fomichev * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2541beecf11bSStanislav Fomichev * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2542beecf11bSStanislav Fomichev * and **BPF_CGROUP_INET6_CONNECT**. 2543beecf11bSStanislav Fomichev * 25447aa79a86SQuentin Monnet * This helper actually implements a subset of **setsockopt()**. 25457aa79a86SQuentin Monnet * It supports the following *level*\ s: 25467aa79a86SQuentin Monnet * 25477aa79a86SQuentin Monnet * * **SOL_SOCKET**, which supports the following *optname*\ s: 25487aa79a86SQuentin Monnet * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2549f9bcf968SDmitry Yakunin * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2550f9bcf968SDmitry Yakunin * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 25517aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports the following *optname*\ s: 25527aa79a86SQuentin Monnet * **TCP_CONGESTION**, **TCP_BPF_IW**, 2553f9bcf968SDmitry Yakunin * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2554f9bcf968SDmitry Yakunin * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2555eca43ee6SNikita V. Shirokov * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**. 25567aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 25577aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 25587aa79a86SQuentin Monnet * Return 25597aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 25607aa79a86SQuentin Monnet * 2561bdb7b79bSAndrii Nakryiko * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2562fa15601aSQuentin Monnet * Description 2563fa15601aSQuentin Monnet * Grow or shrink the room for data in the packet associated to 2564fa15601aSQuentin Monnet * *skb* by *len_diff*, and according to the selected *mode*. 2565fa15601aSQuentin Monnet * 2566836e66c2SDaniel Borkmann * By default, the helper will reset any offloaded checksum 2567836e66c2SDaniel Borkmann * indicator of the skb to CHECKSUM_NONE. This can be avoided 2568836e66c2SDaniel Borkmann * by the following flag: 2569836e66c2SDaniel Borkmann * 2570836e66c2SDaniel Borkmann * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2571836e66c2SDaniel Borkmann * checksum data of the skb to CHECKSUM_NONE. 2572836e66c2SDaniel Borkmann * 257314aa3192SWillem de Bruijn * There are two supported modes at this time: 257414aa3192SWillem de Bruijn * 257514aa3192SWillem de Bruijn * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 257614aa3192SWillem de Bruijn * (room space is added or removed below the layer 2 header). 2577fa15601aSQuentin Monnet * 2578fa15601aSQuentin Monnet * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2579fa15601aSQuentin Monnet * (room space is added or removed below the layer 3 header). 2580fa15601aSQuentin Monnet * 2581868d5235SWillem de Bruijn * The following flags are supported at this time: 25822278f6ccSWillem de Bruijn * 25832278f6ccSWillem de Bruijn * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 25842278f6ccSWillem de Bruijn * Adjusting mss in this way is not allowed for datagrams. 2585fa15601aSQuentin Monnet * 258680867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 258780867c5eSQuentin Monnet * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2588868d5235SWillem de Bruijn * Any new space is reserved to hold a tunnel header. 2589868d5235SWillem de Bruijn * Configure skb offsets and other fields accordingly. 2590868d5235SWillem de Bruijn * 259180867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 259280867c5eSQuentin Monnet * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2593868d5235SWillem de Bruijn * Use with ENCAP_L3 flags to further specify the tunnel type. 2594868d5235SWillem de Bruijn * 259580867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 259658dfc900SAlan Maguire * Use with ENCAP_L3/L4 flags to further specify the tunnel 259780867c5eSQuentin Monnet * type; *len* is the length of the inner MAC header. 259858dfc900SAlan Maguire * 2599d01b59c9SXuesen Huang * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2600d01b59c9SXuesen Huang * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2601d01b59c9SXuesen Huang * L2 type as Ethernet. 2602d01b59c9SXuesen Huang * 260332e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2604fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2605fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2606fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2607fa15601aSQuentin Monnet * direct packet access. 2608fa15601aSQuentin Monnet * Return 2609fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2610fa15601aSQuentin Monnet * 2611bdb7b79bSAndrii Nakryiko * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 2612ab127040SQuentin Monnet * Description 2613ab127040SQuentin Monnet * Redirect the packet to the endpoint referenced by *map* at 2614ab127040SQuentin Monnet * index *key*. Depending on its type, this *map* can contain 2615ab127040SQuentin Monnet * references to net devices (for forwarding packets through other 2616ab127040SQuentin Monnet * ports), or to CPUs (for redirecting XDP frames to another CPU; 2617ab127040SQuentin Monnet * but this is only implemented for native XDP (with driver 2618ab127040SQuentin Monnet * support) as of this writing). 2619ab127040SQuentin Monnet * 262043e74c02SToke Høiland-Jørgensen * The lower two bits of *flags* are used as the return code if 262143e74c02SToke Høiland-Jørgensen * the map lookup fails. This is so that the return value can be 2622ab8d7809SQuentin Monnet * one of the XDP program return codes up to **XDP_TX**, as chosen 2623e624d4edSHangbin Liu * by the caller. The higher bits of *flags* can be set to 2624e624d4edSHangbin Liu * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2625e624d4edSHangbin Liu * 2626e624d4edSHangbin Liu * With BPF_F_BROADCAST the packet will be broadcasted to all the 2627e624d4edSHangbin Liu * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2628e624d4edSHangbin Liu * interface will be excluded when do broadcasting. 2629ab127040SQuentin Monnet * 2630ab8d7809SQuentin Monnet * See also **bpf_redirect**\ (), which only supports redirecting 2631ab8d7809SQuentin Monnet * to an ifindex, but doesn't require a map to do so. 2632ab127040SQuentin Monnet * Return 2633f25975f4SToke Høiland-Jørgensen * **XDP_REDIRECT** on success, or the value of the two lower bits 2634a33d3147SJakub Wilk * of the *flags* argument on error. 2635ab127040SQuentin Monnet * 2636bdb7b79bSAndrii Nakryiko * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2637ab127040SQuentin Monnet * Description 2638ab127040SQuentin Monnet * Redirect the packet to the socket referenced by *map* (of type 2639ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2640ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 2641ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 2642ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 2643ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 2644ab127040SQuentin Monnet * Return 2645ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 2646ab127040SQuentin Monnet * 2647bdb7b79bSAndrii Nakryiko * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2648ab127040SQuentin Monnet * Description 2649ab127040SQuentin Monnet * Add an entry to, or update a *map* referencing sockets. The 2650ab127040SQuentin Monnet * *skops* is used as a new value for the entry associated to 2651ab127040SQuentin Monnet * *key*. *flags* is one of: 2652ab127040SQuentin Monnet * 2653ab127040SQuentin Monnet * **BPF_NOEXIST** 2654ab127040SQuentin Monnet * The entry for *key* must not exist in the map. 2655ab127040SQuentin Monnet * **BPF_EXIST** 2656ab127040SQuentin Monnet * The entry for *key* must already exist in the map. 2657ab127040SQuentin Monnet * **BPF_ANY** 2658ab127040SQuentin Monnet * No condition on the existence of the entry for *key*. 2659ab127040SQuentin Monnet * 2660ab127040SQuentin Monnet * If the *map* has eBPF programs (parser and verdict), those will 2661ab127040SQuentin Monnet * be inherited by the socket being added. If the socket is 2662ab127040SQuentin Monnet * already attached to eBPF programs, this results in an error. 2663ab127040SQuentin Monnet * Return 2664ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 2665ab127040SQuentin Monnet * 2666bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2667fa15601aSQuentin Monnet * Description 2668fa15601aSQuentin Monnet * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2669fa15601aSQuentin Monnet * *delta* (which can be positive or negative). Note that this 2670fa15601aSQuentin Monnet * operation modifies the address stored in *xdp_md*\ **->data**, 2671fa15601aSQuentin Monnet * so the latter must be loaded only after the helper has been 2672fa15601aSQuentin Monnet * called. 2673fa15601aSQuentin Monnet * 2674fa15601aSQuentin Monnet * The use of *xdp_md*\ **->data_meta** is optional and programs 2675fa15601aSQuentin Monnet * are not required to use it. The rationale is that when the 2676fa15601aSQuentin Monnet * packet is processed with XDP (e.g. as DoS filter), it is 2677fa15601aSQuentin Monnet * possible to push further meta data along with it before passing 2678fa15601aSQuentin Monnet * to the stack, and to give the guarantee that an ingress eBPF 2679fa15601aSQuentin Monnet * program attached as a TC classifier on the same device can pick 2680fa15601aSQuentin Monnet * this up for further post-processing. Since TC works with socket 2681fa15601aSQuentin Monnet * buffers, it remains possible to set from XDP the **mark** or 2682fa15601aSQuentin Monnet * **priority** pointers, or other pointers for the socket buffer. 2683fa15601aSQuentin Monnet * Having this scratch space generic and programmable allows for 2684fa15601aSQuentin Monnet * more flexibility as the user is free to store whatever meta 2685fa15601aSQuentin Monnet * data they need. 2686fa15601aSQuentin Monnet * 268732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2688fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2689fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2690fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2691fa15601aSQuentin Monnet * direct packet access. 2692fa15601aSQuentin Monnet * Return 2693fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 26947aa79a86SQuentin Monnet * 2695bdb7b79bSAndrii Nakryiko * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 26967aa79a86SQuentin Monnet * Description 26977aa79a86SQuentin Monnet * Read the value of a perf event counter, and store it into *buf* 26987aa79a86SQuentin Monnet * of size *buf_size*. This helper relies on a *map* of type 26997aa79a86SQuentin Monnet * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 27007aa79a86SQuentin Monnet * counter is selected when *map* is updated with perf event file 27017aa79a86SQuentin Monnet * descriptors. The *map* is an array whose size is the number of 27027aa79a86SQuentin Monnet * available CPUs, and each cell contains a value relative to one 27037aa79a86SQuentin Monnet * CPU. The value to retrieve is indicated by *flags*, that 27047aa79a86SQuentin Monnet * contains the index of the CPU to look up, masked with 27057aa79a86SQuentin Monnet * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 27067aa79a86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 27077aa79a86SQuentin Monnet * current CPU should be retrieved. 27087aa79a86SQuentin Monnet * 27097aa79a86SQuentin Monnet * This helper behaves in a way close to 27107aa79a86SQuentin Monnet * **bpf_perf_event_read**\ () helper, save that instead of 27117aa79a86SQuentin Monnet * just returning the value observed, it fills the *buf* 27127aa79a86SQuentin Monnet * structure. This allows for additional data to be retrieved: in 27137aa79a86SQuentin Monnet * particular, the enabled and running times (in *buf*\ 27147aa79a86SQuentin Monnet * **->enabled** and *buf*\ **->running**, respectively) are 27157aa79a86SQuentin Monnet * copied. In general, **bpf_perf_event_read_value**\ () is 27167aa79a86SQuentin Monnet * recommended over **bpf_perf_event_read**\ (), which has some 27177aa79a86SQuentin Monnet * ABI issues and provides fewer functionalities. 27187aa79a86SQuentin Monnet * 27197aa79a86SQuentin Monnet * These values are interesting, because hardware PMU (Performance 27207aa79a86SQuentin Monnet * Monitoring Unit) counters are limited resources. When there are 27217aa79a86SQuentin Monnet * more PMU based perf events opened than available counters, 27227aa79a86SQuentin Monnet * kernel will multiplex these events so each event gets certain 27237aa79a86SQuentin Monnet * percentage (but not all) of the PMU time. In case that 27247aa79a86SQuentin Monnet * multiplexing happens, the number of samples or counter value 27257aa79a86SQuentin Monnet * will not reflect the case compared to when no multiplexing 27267aa79a86SQuentin Monnet * occurs. This makes comparison between different runs difficult. 27277aa79a86SQuentin Monnet * Typically, the counter value should be normalized before 27287aa79a86SQuentin Monnet * comparing to other experiments. The usual normalization is done 27297aa79a86SQuentin Monnet * as follows. 27307aa79a86SQuentin Monnet * 27317aa79a86SQuentin Monnet * :: 27327aa79a86SQuentin Monnet * 27337aa79a86SQuentin Monnet * normalized_counter = counter * t_enabled / t_running 27347aa79a86SQuentin Monnet * 27357aa79a86SQuentin Monnet * Where t_enabled is the time enabled for event and t_running is 27367aa79a86SQuentin Monnet * the time running for event since last normalization. The 27377aa79a86SQuentin Monnet * enabled and running times are accumulated since the perf event 27387aa79a86SQuentin Monnet * open. To achieve scaling factor between two invocations of an 2739ab8d7809SQuentin Monnet * eBPF program, users can use CPU id as the key (which is 27407aa79a86SQuentin Monnet * typical for perf array usage model) to remember the previous 27417aa79a86SQuentin Monnet * value and do the calculation inside the eBPF program. 27427aa79a86SQuentin Monnet * Return 27437aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 27447aa79a86SQuentin Monnet * 2745bdb7b79bSAndrii Nakryiko * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 27467aa79a86SQuentin Monnet * Description 27477aa79a86SQuentin Monnet * For en eBPF program attached to a perf event, retrieve the 27487aa79a86SQuentin Monnet * value of the event counter associated to *ctx* and store it in 27497aa79a86SQuentin Monnet * the structure pointed by *buf* and of size *buf_size*. Enabled 27507aa79a86SQuentin Monnet * and running times are also stored in the structure (see 27517aa79a86SQuentin Monnet * description of helper **bpf_perf_event_read_value**\ () for 27527aa79a86SQuentin Monnet * more details). 27537aa79a86SQuentin Monnet * Return 27547aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 27557aa79a86SQuentin Monnet * 2756bdb7b79bSAndrii Nakryiko * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 27577aa79a86SQuentin Monnet * Description 27587aa79a86SQuentin Monnet * Emulate a call to **getsockopt()** on the socket associated to 27597aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 27607aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 27617aa79a86SQuentin Monnet * must be specified, see **getsockopt(2)** for more information. 27627aa79a86SQuentin Monnet * The retrieved value is stored in the structure pointed by 27637aa79a86SQuentin Monnet * *opval* and of length *optlen*. 27647aa79a86SQuentin Monnet * 2765beecf11bSStanislav Fomichev * *bpf_socket* should be one of the following: 2766ab8d7809SQuentin Monnet * 2767beecf11bSStanislav Fomichev * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2768beecf11bSStanislav Fomichev * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2769beecf11bSStanislav Fomichev * and **BPF_CGROUP_INET6_CONNECT**. 2770beecf11bSStanislav Fomichev * 27717aa79a86SQuentin Monnet * This helper actually implements a subset of **getsockopt()**. 27727aa79a86SQuentin Monnet * It supports the following *level*\ s: 27737aa79a86SQuentin Monnet * 27747aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports *optname* 27757aa79a86SQuentin Monnet * **TCP_CONGESTION**. 27767aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 27777aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 27787aa79a86SQuentin Monnet * Return 27797aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 27807aa79a86SQuentin Monnet * 2781bdb7b79bSAndrii Nakryiko * long bpf_override_return(struct pt_regs *regs, u64 rc) 27827aa79a86SQuentin Monnet * Description 27837aa79a86SQuentin Monnet * Used for error injection, this helper uses kprobes to override 27847aa79a86SQuentin Monnet * the return value of the probed function, and to set it to *rc*. 27857aa79a86SQuentin Monnet * The first argument is the context *regs* on which the kprobe 27867aa79a86SQuentin Monnet * works. 27877aa79a86SQuentin Monnet * 2788ab8d7809SQuentin Monnet * This helper works by setting the PC (program counter) 27897aa79a86SQuentin Monnet * to an override function which is run in place of the original 27907aa79a86SQuentin Monnet * probed function. This means the probed function is not run at 27917aa79a86SQuentin Monnet * all. The replacement function just returns with the required 27927aa79a86SQuentin Monnet * value. 27937aa79a86SQuentin Monnet * 27947aa79a86SQuentin Monnet * This helper has security implications, and thus is subject to 27957aa79a86SQuentin Monnet * restrictions. It is only available if the kernel was compiled 27967aa79a86SQuentin Monnet * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 27977aa79a86SQuentin Monnet * option, and in this case it only works on functions tagged with 27987aa79a86SQuentin Monnet * **ALLOW_ERROR_INJECTION** in the kernel code. 27997aa79a86SQuentin Monnet * 28007aa79a86SQuentin Monnet * Also, the helper is only available for the architectures having 28017aa79a86SQuentin Monnet * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 28027aa79a86SQuentin Monnet * x86 architecture is the only one to support this feature. 28037aa79a86SQuentin Monnet * Return 28047aa79a86SQuentin Monnet * 0 28057aa79a86SQuentin Monnet * 2806bdb7b79bSAndrii Nakryiko * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 28077aa79a86SQuentin Monnet * Description 28087aa79a86SQuentin Monnet * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 28097aa79a86SQuentin Monnet * for the full TCP socket associated to *bpf_sock_ops* to 28107aa79a86SQuentin Monnet * *argval*. 28117aa79a86SQuentin Monnet * 28127aa79a86SQuentin Monnet * The primary use of this field is to determine if there should 28137aa79a86SQuentin Monnet * be calls to eBPF programs of type 28147aa79a86SQuentin Monnet * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 28157aa79a86SQuentin Monnet * code. A program of the same type can change its value, per 28167aa79a86SQuentin Monnet * connection and as necessary, when the connection is 28177aa79a86SQuentin Monnet * established. This field is directly accessible for reading, but 28187aa79a86SQuentin Monnet * this helper must be used for updates in order to return an 28197aa79a86SQuentin Monnet * error if an eBPF program tries to set a callback that is not 28207aa79a86SQuentin Monnet * supported in the current kernel. 28217aa79a86SQuentin Monnet * 2822725721a6SViet Hoang Tran * *argval* is a flag array which can combine these flags: 28237aa79a86SQuentin Monnet * 28247aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 28257aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 28267aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 282723729ff2SStanislav Fomichev * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 28287aa79a86SQuentin Monnet * 2829725721a6SViet Hoang Tran * Therefore, this function can be used to clear a callback flag by 2830725721a6SViet Hoang Tran * setting the appropriate bit to zero. e.g. to disable the RTO 2831725721a6SViet Hoang Tran * callback: 2832725721a6SViet Hoang Tran * 2833725721a6SViet Hoang Tran * **bpf_sock_ops_cb_flags_set(bpf_sock,** 2834725721a6SViet Hoang Tran * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 2835725721a6SViet Hoang Tran * 28367aa79a86SQuentin Monnet * Here are some examples of where one could call such eBPF 28377aa79a86SQuentin Monnet * program: 28387aa79a86SQuentin Monnet * 28397aa79a86SQuentin Monnet * * When RTO fires. 28407aa79a86SQuentin Monnet * * When a packet is retransmitted. 28417aa79a86SQuentin Monnet * * When the connection terminates. 28427aa79a86SQuentin Monnet * * When a packet is sent. 28437aa79a86SQuentin Monnet * * When a packet is received. 28447aa79a86SQuentin Monnet * Return 28457aa79a86SQuentin Monnet * Code **-EINVAL** if the socket is not a full TCP socket; 28467aa79a86SQuentin Monnet * otherwise, a positive number containing the bits that could not 28477aa79a86SQuentin Monnet * be set is returned (which comes down to 0 if all bits were set 28487aa79a86SQuentin Monnet * as required). 28497aa79a86SQuentin Monnet * 2850bdb7b79bSAndrii Nakryiko * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 2851ab127040SQuentin Monnet * Description 2852ab127040SQuentin Monnet * This helper is used in programs implementing policies at the 2853ab127040SQuentin Monnet * socket level. If the message *msg* is allowed to pass (i.e. if 2854ab127040SQuentin Monnet * the verdict eBPF program returns **SK_PASS**), redirect it to 2855ab127040SQuentin Monnet * the socket referenced by *map* (of type 2856ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2857ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 2858ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 2859ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 2860ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 2861ab127040SQuentin Monnet * Return 2862ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 2863ab127040SQuentin Monnet * 2864bdb7b79bSAndrii Nakryiko * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 2865ab127040SQuentin Monnet * Description 2866ab127040SQuentin Monnet * For socket policies, apply the verdict of the eBPF program to 2867ab127040SQuentin Monnet * the next *bytes* (number of bytes) of message *msg*. 2868ab127040SQuentin Monnet * 2869ab127040SQuentin Monnet * For example, this helper can be used in the following cases: 2870ab127040SQuentin Monnet * 2871ab127040SQuentin Monnet * * A single **sendmsg**\ () or **sendfile**\ () system call 2872ab127040SQuentin Monnet * contains multiple logical messages that the eBPF program is 2873ab127040SQuentin Monnet * supposed to read and for which it should apply a verdict. 2874ab127040SQuentin Monnet * * An eBPF program only cares to read the first *bytes* of a 2875ab127040SQuentin Monnet * *msg*. If the message has a large payload, then setting up 2876ab127040SQuentin Monnet * and calling the eBPF program repeatedly for all bytes, even 2877ab127040SQuentin Monnet * though the verdict is already known, would create unnecessary 2878ab127040SQuentin Monnet * overhead. 2879ab127040SQuentin Monnet * 2880ab127040SQuentin Monnet * When called from within an eBPF program, the helper sets a 2881ab127040SQuentin Monnet * counter internal to the BPF infrastructure, that is used to 2882ab127040SQuentin Monnet * apply the last verdict to the next *bytes*. If *bytes* is 2883ab127040SQuentin Monnet * smaller than the current data being processed from a 2884ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () system call, the first 2885ab127040SQuentin Monnet * *bytes* will be sent and the eBPF program will be re-run with 2886ab127040SQuentin Monnet * the pointer for start of data pointing to byte number *bytes* 2887ab127040SQuentin Monnet * **+ 1**. If *bytes* is larger than the current data being 2888ab127040SQuentin Monnet * processed, then the eBPF verdict will be applied to multiple 2889ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2890ab127040SQuentin Monnet * consumed. 2891ab127040SQuentin Monnet * 2892ab127040SQuentin Monnet * Note that if a socket closes with the internal counter holding 2893ab127040SQuentin Monnet * a non-zero value, this is not a problem because data is not 2894ab127040SQuentin Monnet * being buffered for *bytes* and is sent as it is received. 2895ab127040SQuentin Monnet * Return 2896ab127040SQuentin Monnet * 0 2897ab127040SQuentin Monnet * 2898bdb7b79bSAndrii Nakryiko * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 2899ab127040SQuentin Monnet * Description 2900ab127040SQuentin Monnet * For socket policies, prevent the execution of the verdict eBPF 2901ab127040SQuentin Monnet * program for message *msg* until *bytes* (byte number) have been 2902ab127040SQuentin Monnet * accumulated. 2903ab127040SQuentin Monnet * 2904ab127040SQuentin Monnet * This can be used when one needs a specific number of bytes 2905ab127040SQuentin Monnet * before a verdict can be assigned, even if the data spans 2906ab127040SQuentin Monnet * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 2907ab127040SQuentin Monnet * case would be a user calling **sendmsg**\ () repeatedly with 2908ab127040SQuentin Monnet * 1-byte long message segments. Obviously, this is bad for 2909ab127040SQuentin Monnet * performance, but it is still valid. If the eBPF program needs 2910ab127040SQuentin Monnet * *bytes* bytes to validate a header, this helper can be used to 2911ab127040SQuentin Monnet * prevent the eBPF program to be called again until *bytes* have 2912ab127040SQuentin Monnet * been accumulated. 2913ab127040SQuentin Monnet * Return 2914ab127040SQuentin Monnet * 0 2915ab127040SQuentin Monnet * 2916bdb7b79bSAndrii Nakryiko * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2917ab127040SQuentin Monnet * Description 2918ab127040SQuentin Monnet * For socket policies, pull in non-linear data from user space 2919ab127040SQuentin Monnet * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2920ab127040SQuentin Monnet * **->data_end** to *start* and *end* bytes offsets into *msg*, 2921ab127040SQuentin Monnet * respectively. 2922ab127040SQuentin Monnet * 2923ab127040SQuentin Monnet * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2924ab127040SQuentin Monnet * *msg* it can only parse data that the (**data**, **data_end**) 2925ab127040SQuentin Monnet * pointers have already consumed. For **sendmsg**\ () hooks this 2926ab127040SQuentin Monnet * is likely the first scatterlist element. But for calls relying 2927ab127040SQuentin Monnet * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2928ab127040SQuentin Monnet * be the range (**0**, **0**) because the data is shared with 2929ab127040SQuentin Monnet * user space and by default the objective is to avoid allowing 2930ab127040SQuentin Monnet * user space to modify data while (or after) eBPF verdict is 2931ab127040SQuentin Monnet * being decided. This helper can be used to pull in data and to 2932ab127040SQuentin Monnet * set the start and end pointer to given values. Data will be 2933ab127040SQuentin Monnet * copied if necessary (i.e. if data was not linear and if start 2934ab127040SQuentin Monnet * and end pointers do not point to the same chunk). 2935ab127040SQuentin Monnet * 293632e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2937ab127040SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2938ab127040SQuentin Monnet * previously done by the verifier are invalidated and must be 2939ab127040SQuentin Monnet * performed again, if the helper is used in combination with 2940ab127040SQuentin Monnet * direct packet access. 2941ab127040SQuentin Monnet * 2942ab127040SQuentin Monnet * All values for *flags* are reserved for future usage, and must 2943ab127040SQuentin Monnet * be left at zero. 2944ab127040SQuentin Monnet * Return 2945ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 2946ab127040SQuentin Monnet * 2947bdb7b79bSAndrii Nakryiko * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 29487aa79a86SQuentin Monnet * Description 29497aa79a86SQuentin Monnet * Bind the socket associated to *ctx* to the address pointed by 29507aa79a86SQuentin Monnet * *addr*, of length *addr_len*. This allows for making outgoing 29517aa79a86SQuentin Monnet * connection from the desired IP address, which can be useful for 29527aa79a86SQuentin Monnet * example when all processes inside a cgroup should use one 29537aa79a86SQuentin Monnet * single IP address on a host that has multiple IP configured. 29547aa79a86SQuentin Monnet * 29557aa79a86SQuentin Monnet * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 29567aa79a86SQuentin Monnet * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 29578086fbafSStanislav Fomichev * **AF_INET6**). It's advised to pass zero port (**sin_port** 29588086fbafSStanislav Fomichev * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 29598086fbafSStanislav Fomichev * behavior and lets the kernel efficiently pick up an unused 29608086fbafSStanislav Fomichev * port as long as 4-tuple is unique. Passing non-zero port might 29618086fbafSStanislav Fomichev * lead to degraded performance. 29627aa79a86SQuentin Monnet * Return 29637aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 29642d020dd7SQuentin Monnet * 2965bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 29662d020dd7SQuentin Monnet * Description 29672d020dd7SQuentin Monnet * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2968c8741e2bSJesper Dangaard Brouer * possible to both shrink and grow the packet tail. 2969c8741e2bSJesper Dangaard Brouer * Shrink done via *delta* being a negative integer. 29702d020dd7SQuentin Monnet * 297132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 29722d020dd7SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 29732d020dd7SQuentin Monnet * previously done by the verifier are invalidated and must be 29742d020dd7SQuentin Monnet * performed again, if the helper is used in combination with 29752d020dd7SQuentin Monnet * direct packet access. 29762d020dd7SQuentin Monnet * Return 29772d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 29782d020dd7SQuentin Monnet * 2979bdb7b79bSAndrii Nakryiko * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 29802d020dd7SQuentin Monnet * Description 29812d020dd7SQuentin Monnet * Retrieve the XFRM state (IP transform framework, see also 29822d020dd7SQuentin Monnet * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 29832d020dd7SQuentin Monnet * 29842d020dd7SQuentin Monnet * The retrieved value is stored in the **struct bpf_xfrm_state** 29852d020dd7SQuentin Monnet * pointed by *xfrm_state* and of length *size*. 29862d020dd7SQuentin Monnet * 29872d020dd7SQuentin Monnet * All values for *flags* are reserved for future usage, and must 29882d020dd7SQuentin Monnet * be left at zero. 29892d020dd7SQuentin Monnet * 29902d020dd7SQuentin Monnet * This helper is available only if the kernel was compiled with 29912d020dd7SQuentin Monnet * **CONFIG_XFRM** configuration option. 29922d020dd7SQuentin Monnet * Return 29932d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 2994c195651eSYonghong Song * 2995bdb7b79bSAndrii Nakryiko * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2996c195651eSYonghong Song * Description 2997c195651eSYonghong Song * Return a user or a kernel stack in bpf program provided buffer. 2998c195651eSYonghong Song * To achieve this, the helper needs *ctx*, which is a pointer 2999c195651eSYonghong Song * to the context on which the tracing program is executed. 3000c195651eSYonghong Song * To store the stacktrace, the bpf program provides *buf* with 3001c195651eSYonghong Song * a nonnegative *size*. 3002c195651eSYonghong Song * 3003c195651eSYonghong Song * The last argument, *flags*, holds the number of stack frames to 3004c195651eSYonghong Song * skip (from 0 to 255), masked with 3005c195651eSYonghong Song * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3006c195651eSYonghong Song * the following flags: 3007c195651eSYonghong Song * 3008c195651eSYonghong Song * **BPF_F_USER_STACK** 3009c195651eSYonghong Song * Collect a user space stack instead of a kernel stack. 3010c195651eSYonghong Song * **BPF_F_USER_BUILD_ID** 3011c195651eSYonghong Song * Collect buildid+offset instead of ips for user stack, 3012c195651eSYonghong Song * only valid if **BPF_F_USER_STACK** is also specified. 3013c195651eSYonghong Song * 3014c195651eSYonghong Song * **bpf_get_stack**\ () can collect up to 3015c195651eSYonghong Song * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3016c195651eSYonghong Song * to sufficient large buffer size. Note that 3017c195651eSYonghong Song * this limit can be controlled with the **sysctl** program, and 3018c195651eSYonghong Song * that it should be manually increased in order to profile long 3019c195651eSYonghong Song * user stacks (such as stacks for Java programs). To do so, use: 3020c195651eSYonghong Song * 3021c195651eSYonghong Song * :: 3022c195651eSYonghong Song * 3023c195651eSYonghong Song * # sysctl kernel.perf_event_max_stack=<new value> 3024c195651eSYonghong Song * Return 3025ee2a0988SNamhyung Kim * The non-negative copied *buf* length equal to or less than 3026ee2a0988SNamhyung Kim * *size* on success, or a negative error in case of failure. 30274e1ec56cSDaniel Borkmann * 3028bdb7b79bSAndrii Nakryiko * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 30294e1ec56cSDaniel Borkmann * Description 30304e1ec56cSDaniel Borkmann * This helper is similar to **bpf_skb_load_bytes**\ () in that 30314e1ec56cSDaniel Borkmann * it provides an easy way to load *len* bytes from *offset* 30324e1ec56cSDaniel Borkmann * from the packet associated to *skb*, into the buffer pointed 30334e1ec56cSDaniel Borkmann * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 30344e1ec56cSDaniel Borkmann * a fifth argument *start_header* exists in order to select a 30354e1ec56cSDaniel Borkmann * base offset to start from. *start_header* can be one of: 30364e1ec56cSDaniel Borkmann * 30374e1ec56cSDaniel Borkmann * **BPF_HDR_START_MAC** 30384e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s mac header. 30394e1ec56cSDaniel Borkmann * **BPF_HDR_START_NET** 30404e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s network header. 30414e1ec56cSDaniel Borkmann * 30424e1ec56cSDaniel Borkmann * In general, "direct packet access" is the preferred method to 30434e1ec56cSDaniel Borkmann * access packet data, however, this helper is in particular useful 30444e1ec56cSDaniel Borkmann * in socket filters where *skb*\ **->data** does not always point 30454e1ec56cSDaniel Borkmann * to the start of the mac header and where "direct packet access" 30464e1ec56cSDaniel Borkmann * is not available. 30474e1ec56cSDaniel Borkmann * Return 30484e1ec56cSDaniel Borkmann * 0 on success, or a negative error in case of failure. 30494e1ec56cSDaniel Borkmann * 3050bdb7b79bSAndrii Nakryiko * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 305187f5fc7eSDavid Ahern * Description 305287f5fc7eSDavid Ahern * Do FIB lookup in kernel tables using parameters in *params*. 305387f5fc7eSDavid Ahern * If lookup is successful and result shows packet is to be 305487f5fc7eSDavid Ahern * forwarded, the neighbor tables are searched for the nexthop. 305587f5fc7eSDavid Ahern * If successful (ie., FIB lookup shows forwarding and nexthop 3056fa898d76SDavid Ahern * is resolved), the nexthop address is returned in ipv4_dst 3057fa898d76SDavid Ahern * or ipv6_dst based on family, smac is set to mac address of 3058fa898d76SDavid Ahern * egress device, dmac is set to nexthop mac address, rt_metric 30594c79579bSDavid Ahern * is set to metric from route (IPv4/IPv6 only), and ifindex 30604c79579bSDavid Ahern * is set to the device index of the nexthop from the FIB lookup. 306187f5fc7eSDavid Ahern * 306287f5fc7eSDavid Ahern * *plen* argument is the size of the passed in struct. 30637a279e93SQuentin Monnet * *flags* argument can be a combination of one or more of the 30647a279e93SQuentin Monnet * following values: 306587f5fc7eSDavid Ahern * 30667a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_DIRECT** 30677a279e93SQuentin Monnet * Do a direct table lookup vs full lookup using FIB 30687a279e93SQuentin Monnet * rules. 30697a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_OUTPUT** 30707a279e93SQuentin Monnet * Perform lookup from an egress perspective (default is 30717a279e93SQuentin Monnet * ingress). 307287f5fc7eSDavid Ahern * 307387f5fc7eSDavid Ahern * *ctx* is either **struct xdp_md** for XDP programs or 307487f5fc7eSDavid Ahern * **struct sk_buff** tc cls_act programs. 307587f5fc7eSDavid Ahern * Return 30764c79579bSDavid Ahern * * < 0 if any input argument is invalid 30774c79579bSDavid Ahern * * 0 on success (packet is forwarded, nexthop neighbor exists) 30784c79579bSDavid Ahern * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 30792bae79d2SQuentin Monnet * packet is not forwarded or needs assist from full stack 308081110384SJohn Fastabend * 3081e1850ea9SJesper Dangaard Brouer * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3082e1850ea9SJesper Dangaard Brouer * was exceeded and output params->mtu_result contains the MTU. 3083e1850ea9SJesper Dangaard Brouer * 3084bdb7b79bSAndrii Nakryiko * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 308581110384SJohn Fastabend * Description 308681110384SJohn Fastabend * Add an entry to, or update a sockhash *map* referencing sockets. 308781110384SJohn Fastabend * The *skops* is used as a new value for the entry associated to 308881110384SJohn Fastabend * *key*. *flags* is one of: 308981110384SJohn Fastabend * 309081110384SJohn Fastabend * **BPF_NOEXIST** 309181110384SJohn Fastabend * The entry for *key* must not exist in the map. 309281110384SJohn Fastabend * **BPF_EXIST** 309381110384SJohn Fastabend * The entry for *key* must already exist in the map. 309481110384SJohn Fastabend * **BPF_ANY** 309581110384SJohn Fastabend * No condition on the existence of the entry for *key*. 309681110384SJohn Fastabend * 309781110384SJohn Fastabend * If the *map* has eBPF programs (parser and verdict), those will 309881110384SJohn Fastabend * be inherited by the socket being added. If the socket is 309981110384SJohn Fastabend * already attached to eBPF programs, this results in an error. 310081110384SJohn Fastabend * Return 310181110384SJohn Fastabend * 0 on success, or a negative error in case of failure. 310281110384SJohn Fastabend * 3103bdb7b79bSAndrii Nakryiko * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 310481110384SJohn Fastabend * Description 310581110384SJohn Fastabend * This helper is used in programs implementing policies at the 310681110384SJohn Fastabend * socket level. If the message *msg* is allowed to pass (i.e. if 310781110384SJohn Fastabend * the verdict eBPF program returns **SK_PASS**), redirect it to 310881110384SJohn Fastabend * the socket referenced by *map* (of type 310981110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 311081110384SJohn Fastabend * egress interfaces can be used for redirection. The 311181110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 311281110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 311381110384SJohn Fastabend * egress path otherwise). This is the only flag supported for now. 311481110384SJohn Fastabend * Return 311581110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 311681110384SJohn Fastabend * 3117bdb7b79bSAndrii Nakryiko * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 311881110384SJohn Fastabend * Description 311981110384SJohn Fastabend * This helper is used in programs implementing policies at the 312081110384SJohn Fastabend * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 312149f3d12bSJakub Wilk * if the verdict eBPF program returns **SK_PASS**), redirect it 312281110384SJohn Fastabend * to the socket referenced by *map* (of type 312381110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 312481110384SJohn Fastabend * egress interfaces can be used for redirection. The 312581110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 312681110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 312781110384SJohn Fastabend * egress otherwise). This is the only flag supported for now. 312881110384SJohn Fastabend * Return 312981110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 3130fe94cc29SMathieu Xhonneux * 3131bdb7b79bSAndrii Nakryiko * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3132fe94cc29SMathieu Xhonneux * Description 3133fe94cc29SMathieu Xhonneux * Encapsulate the packet associated to *skb* within a Layer 3 3134fe94cc29SMathieu Xhonneux * protocol header. This header is provided in the buffer at 3135fe94cc29SMathieu Xhonneux * address *hdr*, with *len* its size in bytes. *type* indicates 3136fe94cc29SMathieu Xhonneux * the protocol of the header and can be one of: 3137fe94cc29SMathieu Xhonneux * 3138fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6** 3139fe94cc29SMathieu Xhonneux * IPv6 encapsulation with Segment Routing Header 3140fe94cc29SMathieu Xhonneux * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3141fe94cc29SMathieu Xhonneux * the IPv6 header is computed by the kernel. 3142fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6_INLINE** 3143fe94cc29SMathieu Xhonneux * Only works if *skb* contains an IPv6 packet. Insert a 3144fe94cc29SMathieu Xhonneux * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3145fe94cc29SMathieu Xhonneux * the IPv6 header. 31463e0bd37cSPeter Oskolkov * **BPF_LWT_ENCAP_IP** 31473e0bd37cSPeter Oskolkov * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 31483e0bd37cSPeter Oskolkov * must be IPv4 or IPv6, followed by zero or more 314980867c5eSQuentin Monnet * additional headers, up to **LWT_BPF_MAX_HEADROOM** 315080867c5eSQuentin Monnet * total bytes in all prepended headers. Please note that 315180867c5eSQuentin Monnet * if **skb_is_gso**\ (*skb*) is true, no more than two 315280867c5eSQuentin Monnet * headers can be prepended, and the inner header, if 315380867c5eSQuentin Monnet * present, should be either GRE or UDP/GUE. 31543e0bd37cSPeter Oskolkov * 315580867c5eSQuentin Monnet * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 315680867c5eSQuentin Monnet * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 315780867c5eSQuentin Monnet * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 315880867c5eSQuentin Monnet * **BPF_PROG_TYPE_LWT_XMIT**. 3159fe94cc29SMathieu Xhonneux * 316032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3161fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3162fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3163fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3164fe94cc29SMathieu Xhonneux * direct packet access. 3165fe94cc29SMathieu Xhonneux * Return 3166fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3167fe94cc29SMathieu Xhonneux * 3168bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3169fe94cc29SMathieu Xhonneux * Description 3170fe94cc29SMathieu Xhonneux * Store *len* bytes from address *from* into the packet 3171fe94cc29SMathieu Xhonneux * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3172fe94cc29SMathieu Xhonneux * inside the outermost IPv6 Segment Routing Header can be 3173fe94cc29SMathieu Xhonneux * modified through this helper. 3174fe94cc29SMathieu Xhonneux * 317532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3176fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3177fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3178fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3179fe94cc29SMathieu Xhonneux * direct packet access. 3180fe94cc29SMathieu Xhonneux * Return 3181fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3182fe94cc29SMathieu Xhonneux * 3183bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3184fe94cc29SMathieu Xhonneux * Description 3185fe94cc29SMathieu Xhonneux * Adjust the size allocated to TLVs in the outermost IPv6 3186fe94cc29SMathieu Xhonneux * Segment Routing Header contained in the packet associated to 3187fe94cc29SMathieu Xhonneux * *skb*, at position *offset* by *delta* bytes. Only offsets 3188fe94cc29SMathieu Xhonneux * after the segments are accepted. *delta* can be as well 3189fe94cc29SMathieu Xhonneux * positive (growing) as negative (shrinking). 3190fe94cc29SMathieu Xhonneux * 319132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3192fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3193fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3194fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3195fe94cc29SMathieu Xhonneux * direct packet access. 3196fe94cc29SMathieu Xhonneux * Return 3197fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3198fe94cc29SMathieu Xhonneux * 3199bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3200fe94cc29SMathieu Xhonneux * Description 3201fe94cc29SMathieu Xhonneux * Apply an IPv6 Segment Routing action of type *action* to the 3202fe94cc29SMathieu Xhonneux * packet associated to *skb*. Each action takes a parameter 3203fe94cc29SMathieu Xhonneux * contained at address *param*, and of length *param_len* bytes. 3204fe94cc29SMathieu Xhonneux * *action* can be one of: 3205fe94cc29SMathieu Xhonneux * 3206fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_X** 3207fe94cc29SMathieu Xhonneux * End.X action: Endpoint with Layer-3 cross-connect. 3208fe94cc29SMathieu Xhonneux * Type of *param*: **struct in6_addr**. 3209fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_T** 3210fe94cc29SMathieu Xhonneux * End.T action: Endpoint with specific IPv6 table lookup. 3211fe94cc29SMathieu Xhonneux * Type of *param*: **int**. 3212fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6** 3213fe94cc29SMathieu Xhonneux * End.B6 action: Endpoint bound to an SRv6 policy. 321480867c5eSQuentin Monnet * Type of *param*: **struct ipv6_sr_hdr**. 3215fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3216fe94cc29SMathieu Xhonneux * End.B6.Encap action: Endpoint bound to an SRv6 3217fe94cc29SMathieu Xhonneux * encapsulation policy. 321880867c5eSQuentin Monnet * Type of *param*: **struct ipv6_sr_hdr**. 3219fe94cc29SMathieu Xhonneux * 322032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3221fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3222fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3223fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3224fe94cc29SMathieu Xhonneux * direct packet access. 3225fe94cc29SMathieu Xhonneux * Return 3226fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3227f4364dcfSSean Young * 3228bdb7b79bSAndrii Nakryiko * long bpf_rc_repeat(void *ctx) 322962369db2SQuentin Monnet * Description 323062369db2SQuentin Monnet * This helper is used in programs implementing IR decoding, to 323162369db2SQuentin Monnet * report a successfully decoded repeat key message. This delays 323262369db2SQuentin Monnet * the generation of a key up event for previously generated 323362369db2SQuentin Monnet * key down event. 323462369db2SQuentin Monnet * 323562369db2SQuentin Monnet * Some IR protocols like NEC have a special IR message for 323662369db2SQuentin Monnet * repeating last button, for when a button is held down. 323762369db2SQuentin Monnet * 323862369db2SQuentin Monnet * The *ctx* should point to the lirc sample as passed into 323962369db2SQuentin Monnet * the program. 324062369db2SQuentin Monnet * 324162369db2SQuentin Monnet * This helper is only available is the kernel was compiled with 324262369db2SQuentin Monnet * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 324362369db2SQuentin Monnet * "**y**". 324462369db2SQuentin Monnet * Return 324562369db2SQuentin Monnet * 0 324662369db2SQuentin Monnet * 3247bdb7b79bSAndrii Nakryiko * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3248f4364dcfSSean Young * Description 3249f4364dcfSSean Young * This helper is used in programs implementing IR decoding, to 3250f4364dcfSSean Young * report a successfully decoded key press with *scancode*, 3251f4364dcfSSean Young * *toggle* value in the given *protocol*. The scancode will be 3252f4364dcfSSean Young * translated to a keycode using the rc keymap, and reported as 3253f4364dcfSSean Young * an input key down event. After a period a key up event is 3254f4364dcfSSean Young * generated. This period can be extended by calling either 325590b1023fSQuentin Monnet * **bpf_rc_keydown**\ () again with the same values, or calling 325690b1023fSQuentin Monnet * **bpf_rc_repeat**\ (). 3257f4364dcfSSean Young * 3258f4364dcfSSean Young * Some protocols include a toggle bit, in case the button was 3259f4364dcfSSean Young * released and pressed again between consecutive scancodes. 3260f4364dcfSSean Young * 3261f4364dcfSSean Young * The *ctx* should point to the lirc sample as passed into 3262f4364dcfSSean Young * the program. 3263f4364dcfSSean Young * 3264f4364dcfSSean Young * The *protocol* is the decoded protocol number (see 3265f4364dcfSSean Young * **enum rc_proto** for some predefined values). 3266f4364dcfSSean Young * 3267f4364dcfSSean Young * This helper is only available is the kernel was compiled with 3268f4364dcfSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3269f4364dcfSSean Young * "**y**". 3270f4364dcfSSean Young * Return 3271f4364dcfSSean Young * 0 3272f4364dcfSSean Young * 327362369db2SQuentin Monnet * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3274cb20b08eSDaniel Borkmann * Description 3275cb20b08eSDaniel Borkmann * Return the cgroup v2 id of the socket associated with the *skb*. 3276cb20b08eSDaniel Borkmann * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3277cb20b08eSDaniel Borkmann * helper for cgroup v1 by providing a tag resp. identifier that 3278cb20b08eSDaniel Borkmann * can be matched on or used for map lookups e.g. to implement 3279cb20b08eSDaniel Borkmann * policy. The cgroup v2 id of a given path in the hierarchy is 3280cb20b08eSDaniel Borkmann * exposed in user space through the f_handle API in order to get 3281cb20b08eSDaniel Borkmann * to the same 64-bit id. 3282cb20b08eSDaniel Borkmann * 3283cb20b08eSDaniel Borkmann * This helper can be used on TC egress path, but not on ingress, 3284cb20b08eSDaniel Borkmann * and is available only if the kernel was compiled with the 3285cb20b08eSDaniel Borkmann * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3286cb20b08eSDaniel Borkmann * Return 3287cb20b08eSDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 3288bf6fa2c8SYonghong Song * 3289bf6fa2c8SYonghong Song * u64 bpf_get_current_cgroup_id(void) 3290e40fbbf0SUsama Arif * Description 3291e40fbbf0SUsama Arif * Get the current cgroup id based on the cgroup within which 3292e40fbbf0SUsama Arif * the current task is running. 3293bf6fa2c8SYonghong Song * Return 3294bf6fa2c8SYonghong Song * A 64-bit integer containing the current cgroup id based 3295bf6fa2c8SYonghong Song * on the cgroup within which the current task is running. 3296cd339431SRoman Gushchin * 329762369db2SQuentin Monnet * void *bpf_get_local_storage(void *map, u64 flags) 3298cd339431SRoman Gushchin * Description 3299cd339431SRoman Gushchin * Get the pointer to the local storage area. 3300cd339431SRoman Gushchin * The type and the size of the local storage is defined 3301cd339431SRoman Gushchin * by the *map* argument. 3302cd339431SRoman Gushchin * The *flags* meaning is specific for each map type, 3303cd339431SRoman Gushchin * and has to be 0 for cgroup local storage. 3304cd339431SRoman Gushchin * 330590b1023fSQuentin Monnet * Depending on the BPF program type, a local storage area 330690b1023fSQuentin Monnet * can be shared between multiple instances of the BPF program, 3307cd339431SRoman Gushchin * running simultaneously. 3308cd339431SRoman Gushchin * 3309cd339431SRoman Gushchin * A user should care about the synchronization by himself. 331091c960b0SBrendan Jackman * For example, by using the **BPF_ATOMIC** instructions to alter 3311cd339431SRoman Gushchin * the shared data. 3312cd339431SRoman Gushchin * Return 331390b1023fSQuentin Monnet * A pointer to the local storage area. 33142dbb9b9eSMartin KaFai Lau * 3315bdb7b79bSAndrii Nakryiko * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 33162dbb9b9eSMartin KaFai Lau * Description 331790b1023fSQuentin Monnet * Select a **SO_REUSEPORT** socket from a 3318f170acdaSKuniyuki Iwashima * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 331990b1023fSQuentin Monnet * It checks the selected socket is matching the incoming 332090b1023fSQuentin Monnet * request in the socket buffer. 33212dbb9b9eSMartin KaFai Lau * Return 33222dbb9b9eSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 33236acc9b43SJoe Stringer * 332462369db2SQuentin Monnet * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 332562369db2SQuentin Monnet * Description 332662369db2SQuentin Monnet * Return id of cgroup v2 that is ancestor of cgroup associated 332762369db2SQuentin Monnet * with the *skb* at the *ancestor_level*. The root cgroup is at 332862369db2SQuentin Monnet * *ancestor_level* zero and each step down the hierarchy 332962369db2SQuentin Monnet * increments the level. If *ancestor_level* == level of cgroup 333062369db2SQuentin Monnet * associated with *skb*, then return value will be same as that 333162369db2SQuentin Monnet * of **bpf_skb_cgroup_id**\ (). 333262369db2SQuentin Monnet * 333362369db2SQuentin Monnet * The helper is useful to implement policies based on cgroups 333462369db2SQuentin Monnet * that are upper in hierarchy than immediate cgroup associated 333562369db2SQuentin Monnet * with *skb*. 333662369db2SQuentin Monnet * 333762369db2SQuentin Monnet * The format of returned id and helper limitations are same as in 333862369db2SQuentin Monnet * **bpf_skb_cgroup_id**\ (). 333962369db2SQuentin Monnet * Return 334062369db2SQuentin Monnet * The id is returned or 0 in case the id could not be retrieved. 334162369db2SQuentin Monnet * 3342f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 33436acc9b43SJoe Stringer * Description 33446acc9b43SJoe Stringer * Look for TCP socket matching *tuple*, optionally in a child 33456acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 334690b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 33476acc9b43SJoe Stringer * 33486acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 33496acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 33506acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 33516acc9b43SJoe Stringer * 33526acc9b43SJoe Stringer * *tuple_size* must be one of: 33536acc9b43SJoe Stringer * 33546acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 33556acc9b43SJoe Stringer * Look for an IPv4 socket. 33566acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 33576acc9b43SJoe Stringer * Look for an IPv6 socket. 33586acc9b43SJoe Stringer * 3359f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 3360bfdfa517SRandy Dunlap * socket lookup table in the netns associated with the *ctx* 3361f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 3362f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 3363f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 3364f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 3365f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 3366f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 33676acc9b43SJoe Stringer * 33686acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 33696acc9b43SJoe Stringer * be left at zero. 33706acc9b43SJoe Stringer * 33716acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 33726acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 33736acc9b43SJoe Stringer * Return 33740bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 33750bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 337680867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 337780867c5eSQuentin Monnet * tuple. 33786acc9b43SJoe Stringer * 3379f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 33806acc9b43SJoe Stringer * Description 33816acc9b43SJoe Stringer * Look for UDP socket matching *tuple*, optionally in a child 33826acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 338390b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 33846acc9b43SJoe Stringer * 33856acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 33866acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 33876acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 33886acc9b43SJoe Stringer * 33896acc9b43SJoe Stringer * *tuple_size* must be one of: 33906acc9b43SJoe Stringer * 33916acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 33926acc9b43SJoe Stringer * Look for an IPv4 socket. 33936acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 33946acc9b43SJoe Stringer * Look for an IPv6 socket. 33956acc9b43SJoe Stringer * 3396f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 3397bfdfa517SRandy Dunlap * socket lookup table in the netns associated with the *ctx* 3398f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 3399f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 3400f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 3401f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 3402f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 3403f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 34046acc9b43SJoe Stringer * 34056acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 34066acc9b43SJoe Stringer * be left at zero. 34076acc9b43SJoe Stringer * 34086acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 34096acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 34106acc9b43SJoe Stringer * Return 34110bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 34120bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 341380867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 341480867c5eSQuentin Monnet * tuple. 34156acc9b43SJoe Stringer * 3416a5fa25adSMartin KaFai Lau * long bpf_sk_release(void *sock) 34176acc9b43SJoe Stringer * Description 341890b1023fSQuentin Monnet * Release the reference held by *sock*. *sock* must be a 341990b1023fSQuentin Monnet * non-**NULL** pointer that was returned from 342090b1023fSQuentin Monnet * **bpf_sk_lookup_xxx**\ (). 342190b1023fSQuentin Monnet * Return 342290b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 342390b1023fSQuentin Monnet * 3424bdb7b79bSAndrii Nakryiko * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 342562369db2SQuentin Monnet * Description 342662369db2SQuentin Monnet * Push an element *value* in *map*. *flags* is one of: 342762369db2SQuentin Monnet * 342862369db2SQuentin Monnet * **BPF_EXIST** 342962369db2SQuentin Monnet * If the queue/stack is full, the oldest element is 343062369db2SQuentin Monnet * removed to make room for this. 343162369db2SQuentin Monnet * Return 343262369db2SQuentin Monnet * 0 on success, or a negative error in case of failure. 343362369db2SQuentin Monnet * 3434bdb7b79bSAndrii Nakryiko * long bpf_map_pop_elem(struct bpf_map *map, void *value) 343590b1023fSQuentin Monnet * Description 343690b1023fSQuentin Monnet * Pop an element from *map*. 343790b1023fSQuentin Monnet * Return 343890b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 343990b1023fSQuentin Monnet * 3440bdb7b79bSAndrii Nakryiko * long bpf_map_peek_elem(struct bpf_map *map, void *value) 344190b1023fSQuentin Monnet * Description 344290b1023fSQuentin Monnet * Get an element from *map* without removing it. 34436acc9b43SJoe Stringer * Return 34446acc9b43SJoe Stringer * 0 on success, or a negative error in case of failure. 34456fff607eSJohn Fastabend * 3446bdb7b79bSAndrii Nakryiko * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 34476fff607eSJohn Fastabend * Description 344890b1023fSQuentin Monnet * For socket policies, insert *len* bytes into *msg* at offset 34496fff607eSJohn Fastabend * *start*. 34506fff607eSJohn Fastabend * 34516fff607eSJohn Fastabend * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 345290b1023fSQuentin Monnet * *msg* it may want to insert metadata or options into the *msg*. 34536fff607eSJohn Fastabend * This can later be read and used by any of the lower layer BPF 34546fff607eSJohn Fastabend * hooks. 34556fff607eSJohn Fastabend * 34566fff607eSJohn Fastabend * This helper may fail if under memory pressure (a malloc 34576fff607eSJohn Fastabend * fails) in these cases BPF programs will get an appropriate 34586fff607eSJohn Fastabend * error and BPF programs will need to handle them. 34596fff607eSJohn Fastabend * Return 34606fff607eSJohn Fastabend * 0 on success, or a negative error in case of failure. 34617246d8edSJohn Fastabend * 3462bdb7b79bSAndrii Nakryiko * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 34637246d8edSJohn Fastabend * Description 34645f0e5412SAndrii Nakryiko * Will remove *len* bytes from a *msg* starting at byte *start*. 34657246d8edSJohn Fastabend * This may result in **ENOMEM** errors under certain situations if 34667246d8edSJohn Fastabend * an allocation and copy are required due to a full ring buffer. 34677246d8edSJohn Fastabend * However, the helper will try to avoid doing the allocation 34687246d8edSJohn Fastabend * if possible. Other errors can occur if input parameters are 346990b1023fSQuentin Monnet * invalid either due to *start* byte not being valid part of *msg* 34707246d8edSJohn Fastabend * payload and/or *pop* value being to large. 34717246d8edSJohn Fastabend * Return 347290b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 347301d3240aSSean Young * 3474bdb7b79bSAndrii Nakryiko * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 347501d3240aSSean Young * Description 347601d3240aSSean Young * This helper is used in programs implementing IR decoding, to 347701d3240aSSean Young * report a successfully decoded pointer movement. 347801d3240aSSean Young * 347901d3240aSSean Young * The *ctx* should point to the lirc sample as passed into 348001d3240aSSean Young * the program. 348101d3240aSSean Young * 348201d3240aSSean Young * This helper is only available is the kernel was compiled with 348301d3240aSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 348401d3240aSSean Young * "**y**". 348501d3240aSSean Young * Return 348601d3240aSSean Young * 0 348746f8bc92SMartin KaFai Lau * 3488bdb7b79bSAndrii Nakryiko * long bpf_spin_lock(struct bpf_spin_lock *lock) 34890eb09785SQuentin Monnet * Description 34900eb09785SQuentin Monnet * Acquire a spinlock represented by the pointer *lock*, which is 34910eb09785SQuentin Monnet * stored as part of a value of a map. Taking the lock allows to 34920eb09785SQuentin Monnet * safely update the rest of the fields in that value. The 34930eb09785SQuentin Monnet * spinlock can (and must) later be released with a call to 34940eb09785SQuentin Monnet * **bpf_spin_unlock**\ (\ *lock*\ ). 34950eb09785SQuentin Monnet * 34960eb09785SQuentin Monnet * Spinlocks in BPF programs come with a number of restrictions 34970eb09785SQuentin Monnet * and constraints: 34980eb09785SQuentin Monnet * 34990eb09785SQuentin Monnet * * **bpf_spin_lock** objects are only allowed inside maps of 35000eb09785SQuentin Monnet * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 35010eb09785SQuentin Monnet * list could be extended in the future). 35020eb09785SQuentin Monnet * * BTF description of the map is mandatory. 35030eb09785SQuentin Monnet * * The BPF program can take ONE lock at a time, since taking two 35040eb09785SQuentin Monnet * or more could cause dead locks. 35050eb09785SQuentin Monnet * * Only one **struct bpf_spin_lock** is allowed per map element. 35060eb09785SQuentin Monnet * * When the lock is taken, calls (either BPF to BPF or helpers) 35070eb09785SQuentin Monnet * are not allowed. 35080eb09785SQuentin Monnet * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 35090eb09785SQuentin Monnet * allowed inside a spinlock-ed region. 35100eb09785SQuentin Monnet * * The BPF program MUST call **bpf_spin_unlock**\ () to release 35110eb09785SQuentin Monnet * the lock, on all execution paths, before it returns. 35120eb09785SQuentin Monnet * * The BPF program can access **struct bpf_spin_lock** only via 35130eb09785SQuentin Monnet * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 35140eb09785SQuentin Monnet * helpers. Loading or storing data into the **struct 35150eb09785SQuentin Monnet * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 35160eb09785SQuentin Monnet * * To use the **bpf_spin_lock**\ () helper, the BTF description 35170eb09785SQuentin Monnet * of the map value must be a struct and have **struct 35180eb09785SQuentin Monnet * bpf_spin_lock** *anyname*\ **;** field at the top level. 35190eb09785SQuentin Monnet * Nested lock inside another struct is not allowed. 35200eb09785SQuentin Monnet * * The **struct bpf_spin_lock** *lock* field in a map value must 35210eb09785SQuentin Monnet * be aligned on a multiple of 4 bytes in that value. 35220eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 35230eb09785SQuentin Monnet * the **bpf_spin_lock** field to user space. 35240eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 35250eb09785SQuentin Monnet * a BPF program, do not update the **bpf_spin_lock** field. 35260eb09785SQuentin Monnet * * **bpf_spin_lock** cannot be on the stack or inside a 35270eb09785SQuentin Monnet * networking packet (it can only be inside of a map values). 35280eb09785SQuentin Monnet * * **bpf_spin_lock** is available to root only. 35290eb09785SQuentin Monnet * * Tracing programs and socket filter programs cannot use 35300eb09785SQuentin Monnet * **bpf_spin_lock**\ () due to insufficient preemption checks 35310eb09785SQuentin Monnet * (but this may change in the future). 35320eb09785SQuentin Monnet * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 35330eb09785SQuentin Monnet * Return 35340eb09785SQuentin Monnet * 0 35350eb09785SQuentin Monnet * 3536bdb7b79bSAndrii Nakryiko * long bpf_spin_unlock(struct bpf_spin_lock *lock) 35370eb09785SQuentin Monnet * Description 35380eb09785SQuentin Monnet * Release the *lock* previously locked by a call to 35390eb09785SQuentin Monnet * **bpf_spin_lock**\ (\ *lock*\ ). 35400eb09785SQuentin Monnet * Return 35410eb09785SQuentin Monnet * 0 35420eb09785SQuentin Monnet * 354346f8bc92SMartin KaFai Lau * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 354446f8bc92SMartin KaFai Lau * Description 354546f8bc92SMartin KaFai Lau * This helper gets a **struct bpf_sock** pointer such 354662369db2SQuentin Monnet * that all the fields in this **bpf_sock** can be accessed. 354746f8bc92SMartin KaFai Lau * Return 354862369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 354946f8bc92SMartin KaFai Lau * case of failure. 3550655a51e5SMartin KaFai Lau * 3551655a51e5SMartin KaFai Lau * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3552655a51e5SMartin KaFai Lau * Description 3553655a51e5SMartin KaFai Lau * This helper gets a **struct bpf_tcp_sock** pointer from a 3554655a51e5SMartin KaFai Lau * **struct bpf_sock** pointer. 3555655a51e5SMartin KaFai Lau * Return 355662369db2SQuentin Monnet * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3557655a51e5SMartin KaFai Lau * case of failure. 3558f7c917baSbrakmo * 3559bdb7b79bSAndrii Nakryiko * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3560f7c917baSbrakmo * Description 356162369db2SQuentin Monnet * Set ECN (Explicit Congestion Notification) field of IP header 356262369db2SQuentin Monnet * to **CE** (Congestion Encountered) if current value is **ECT** 356362369db2SQuentin Monnet * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 356462369db2SQuentin Monnet * and IPv4. 3565f7c917baSbrakmo * Return 356662369db2SQuentin Monnet * 1 if the **CE** flag is set (either by the current helper call 356762369db2SQuentin Monnet * or because it was already present), 0 if it is not set. 3568dbafd7ddSMartin KaFai Lau * 3569dbafd7ddSMartin KaFai Lau * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3570dbafd7ddSMartin KaFai Lau * Description 357162369db2SQuentin Monnet * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 357262369db2SQuentin Monnet * **bpf_sk_release**\ () is unnecessary and not allowed. 3573dbafd7ddSMartin KaFai Lau * Return 357462369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 3575dbafd7ddSMartin KaFai Lau * case of failure. 3576edbf8c01SLorenz Bauer * 3577edbf8c01SLorenz Bauer * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3578edbf8c01SLorenz Bauer * Description 3579edbf8c01SLorenz Bauer * Look for TCP socket matching *tuple*, optionally in a child 3580edbf8c01SLorenz Bauer * network namespace *netns*. The return value must be checked, 3581edbf8c01SLorenz Bauer * and if non-**NULL**, released via **bpf_sk_release**\ (). 3582edbf8c01SLorenz Bauer * 358380867c5eSQuentin Monnet * This function is identical to **bpf_sk_lookup_tcp**\ (), except 358480867c5eSQuentin Monnet * that it also returns timewait or request sockets. Use 358580867c5eSQuentin Monnet * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 358680867c5eSQuentin Monnet * full structure. 3587edbf8c01SLorenz Bauer * 3588edbf8c01SLorenz Bauer * This helper is available only if the kernel was compiled with 3589edbf8c01SLorenz Bauer * **CONFIG_NET** configuration option. 3590edbf8c01SLorenz Bauer * Return 3591edbf8c01SLorenz Bauer * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3592edbf8c01SLorenz Bauer * For sockets with reuseport option, the **struct bpf_sock** 359380867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 359480867c5eSQuentin Monnet * tuple. 359539904084SLorenz Bauer * 3596c0df236eSMartin KaFai Lau * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 359739904084SLorenz Bauer * Description 359880867c5eSQuentin Monnet * Check whether *iph* and *th* contain a valid SYN cookie ACK for 359980867c5eSQuentin Monnet * the listening socket in *sk*. 360039904084SLorenz Bauer * 360180867c5eSQuentin Monnet * *iph* points to the start of the IPv4 or IPv6 header, while 360280867c5eSQuentin Monnet * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3603ac80287aSMaxim Mikityanskiy * **sizeof**\ (**struct ipv6hdr**). 360439904084SLorenz Bauer * 360580867c5eSQuentin Monnet * *th* points to the start of the TCP header, while *th_len* 3606ac80287aSMaxim Mikityanskiy * contains the length of the TCP header (at least 3607ac80287aSMaxim Mikityanskiy * **sizeof**\ (**struct tcphdr**)). 360839904084SLorenz Bauer * Return 360980867c5eSQuentin Monnet * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 361080867c5eSQuentin Monnet * error otherwise. 3611808649fbSAndrey Ignatov * 3612bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3613808649fbSAndrey Ignatov * Description 3614808649fbSAndrey Ignatov * Get name of sysctl in /proc/sys/ and copy it into provided by 3615808649fbSAndrey Ignatov * program buffer *buf* of size *buf_len*. 3616808649fbSAndrey Ignatov * 3617808649fbSAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 3618808649fbSAndrey Ignatov * 3619808649fbSAndrey Ignatov * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3620808649fbSAndrey Ignatov * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3621808649fbSAndrey Ignatov * only (e.g. "tcp_mem"). 3622808649fbSAndrey Ignatov * Return 3623808649fbSAndrey Ignatov * Number of character copied (not including the trailing NUL). 3624808649fbSAndrey Ignatov * 3625808649fbSAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3626808649fbSAndrey Ignatov * truncated name in this case). 36271d11b301SAndrey Ignatov * 3628bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 36291d11b301SAndrey Ignatov * Description 36301d11b301SAndrey Ignatov * Get current value of sysctl as it is presented in /proc/sys 36311d11b301SAndrey Ignatov * (incl. newline, etc), and copy it as a string into provided 36321d11b301SAndrey Ignatov * by program buffer *buf* of size *buf_len*. 36331d11b301SAndrey Ignatov * 36341d11b301SAndrey Ignatov * The whole value is copied, no matter what file position user 36351d11b301SAndrey Ignatov * space issued e.g. sys_read at. 36361d11b301SAndrey Ignatov * 36371d11b301SAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 36381d11b301SAndrey Ignatov * Return 36391d11b301SAndrey Ignatov * Number of character copied (not including the trailing NUL). 36401d11b301SAndrey Ignatov * 36411d11b301SAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 36421d11b301SAndrey Ignatov * truncated name in this case). 36431d11b301SAndrey Ignatov * 36441d11b301SAndrey Ignatov * **-EINVAL** if current value was unavailable, e.g. because 36451d11b301SAndrey Ignatov * sysctl is uninitialized and read returns -EIO for it. 36464e63acdfSAndrey Ignatov * 3647bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 36484e63acdfSAndrey Ignatov * Description 36494e63acdfSAndrey Ignatov * Get new value being written by user space to sysctl (before 36504e63acdfSAndrey Ignatov * the actual write happens) and copy it as a string into 36514e63acdfSAndrey Ignatov * provided by program buffer *buf* of size *buf_len*. 36524e63acdfSAndrey Ignatov * 36534e63acdfSAndrey Ignatov * User space may write new value at file position > 0. 36544e63acdfSAndrey Ignatov * 36554e63acdfSAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 36564e63acdfSAndrey Ignatov * Return 36574e63acdfSAndrey Ignatov * Number of character copied (not including the trailing NUL). 36584e63acdfSAndrey Ignatov * 36594e63acdfSAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 36604e63acdfSAndrey Ignatov * truncated name in this case). 36614e63acdfSAndrey Ignatov * 36624e63acdfSAndrey Ignatov * **-EINVAL** if sysctl is being read. 36634e63acdfSAndrey Ignatov * 3664bdb7b79bSAndrii Nakryiko * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 36654e63acdfSAndrey Ignatov * Description 36664e63acdfSAndrey Ignatov * Override new value being written by user space to sysctl with 36674e63acdfSAndrey Ignatov * value provided by program in buffer *buf* of size *buf_len*. 36684e63acdfSAndrey Ignatov * 36694e63acdfSAndrey Ignatov * *buf* should contain a string in same form as provided by user 36704e63acdfSAndrey Ignatov * space on sysctl write. 36714e63acdfSAndrey Ignatov * 36724e63acdfSAndrey Ignatov * User space may write new value at file position > 0. To override 36734e63acdfSAndrey Ignatov * the whole sysctl value file position should be set to zero. 36744e63acdfSAndrey Ignatov * Return 36754e63acdfSAndrey Ignatov * 0 on success. 36764e63acdfSAndrey Ignatov * 36774e63acdfSAndrey Ignatov * **-E2BIG** if the *buf_len* is too big. 36784e63acdfSAndrey Ignatov * 36794e63acdfSAndrey Ignatov * **-EINVAL** if sysctl is being read. 3680d7a4cb9bSAndrey Ignatov * 3681bdb7b79bSAndrii Nakryiko * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3682d7a4cb9bSAndrey Ignatov * Description 3683d7a4cb9bSAndrey Ignatov * Convert the initial part of the string from buffer *buf* of 3684d7a4cb9bSAndrey Ignatov * size *buf_len* to a long integer according to the given base 3685d7a4cb9bSAndrey Ignatov * and save the result in *res*. 3686d7a4cb9bSAndrey Ignatov * 3687d7a4cb9bSAndrey Ignatov * The string may begin with an arbitrary amount of white space 368880867c5eSQuentin Monnet * (as determined by **isspace**\ (3)) followed by a single 368980867c5eSQuentin Monnet * optional '**-**' sign. 3690d7a4cb9bSAndrey Ignatov * 3691d7a4cb9bSAndrey Ignatov * Five least significant bits of *flags* encode base, other bits 3692d7a4cb9bSAndrey Ignatov * are currently unused. 3693d7a4cb9bSAndrey Ignatov * 3694d7a4cb9bSAndrey Ignatov * Base must be either 8, 10, 16 or 0 to detect it automatically 369580867c5eSQuentin Monnet * similar to user space **strtol**\ (3). 3696d7a4cb9bSAndrey Ignatov * Return 3697d7a4cb9bSAndrey Ignatov * Number of characters consumed on success. Must be positive but 369880867c5eSQuentin Monnet * no more than *buf_len*. 3699d7a4cb9bSAndrey Ignatov * 3700d7a4cb9bSAndrey Ignatov * **-EINVAL** if no valid digits were found or unsupported base 3701d7a4cb9bSAndrey Ignatov * was provided. 3702d7a4cb9bSAndrey Ignatov * 3703d7a4cb9bSAndrey Ignatov * **-ERANGE** if resulting value was out of range. 3704d7a4cb9bSAndrey Ignatov * 3705bdb7b79bSAndrii Nakryiko * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3706d7a4cb9bSAndrey Ignatov * Description 3707d7a4cb9bSAndrey Ignatov * Convert the initial part of the string from buffer *buf* of 3708d7a4cb9bSAndrey Ignatov * size *buf_len* to an unsigned long integer according to the 3709d7a4cb9bSAndrey Ignatov * given base and save the result in *res*. 3710d7a4cb9bSAndrey Ignatov * 3711d7a4cb9bSAndrey Ignatov * The string may begin with an arbitrary amount of white space 371280867c5eSQuentin Monnet * (as determined by **isspace**\ (3)). 3713d7a4cb9bSAndrey Ignatov * 3714d7a4cb9bSAndrey Ignatov * Five least significant bits of *flags* encode base, other bits 3715d7a4cb9bSAndrey Ignatov * are currently unused. 3716d7a4cb9bSAndrey Ignatov * 3717d7a4cb9bSAndrey Ignatov * Base must be either 8, 10, 16 or 0 to detect it automatically 371880867c5eSQuentin Monnet * similar to user space **strtoul**\ (3). 3719d7a4cb9bSAndrey Ignatov * Return 3720d7a4cb9bSAndrey Ignatov * Number of characters consumed on success. Must be positive but 372180867c5eSQuentin Monnet * no more than *buf_len*. 3722d7a4cb9bSAndrey Ignatov * 3723d7a4cb9bSAndrey Ignatov * **-EINVAL** if no valid digits were found or unsupported base 3724d7a4cb9bSAndrey Ignatov * was provided. 3725d7a4cb9bSAndrey Ignatov * 3726d7a4cb9bSAndrey Ignatov * **-ERANGE** if resulting value was out of range. 37276ac99e8fSMartin KaFai Lau * 372830897832SKP Singh * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 37296ac99e8fSMartin KaFai Lau * Description 373080867c5eSQuentin Monnet * Get a bpf-local-storage from a *sk*. 37316ac99e8fSMartin KaFai Lau * 37326ac99e8fSMartin KaFai Lau * Logically, it could be thought of getting the value from 37336ac99e8fSMartin KaFai Lau * a *map* with *sk* as the **key**. From this 37346ac99e8fSMartin KaFai Lau * perspective, the usage is not much different from 373580867c5eSQuentin Monnet * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 373680867c5eSQuentin Monnet * helper enforces the key must be a full socket and the map must 373780867c5eSQuentin Monnet * be a **BPF_MAP_TYPE_SK_STORAGE** also. 37386ac99e8fSMartin KaFai Lau * 37396ac99e8fSMartin KaFai Lau * Underneath, the value is stored locally at *sk* instead of 374080867c5eSQuentin Monnet * the *map*. The *map* is used as the bpf-local-storage 374180867c5eSQuentin Monnet * "type". The bpf-local-storage "type" (i.e. the *map*) is 374280867c5eSQuentin Monnet * searched against all bpf-local-storages residing at *sk*. 37436ac99e8fSMartin KaFai Lau * 374430897832SKP Singh * *sk* is a kernel **struct sock** pointer for LSM program. 374530897832SKP Singh * *sk* is a **struct bpf_sock** pointer for other program types. 374630897832SKP Singh * 374780867c5eSQuentin Monnet * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 37486ac99e8fSMartin KaFai Lau * used such that a new bpf-local-storage will be 37496ac99e8fSMartin KaFai Lau * created if one does not exist. *value* can be used 375080867c5eSQuentin Monnet * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 37516ac99e8fSMartin KaFai Lau * the initial value of a bpf-local-storage. If *value* is 375280867c5eSQuentin Monnet * **NULL**, the new bpf-local-storage will be zero initialized. 37536ac99e8fSMartin KaFai Lau * Return 37546ac99e8fSMartin KaFai Lau * A bpf-local-storage pointer is returned on success. 37556ac99e8fSMartin KaFai Lau * 37566ac99e8fSMartin KaFai Lau * **NULL** if not found or there was an error in adding 37576ac99e8fSMartin KaFai Lau * a new bpf-local-storage. 37586ac99e8fSMartin KaFai Lau * 375930897832SKP Singh * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 37606ac99e8fSMartin KaFai Lau * Description 376180867c5eSQuentin Monnet * Delete a bpf-local-storage from a *sk*. 37626ac99e8fSMartin KaFai Lau * Return 37636ac99e8fSMartin KaFai Lau * 0 on success. 37646ac99e8fSMartin KaFai Lau * 37656ac99e8fSMartin KaFai Lau * **-ENOENT** if the bpf-local-storage cannot be found. 3766592a3498SMartin KaFai Lau * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 37678b401f9eSYonghong Song * 3768bdb7b79bSAndrii Nakryiko * long bpf_send_signal(u32 sig) 37698b401f9eSYonghong Song * Description 37708482941fSYonghong Song * Send signal *sig* to the process of the current task. 37718482941fSYonghong Song * The signal may be delivered to any of this process's threads. 37728b401f9eSYonghong Song * Return 37738b401f9eSYonghong Song * 0 on success or successfully queued. 37748b401f9eSYonghong Song * 37758b401f9eSYonghong Song * **-EBUSY** if work queue under nmi is full. 37768b401f9eSYonghong Song * 37778b401f9eSYonghong Song * **-EINVAL** if *sig* is invalid. 37788b401f9eSYonghong Song * 37798b401f9eSYonghong Song * **-EPERM** if no permission to send the *sig*. 37808b401f9eSYonghong Song * 37818b401f9eSYonghong Song * **-EAGAIN** if bpf program can try again. 378270d66244SPetar Penkov * 3783c0df236eSMartin KaFai Lau * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 378470d66244SPetar Penkov * Description 378570d66244SPetar Penkov * Try to issue a SYN cookie for the packet with corresponding 378670d66244SPetar Penkov * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 378770d66244SPetar Penkov * 378870d66244SPetar Penkov * *iph* points to the start of the IPv4 or IPv6 header, while 378970d66244SPetar Penkov * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3790ac80287aSMaxim Mikityanskiy * **sizeof**\ (**struct ipv6hdr**). 379170d66244SPetar Penkov * 379270d66244SPetar Penkov * *th* points to the start of the TCP header, while *th_len* 3793ac80287aSMaxim Mikityanskiy * contains the length of the TCP header with options (at least 3794ac80287aSMaxim Mikityanskiy * **sizeof**\ (**struct tcphdr**)). 379570d66244SPetar Penkov * Return 379670d66244SPetar Penkov * On success, lower 32 bits hold the generated SYN cookie in 379770d66244SPetar Penkov * followed by 16 bits which hold the MSS value for that cookie, 379870d66244SPetar Penkov * and the top 16 bits are unused. 379970d66244SPetar Penkov * 380070d66244SPetar Penkov * On failure, the returned value is one of the following: 380170d66244SPetar Penkov * 380270d66244SPetar Penkov * **-EINVAL** SYN cookie cannot be issued due to error 380370d66244SPetar Penkov * 380470d66244SPetar Penkov * **-ENOENT** SYN cookie should not be issued (no SYN flood) 380570d66244SPetar Penkov * 380670d66244SPetar Penkov * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 380770d66244SPetar Penkov * 380870d66244SPetar Penkov * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 3809a7658e1aSAlexei Starovoitov * 3810bdb7b79bSAndrii Nakryiko * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3811a7658e1aSAlexei Starovoitov * Description 3812a7658e1aSAlexei Starovoitov * Write raw *data* blob into a special BPF perf event held by 3813a7658e1aSAlexei Starovoitov * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3814a7658e1aSAlexei Starovoitov * event must have the following attributes: **PERF_SAMPLE_RAW** 3815a7658e1aSAlexei Starovoitov * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3816a7658e1aSAlexei Starovoitov * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3817a7658e1aSAlexei Starovoitov * 3818a7658e1aSAlexei Starovoitov * The *flags* are used to indicate the index in *map* for which 3819a7658e1aSAlexei Starovoitov * the value must be put, masked with **BPF_F_INDEX_MASK**. 3820a7658e1aSAlexei Starovoitov * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3821a7658e1aSAlexei Starovoitov * to indicate that the index of the current CPU core should be 3822a7658e1aSAlexei Starovoitov * used. 3823a7658e1aSAlexei Starovoitov * 3824a7658e1aSAlexei Starovoitov * The value to write, of *size*, is passed through eBPF stack and 3825a7658e1aSAlexei Starovoitov * pointed by *data*. 3826a7658e1aSAlexei Starovoitov * 3827a7658e1aSAlexei Starovoitov * *ctx* is a pointer to in-kernel struct sk_buff. 3828a7658e1aSAlexei Starovoitov * 3829a7658e1aSAlexei Starovoitov * This helper is similar to **bpf_perf_event_output**\ () but 3830a7658e1aSAlexei Starovoitov * restricted to raw_tracepoint bpf programs. 3831a7658e1aSAlexei Starovoitov * Return 3832a7658e1aSAlexei Starovoitov * 0 on success, or a negative error in case of failure. 38336ae08ae3SDaniel Borkmann * 3834bdb7b79bSAndrii Nakryiko * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 38356ae08ae3SDaniel Borkmann * Description 38366ae08ae3SDaniel Borkmann * Safely attempt to read *size* bytes from user space address 38376ae08ae3SDaniel Borkmann * *unsafe_ptr* and store the data in *dst*. 38386ae08ae3SDaniel Borkmann * Return 38396ae08ae3SDaniel Borkmann * 0 on success, or a negative error in case of failure. 38406ae08ae3SDaniel Borkmann * 3841bdb7b79bSAndrii Nakryiko * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 38426ae08ae3SDaniel Borkmann * Description 38436ae08ae3SDaniel Borkmann * Safely attempt to read *size* bytes from kernel space address 38446ae08ae3SDaniel Borkmann * *unsafe_ptr* and store the data in *dst*. 38456ae08ae3SDaniel Borkmann * Return 38466ae08ae3SDaniel Borkmann * 0 on success, or a negative error in case of failure. 38476ae08ae3SDaniel Borkmann * 3848bdb7b79bSAndrii Nakryiko * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 38496ae08ae3SDaniel Borkmann * Description 38506ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe user address 38516ae08ae3SDaniel Borkmann * *unsafe_ptr* to *dst*. The *size* should include the 38526ae08ae3SDaniel Borkmann * terminating NUL byte. In case the string length is smaller than 38536ae08ae3SDaniel Borkmann * *size*, the target is not padded with further NUL bytes. If the 38546ae08ae3SDaniel Borkmann * string length is larger than *size*, just *size*-1 bytes are 38556ae08ae3SDaniel Borkmann * copied and the last byte is set to NUL. 38566ae08ae3SDaniel Borkmann * 3857c6458e72SBrendan Jackman * On success, returns the number of bytes that were written, 3858c6458e72SBrendan Jackman * including the terminal NUL. This makes this helper useful in 3859c6458e72SBrendan Jackman * tracing programs for reading strings, and more importantly to 3860c6458e72SBrendan Jackman * get its length at runtime. See the following snippet: 38616ae08ae3SDaniel Borkmann * 38626ae08ae3SDaniel Borkmann * :: 38636ae08ae3SDaniel Borkmann * 38646ae08ae3SDaniel Borkmann * SEC("kprobe/sys_open") 38656ae08ae3SDaniel Borkmann * void bpf_sys_open(struct pt_regs *ctx) 38666ae08ae3SDaniel Borkmann * { 38676ae08ae3SDaniel Borkmann * char buf[PATHLEN]; // PATHLEN is defined to 256 38686ae08ae3SDaniel Borkmann * int res = bpf_probe_read_user_str(buf, sizeof(buf), 38696ae08ae3SDaniel Borkmann * ctx->di); 38706ae08ae3SDaniel Borkmann * 38716ae08ae3SDaniel Borkmann * // Consume buf, for example push it to 38726ae08ae3SDaniel Borkmann * // userspace via bpf_perf_event_output(); we 38736ae08ae3SDaniel Borkmann * // can use res (the string length) as event 38746ae08ae3SDaniel Borkmann * // size, after checking its boundaries. 38756ae08ae3SDaniel Borkmann * } 38766ae08ae3SDaniel Borkmann * 3877ab8d7809SQuentin Monnet * In comparison, using **bpf_probe_read_user**\ () helper here 38786ae08ae3SDaniel Borkmann * instead to read the string would require to estimate the length 38796ae08ae3SDaniel Borkmann * at compile time, and would often result in copying more memory 38806ae08ae3SDaniel Borkmann * than necessary. 38816ae08ae3SDaniel Borkmann * 38826ae08ae3SDaniel Borkmann * Another useful use case is when parsing individual process 38836ae08ae3SDaniel Borkmann * arguments or individual environment variables navigating 38846ae08ae3SDaniel Borkmann * *current*\ **->mm->arg_start** and *current*\ 38856ae08ae3SDaniel Borkmann * **->mm->env_start**: using this helper and the return value, 38866ae08ae3SDaniel Borkmann * one can quickly iterate at the right offset of the memory area. 38876ae08ae3SDaniel Borkmann * Return 3888c6458e72SBrendan Jackman * On success, the strictly positive length of the output string, 38896ae08ae3SDaniel Borkmann * including the trailing NUL character. On error, a negative 38906ae08ae3SDaniel Borkmann * value. 38916ae08ae3SDaniel Borkmann * 3892bdb7b79bSAndrii Nakryiko * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 38936ae08ae3SDaniel Borkmann * Description 38946ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 3895ab8d7809SQuentin Monnet * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 38966ae08ae3SDaniel Borkmann * Return 38976ae08ae3SDaniel Borkmann * On success, the strictly positive length of the string, including 38986ae08ae3SDaniel Borkmann * the trailing NUL character. On error, a negative value. 3899206057feSMartin KaFai Lau * 3900bdb7b79bSAndrii Nakryiko * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 3901206057feSMartin KaFai Lau * Description 3902ab8d7809SQuentin Monnet * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 3903206057feSMartin KaFai Lau * *rcv_nxt* is the ack_seq to be sent out. 3904206057feSMartin KaFai Lau * Return 3905206057feSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 3906206057feSMartin KaFai Lau * 3907bdb7b79bSAndrii Nakryiko * long bpf_send_signal_thread(u32 sig) 39088482941fSYonghong Song * Description 39098482941fSYonghong Song * Send signal *sig* to the thread corresponding to the current task. 39108482941fSYonghong Song * Return 39118482941fSYonghong Song * 0 on success or successfully queued. 39128482941fSYonghong Song * 39138482941fSYonghong Song * **-EBUSY** if work queue under nmi is full. 39148482941fSYonghong Song * 39158482941fSYonghong Song * **-EINVAL** if *sig* is invalid. 39168482941fSYonghong Song * 39178482941fSYonghong Song * **-EPERM** if no permission to send the *sig*. 39188482941fSYonghong Song * 39198482941fSYonghong Song * **-EAGAIN** if bpf program can try again. 39205576b991SMartin KaFai Lau * 39215576b991SMartin KaFai Lau * u64 bpf_jiffies64(void) 39225576b991SMartin KaFai Lau * Description 39235576b991SMartin KaFai Lau * Obtain the 64bit jiffies 39245576b991SMartin KaFai Lau * Return 39255576b991SMartin KaFai Lau * The 64 bit jiffies 3926fff7b643SDaniel Xu * 3927bdb7b79bSAndrii Nakryiko * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3928fff7b643SDaniel Xu * Description 3929fff7b643SDaniel Xu * For an eBPF program attached to a perf event, retrieve the 3930ab8d7809SQuentin Monnet * branch records (**struct perf_branch_entry**) associated to *ctx* 3931fff7b643SDaniel Xu * and store it in the buffer pointed by *buf* up to size 3932fff7b643SDaniel Xu * *size* bytes. 3933fff7b643SDaniel Xu * Return 3934fff7b643SDaniel Xu * On success, number of bytes written to *buf*. On error, a 3935fff7b643SDaniel Xu * negative value. 3936fff7b643SDaniel Xu * 3937fff7b643SDaniel Xu * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3938fff7b643SDaniel Xu * instead return the number of bytes required to store all the 3939fff7b643SDaniel Xu * branch entries. If this flag is set, *buf* may be NULL. 3940fff7b643SDaniel Xu * 3941fff7b643SDaniel Xu * **-EINVAL** if arguments invalid or **size** not a multiple 3942ab8d7809SQuentin Monnet * of **sizeof**\ (**struct perf_branch_entry**\ ). 3943fff7b643SDaniel Xu * 3944fff7b643SDaniel Xu * **-ENOENT** if architecture does not support branch records. 3945b4490c5cSCarlos Neira * 3946bdb7b79bSAndrii Nakryiko * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3947b4490c5cSCarlos Neira * Description 3948b4490c5cSCarlos Neira * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3949b4490c5cSCarlos Neira * *namespace* will be returned in *nsdata*. 3950ab8d7809SQuentin Monnet * Return 3951ab8d7809SQuentin Monnet * 0 on success, or one of the following in case of failure: 3952b4490c5cSCarlos Neira * 3953b4490c5cSCarlos Neira * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3954b4490c5cSCarlos Neira * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3955b4490c5cSCarlos Neira * 3956b4490c5cSCarlos Neira * **-ENOENT** if pidns does not exists for the current task. 3957b4490c5cSCarlos Neira * 3958bdb7b79bSAndrii Nakryiko * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3959d831ee84SEelco Chaudron * Description 3960d831ee84SEelco Chaudron * Write raw *data* blob into a special BPF perf event held by 3961d831ee84SEelco Chaudron * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3962d831ee84SEelco Chaudron * event must have the following attributes: **PERF_SAMPLE_RAW** 3963d831ee84SEelco Chaudron * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3964d831ee84SEelco Chaudron * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3965d831ee84SEelco Chaudron * 3966d831ee84SEelco Chaudron * The *flags* are used to indicate the index in *map* for which 3967d831ee84SEelco Chaudron * the value must be put, masked with **BPF_F_INDEX_MASK**. 3968d831ee84SEelco Chaudron * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3969d831ee84SEelco Chaudron * to indicate that the index of the current CPU core should be 3970d831ee84SEelco Chaudron * used. 3971d831ee84SEelco Chaudron * 3972d831ee84SEelco Chaudron * The value to write, of *size*, is passed through eBPF stack and 3973d831ee84SEelco Chaudron * pointed by *data*. 3974d831ee84SEelco Chaudron * 3975d831ee84SEelco Chaudron * *ctx* is a pointer to in-kernel struct xdp_buff. 3976d831ee84SEelco Chaudron * 3977d831ee84SEelco Chaudron * This helper is similar to **bpf_perf_eventoutput**\ () but 3978d831ee84SEelco Chaudron * restricted to raw_tracepoint bpf programs. 3979d831ee84SEelco Chaudron * Return 3980d831ee84SEelco Chaudron * 0 on success, or a negative error in case of failure. 3981f318903cSDaniel Borkmann * 3982f318903cSDaniel Borkmann * u64 bpf_get_netns_cookie(void *ctx) 3983f318903cSDaniel Borkmann * Description 3984f318903cSDaniel Borkmann * Retrieve the cookie (generated by the kernel) of the network 3985f318903cSDaniel Borkmann * namespace the input *ctx* is associated with. The network 3986f318903cSDaniel Borkmann * namespace cookie remains stable for its lifetime and provides 3987f318903cSDaniel Borkmann * a global identifier that can be assumed unique. If *ctx* is 3988f318903cSDaniel Borkmann * NULL, then the helper returns the cookie for the initial 3989f318903cSDaniel Borkmann * network namespace. The cookie itself is very similar to that 3990ab8d7809SQuentin Monnet * of **bpf_get_socket_cookie**\ () helper, but for network 3991ab8d7809SQuentin Monnet * namespaces instead of sockets. 3992f318903cSDaniel Borkmann * Return 3993f318903cSDaniel Borkmann * A 8-byte long opaque number. 39940f09abd1SDaniel Borkmann * 39950f09abd1SDaniel Borkmann * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 39960f09abd1SDaniel Borkmann * Description 39970f09abd1SDaniel Borkmann * Return id of cgroup v2 that is ancestor of the cgroup associated 39980f09abd1SDaniel Borkmann * with the current task at the *ancestor_level*. The root cgroup 39990f09abd1SDaniel Borkmann * is at *ancestor_level* zero and each step down the hierarchy 40000f09abd1SDaniel Borkmann * increments the level. If *ancestor_level* == level of cgroup 40010f09abd1SDaniel Borkmann * associated with the current task, then return value will be the 40020f09abd1SDaniel Borkmann * same as that of **bpf_get_current_cgroup_id**\ (). 40030f09abd1SDaniel Borkmann * 40040f09abd1SDaniel Borkmann * The helper is useful to implement policies based on cgroups 40050f09abd1SDaniel Borkmann * that are upper in hierarchy than immediate cgroup associated 40060f09abd1SDaniel Borkmann * with the current task. 40070f09abd1SDaniel Borkmann * 40080f09abd1SDaniel Borkmann * The format of returned id and helper limitations are same as in 40090f09abd1SDaniel Borkmann * **bpf_get_current_cgroup_id**\ (). 40100f09abd1SDaniel Borkmann * Return 40110f09abd1SDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 4012cf7fbe66SJoe Stringer * 401327e5203bSMartin KaFai Lau * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 4014cf7fbe66SJoe Stringer * Description 4015e9ddbb77SJakub Sitnicki * Helper is overloaded depending on BPF program type. This 4016e9ddbb77SJakub Sitnicki * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 4017e9ddbb77SJakub Sitnicki * **BPF_PROG_TYPE_SCHED_ACT** programs. 4018e9ddbb77SJakub Sitnicki * 4019cf7fbe66SJoe Stringer * Assign the *sk* to the *skb*. When combined with appropriate 4020cf7fbe66SJoe Stringer * routing configuration to receive the packet towards the socket, 4021cf7fbe66SJoe Stringer * will cause *skb* to be delivered to the specified socket. 4022cf7fbe66SJoe Stringer * Subsequent redirection of *skb* via **bpf_redirect**\ (), 4023cf7fbe66SJoe Stringer * **bpf_clone_redirect**\ () or other methods outside of BPF may 4024cf7fbe66SJoe Stringer * interfere with successful delivery to the socket. 4025cf7fbe66SJoe Stringer * 4026cf7fbe66SJoe Stringer * This operation is only valid from TC ingress path. 4027cf7fbe66SJoe Stringer * 4028cf7fbe66SJoe Stringer * The *flags* argument must be zero. 4029cf7fbe66SJoe Stringer * Return 4030ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 4031cf7fbe66SJoe Stringer * 4032ab8d7809SQuentin Monnet * **-EINVAL** if specified *flags* are not supported. 4033ab8d7809SQuentin Monnet * 4034ab8d7809SQuentin Monnet * **-ENOENT** if the socket is unavailable for assignment. 4035ab8d7809SQuentin Monnet * 4036ab8d7809SQuentin Monnet * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4037ab8d7809SQuentin Monnet * 4038ab8d7809SQuentin Monnet * **-EOPNOTSUPP** if the operation is not supported, for example 4039ab8d7809SQuentin Monnet * a call from outside of TC ingress. 4040ab8d7809SQuentin Monnet * 4041ab8d7809SQuentin Monnet * **-ESOCKTNOSUPPORT** if the socket type is not supported 4042ab8d7809SQuentin Monnet * (reuseport). 404371d19214SMaciej Żenczykowski * 4044e9ddbb77SJakub Sitnicki * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4045e9ddbb77SJakub Sitnicki * Description 4046e9ddbb77SJakub Sitnicki * Helper is overloaded depending on BPF program type. This 4047e9ddbb77SJakub Sitnicki * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4048e9ddbb77SJakub Sitnicki * 4049e9ddbb77SJakub Sitnicki * Select the *sk* as a result of a socket lookup. 4050e9ddbb77SJakub Sitnicki * 4051e9ddbb77SJakub Sitnicki * For the operation to succeed passed socket must be compatible 4052e9ddbb77SJakub Sitnicki * with the packet description provided by the *ctx* object. 4053e9ddbb77SJakub Sitnicki * 4054e9ddbb77SJakub Sitnicki * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4055e9ddbb77SJakub Sitnicki * be an exact match. While IP family (**AF_INET** or 4056e9ddbb77SJakub Sitnicki * **AF_INET6**) must be compatible, that is IPv6 sockets 4057e9ddbb77SJakub Sitnicki * that are not v6-only can be selected for IPv4 packets. 4058e9ddbb77SJakub Sitnicki * 4059e9ddbb77SJakub Sitnicki * Only TCP listeners and UDP unconnected sockets can be 4060e9ddbb77SJakub Sitnicki * selected. *sk* can also be NULL to reset any previous 4061e9ddbb77SJakub Sitnicki * selection. 4062e9ddbb77SJakub Sitnicki * 4063e9ddbb77SJakub Sitnicki * *flags* argument can combination of following values: 4064e9ddbb77SJakub Sitnicki * 4065e9ddbb77SJakub Sitnicki * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4066e9ddbb77SJakub Sitnicki * socket selection, potentially done by a BPF program 4067e9ddbb77SJakub Sitnicki * that ran before us. 4068e9ddbb77SJakub Sitnicki * 4069e9ddbb77SJakub Sitnicki * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4070e9ddbb77SJakub Sitnicki * load-balancing within reuseport group for the socket 4071e9ddbb77SJakub Sitnicki * being selected. 4072e9ddbb77SJakub Sitnicki * 4073e9ddbb77SJakub Sitnicki * On success *ctx->sk* will point to the selected socket. 4074e9ddbb77SJakub Sitnicki * 4075e9ddbb77SJakub Sitnicki * Return 4076e9ddbb77SJakub Sitnicki * 0 on success, or a negative errno in case of failure. 4077e9ddbb77SJakub Sitnicki * 4078e9ddbb77SJakub Sitnicki * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4079e9ddbb77SJakub Sitnicki * not compatible with packet family (*ctx->family*). 4080e9ddbb77SJakub Sitnicki * 4081e9ddbb77SJakub Sitnicki * * **-EEXIST** if socket has been already selected, 4082e9ddbb77SJakub Sitnicki * potentially by another program, and 4083e9ddbb77SJakub Sitnicki * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4084e9ddbb77SJakub Sitnicki * 4085e9ddbb77SJakub Sitnicki * * **-EINVAL** if unsupported flags were specified. 4086e9ddbb77SJakub Sitnicki * 4087e9ddbb77SJakub Sitnicki * * **-EPROTOTYPE** if socket L4 protocol 4088e9ddbb77SJakub Sitnicki * (*sk->protocol*) doesn't match packet protocol 4089e9ddbb77SJakub Sitnicki * (*ctx->protocol*). 4090e9ddbb77SJakub Sitnicki * 4091e9ddbb77SJakub Sitnicki * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4092e9ddbb77SJakub Sitnicki * state (TCP listening or UDP unconnected). 4093e9ddbb77SJakub Sitnicki * 409471d19214SMaciej Żenczykowski * u64 bpf_ktime_get_boot_ns(void) 409571d19214SMaciej Żenczykowski * Description 409671d19214SMaciej Żenczykowski * Return the time elapsed since system boot, in nanoseconds. 409771d19214SMaciej Żenczykowski * Does include the time the system was suspended. 4098ab8d7809SQuentin Monnet * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 409971d19214SMaciej Żenczykowski * Return 410071d19214SMaciej Żenczykowski * Current *ktime*. 4101492e639fSYonghong Song * 4102bdb7b79bSAndrii Nakryiko * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4103492e639fSYonghong Song * Description 4104ab8d7809SQuentin Monnet * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4105ab8d7809SQuentin Monnet * out the format string. 4106492e639fSYonghong Song * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4107492e639fSYonghong Song * the format string itself. The *data* and *data_len* are format string 4108ab8d7809SQuentin Monnet * arguments. The *data* are a **u64** array and corresponding format string 4109492e639fSYonghong Song * values are stored in the array. For strings and pointers where pointees 4110492e639fSYonghong Song * are accessed, only the pointer values are stored in the *data* array. 4111a42effb0SDave Marchevsky * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4112492e639fSYonghong Song * 4113492e639fSYonghong Song * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4114492e639fSYonghong Song * Reading kernel memory may fail due to either invalid address or 4115492e639fSYonghong Song * valid address but requiring a major memory fault. If reading kernel memory 4116492e639fSYonghong Song * fails, the string for **%s** will be an empty string, and the ip 4117492e639fSYonghong Song * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4118ab8d7809SQuentin Monnet * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4119492e639fSYonghong Song * Return 4120ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 4121492e639fSYonghong Song * 4122ab8d7809SQuentin Monnet * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4123492e639fSYonghong Song * by returning 1 from bpf program. 4124ab8d7809SQuentin Monnet * 4125ab8d7809SQuentin Monnet * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4126ab8d7809SQuentin Monnet * 4127ab8d7809SQuentin Monnet * **-E2BIG** if *fmt* contains too many format specifiers. 4128ab8d7809SQuentin Monnet * 4129ab8d7809SQuentin Monnet * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4130492e639fSYonghong Song * 4131bdb7b79bSAndrii Nakryiko * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4132492e639fSYonghong Song * Description 4133ab8d7809SQuentin Monnet * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4134492e639fSYonghong Song * The *m* represents the seq_file. The *data* and *len* represent the 4135492e639fSYonghong Song * data to write in bytes. 4136492e639fSYonghong Song * Return 4137ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 4138492e639fSYonghong Song * 4139ab8d7809SQuentin Monnet * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4140f307fa2cSAndrey Ignatov * 4141a5fa25adSMartin KaFai Lau * u64 bpf_sk_cgroup_id(void *sk) 4142f307fa2cSAndrey Ignatov * Description 4143f307fa2cSAndrey Ignatov * Return the cgroup v2 id of the socket *sk*. 4144f307fa2cSAndrey Ignatov * 4145a5fa25adSMartin KaFai Lau * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4146f307fa2cSAndrey Ignatov * returned from **bpf_sk_lookup_xxx**\ (), 4147f307fa2cSAndrey Ignatov * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4148f307fa2cSAndrey Ignatov * same as in **bpf_skb_cgroup_id**\ (). 4149f307fa2cSAndrey Ignatov * 4150f307fa2cSAndrey Ignatov * This helper is available only if the kernel was compiled with 4151f307fa2cSAndrey Ignatov * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4152f307fa2cSAndrey Ignatov * Return 4153f307fa2cSAndrey Ignatov * The id is returned or 0 in case the id could not be retrieved. 4154f307fa2cSAndrey Ignatov * 4155a5fa25adSMartin KaFai Lau * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4156f307fa2cSAndrey Ignatov * Description 4157f307fa2cSAndrey Ignatov * Return id of cgroup v2 that is ancestor of cgroup associated 4158f307fa2cSAndrey Ignatov * with the *sk* at the *ancestor_level*. The root cgroup is at 4159f307fa2cSAndrey Ignatov * *ancestor_level* zero and each step down the hierarchy 4160f307fa2cSAndrey Ignatov * increments the level. If *ancestor_level* == level of cgroup 4161f307fa2cSAndrey Ignatov * associated with *sk*, then return value will be same as that 4162f307fa2cSAndrey Ignatov * of **bpf_sk_cgroup_id**\ (). 4163f307fa2cSAndrey Ignatov * 4164f307fa2cSAndrey Ignatov * The helper is useful to implement policies based on cgroups 4165f307fa2cSAndrey Ignatov * that are upper in hierarchy than immediate cgroup associated 4166f307fa2cSAndrey Ignatov * with *sk*. 4167f307fa2cSAndrey Ignatov * 4168f307fa2cSAndrey Ignatov * The format of returned id and helper limitations are same as in 4169f307fa2cSAndrey Ignatov * **bpf_sk_cgroup_id**\ (). 4170f307fa2cSAndrey Ignatov * Return 4171f307fa2cSAndrey Ignatov * The id is returned or 0 in case the id could not be retrieved. 4172457f4436SAndrii Nakryiko * 4173e1613b57SAndrii Nakryiko * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4174457f4436SAndrii Nakryiko * Description 4175457f4436SAndrii Nakryiko * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4176bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4177bcc7f554SQuentin Monnet * of new data availability is sent. 4178bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4179bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 41805c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 41815c507329SPedro Tammela * of new data availability is sent. 41825c507329SPedro Tammela * 41835c507329SPedro Tammela * An adaptive notification is a notification sent whenever the user-space 41845c507329SPedro Tammela * process has caught up and consumed all available payloads. In case the user-space 41855c507329SPedro Tammela * process is still processing a previous payload, then no notification is needed 41865c507329SPedro Tammela * as it will process the newly added payload automatically. 4187457f4436SAndrii Nakryiko * Return 4188bcc7f554SQuentin Monnet * 0 on success, or a negative error in case of failure. 4189457f4436SAndrii Nakryiko * 4190457f4436SAndrii Nakryiko * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4191457f4436SAndrii Nakryiko * Description 4192457f4436SAndrii Nakryiko * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 41935c507329SPedro Tammela * *flags* must be 0. 4194457f4436SAndrii Nakryiko * Return 4195457f4436SAndrii Nakryiko * Valid pointer with *size* bytes of memory available; NULL, 4196457f4436SAndrii Nakryiko * otherwise. 4197457f4436SAndrii Nakryiko * 4198457f4436SAndrii Nakryiko * void bpf_ringbuf_submit(void *data, u64 flags) 4199457f4436SAndrii Nakryiko * Description 4200457f4436SAndrii Nakryiko * Submit reserved ring buffer sample, pointed to by *data*. 4201bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4202bcc7f554SQuentin Monnet * of new data availability is sent. 4203bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4204bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 42055c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 42065c507329SPedro Tammela * of new data availability is sent. 42075c507329SPedro Tammela * 42085c507329SPedro Tammela * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4209457f4436SAndrii Nakryiko * Return 4210457f4436SAndrii Nakryiko * Nothing. Always succeeds. 4211457f4436SAndrii Nakryiko * 4212457f4436SAndrii Nakryiko * void bpf_ringbuf_discard(void *data, u64 flags) 4213457f4436SAndrii Nakryiko * Description 4214457f4436SAndrii Nakryiko * Discard reserved ring buffer sample, pointed to by *data*. 4215bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4216bcc7f554SQuentin Monnet * of new data availability is sent. 4217bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4218bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 42195c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 42205c507329SPedro Tammela * of new data availability is sent. 42215c507329SPedro Tammela * 42225c507329SPedro Tammela * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4223457f4436SAndrii Nakryiko * Return 4224457f4436SAndrii Nakryiko * Nothing. Always succeeds. 4225457f4436SAndrii Nakryiko * 4226457f4436SAndrii Nakryiko * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4227457f4436SAndrii Nakryiko * Description 4228457f4436SAndrii Nakryiko * Query various characteristics of provided ring buffer. What 4229457f4436SAndrii Nakryiko * exactly is queries is determined by *flags*: 4230bcc7f554SQuentin Monnet * 4231bcc7f554SQuentin Monnet * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4232bcc7f554SQuentin Monnet * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4233bcc7f554SQuentin Monnet * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4234bcc7f554SQuentin Monnet * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4235bcc7f554SQuentin Monnet * 4236bcc7f554SQuentin Monnet * Data returned is just a momentary snapshot of actual values 4237457f4436SAndrii Nakryiko * and could be inaccurate, so this facility should be used to 4238457f4436SAndrii Nakryiko * power heuristics and for reporting, not to make 100% correct 4239457f4436SAndrii Nakryiko * calculation. 4240457f4436SAndrii Nakryiko * Return 4241bcc7f554SQuentin Monnet * Requested value, or 0, if *flags* are not recognized. 42427cdec54fSDaniel Borkmann * 4243bdb7b79bSAndrii Nakryiko * long bpf_csum_level(struct sk_buff *skb, u64 level) 42447cdec54fSDaniel Borkmann * Description 42457cdec54fSDaniel Borkmann * Change the skbs checksum level by one layer up or down, or 42467cdec54fSDaniel Borkmann * reset it entirely to none in order to have the stack perform 42477cdec54fSDaniel Borkmann * checksum validation. The level is applicable to the following 42487cdec54fSDaniel Borkmann * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 42497cdec54fSDaniel Borkmann * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 42507cdec54fSDaniel Borkmann * through **bpf_skb_adjust_room**\ () helper with passing in 42517cdec54fSDaniel Borkmann * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 42527cdec54fSDaniel Borkmann * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 42537cdec54fSDaniel Borkmann * the UDP header is removed. Similarly, an encap of the latter 42547cdec54fSDaniel Borkmann * into the former could be accompanied by a helper call to 42557cdec54fSDaniel Borkmann * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 42567cdec54fSDaniel Borkmann * skb is still intended to be processed in higher layers of the 42577cdec54fSDaniel Borkmann * stack instead of just egressing at tc. 42587cdec54fSDaniel Borkmann * 42597cdec54fSDaniel Borkmann * There are three supported level settings at this time: 42607cdec54fSDaniel Borkmann * 42617cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 42627cdec54fSDaniel Borkmann * with CHECKSUM_UNNECESSARY. 42637cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 42647cdec54fSDaniel Borkmann * with CHECKSUM_UNNECESSARY. 42657cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 42667cdec54fSDaniel Borkmann * sets CHECKSUM_NONE to force checksum validation by the stack. 42677cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 42687cdec54fSDaniel Borkmann * skb->csum_level. 42697cdec54fSDaniel Borkmann * Return 42707cdec54fSDaniel Borkmann * 0 on success, or a negative error in case of failure. In the 42717cdec54fSDaniel Borkmann * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 42727cdec54fSDaniel Borkmann * is returned or the error code -EACCES in case the skb is not 42737cdec54fSDaniel Borkmann * subject to CHECKSUM_UNNECESSARY. 4274af7ec138SYonghong Song * 4275af7ec138SYonghong Song * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4276af7ec138SYonghong Song * Description 4277af7ec138SYonghong Song * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4278af7ec138SYonghong Song * Return 4279938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4280478cfbdfSYonghong Song * 4281478cfbdfSYonghong Song * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4282478cfbdfSYonghong Song * Description 4283478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4284478cfbdfSYonghong Song * Return 4285938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4286478cfbdfSYonghong Song * 4287478cfbdfSYonghong Song * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4288478cfbdfSYonghong Song * Description 4289478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4290478cfbdfSYonghong Song * Return 4291938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4292478cfbdfSYonghong Song * 4293478cfbdfSYonghong Song * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4294478cfbdfSYonghong Song * Description 4295478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4296478cfbdfSYonghong Song * Return 4297938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 42980d4fad3eSYonghong Song * 42990d4fad3eSYonghong Song * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 43000d4fad3eSYonghong Song * Description 43010d4fad3eSYonghong Song * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 43020d4fad3eSYonghong Song * Return 4303938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4304fa28dcb8SSong Liu * 4305fa28dcb8SSong Liu * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4306fa28dcb8SSong Liu * Description 4307fa28dcb8SSong Liu * Return a user or a kernel stack in bpf program provided buffer. 4308fa28dcb8SSong Liu * To achieve this, the helper needs *task*, which is a valid 4309938c3efdSQuentin Monnet * pointer to **struct task_struct**. To store the stacktrace, the 4310fa28dcb8SSong Liu * bpf program provides *buf* with a nonnegative *size*. 4311fa28dcb8SSong Liu * 4312fa28dcb8SSong Liu * The last argument, *flags*, holds the number of stack frames to 4313fa28dcb8SSong Liu * skip (from 0 to 255), masked with 4314fa28dcb8SSong Liu * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4315fa28dcb8SSong Liu * the following flags: 4316fa28dcb8SSong Liu * 4317fa28dcb8SSong Liu * **BPF_F_USER_STACK** 4318fa28dcb8SSong Liu * Collect a user space stack instead of a kernel stack. 4319fa28dcb8SSong Liu * **BPF_F_USER_BUILD_ID** 4320fa28dcb8SSong Liu * Collect buildid+offset instead of ips for user stack, 4321fa28dcb8SSong Liu * only valid if **BPF_F_USER_STACK** is also specified. 4322fa28dcb8SSong Liu * 4323fa28dcb8SSong Liu * **bpf_get_task_stack**\ () can collect up to 4324fa28dcb8SSong Liu * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4325fa28dcb8SSong Liu * to sufficient large buffer size. Note that 4326fa28dcb8SSong Liu * this limit can be controlled with the **sysctl** program, and 4327fa28dcb8SSong Liu * that it should be manually increased in order to profile long 4328fa28dcb8SSong Liu * user stacks (such as stacks for Java programs). To do so, use: 4329fa28dcb8SSong Liu * 4330fa28dcb8SSong Liu * :: 4331fa28dcb8SSong Liu * 4332fa28dcb8SSong Liu * # sysctl kernel.perf_event_max_stack=<new value> 4333fa28dcb8SSong Liu * Return 4334ee2a0988SNamhyung Kim * The non-negative copied *buf* length equal to or less than 4335ee2a0988SNamhyung Kim * *size* on success, or a negative error in case of failure. 4336fa28dcb8SSong Liu * 43370813a841SMartin KaFai Lau * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 43380813a841SMartin KaFai Lau * Description 43390813a841SMartin KaFai Lau * Load header option. Support reading a particular TCP header 4340938c3efdSQuentin Monnet * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 43410813a841SMartin KaFai Lau * 43420813a841SMartin KaFai Lau * If *flags* is 0, it will search the option from the 4343938c3efdSQuentin Monnet * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 43440813a841SMartin KaFai Lau * has details on what skb_data contains under different 4345938c3efdSQuentin Monnet * *skops*\ **->op**. 43460813a841SMartin KaFai Lau * 43470813a841SMartin KaFai Lau * The first byte of the *searchby_res* specifies the 43480813a841SMartin KaFai Lau * kind that it wants to search. 43490813a841SMartin KaFai Lau * 43500813a841SMartin KaFai Lau * If the searching kind is an experimental kind 43510813a841SMartin KaFai Lau * (i.e. 253 or 254 according to RFC6994). It also 43520813a841SMartin KaFai Lau * needs to specify the "magic" which is either 43530813a841SMartin KaFai Lau * 2 bytes or 4 bytes. It then also needs to 43540813a841SMartin KaFai Lau * specify the size of the magic by using 43550813a841SMartin KaFai Lau * the 2nd byte which is "kind-length" of a TCP 43560813a841SMartin KaFai Lau * header option and the "kind-length" also 43570813a841SMartin KaFai Lau * includes the first 2 bytes "kind" and "kind-length" 43580813a841SMartin KaFai Lau * itself as a normal TCP header option also does. 43590813a841SMartin KaFai Lau * 43600813a841SMartin KaFai Lau * For example, to search experimental kind 254 with 43610813a841SMartin KaFai Lau * 2 byte magic 0xeB9F, the searchby_res should be 43620813a841SMartin KaFai Lau * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 43630813a841SMartin KaFai Lau * 43640813a841SMartin KaFai Lau * To search for the standard window scale option (3), 4365938c3efdSQuentin Monnet * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 43660813a841SMartin KaFai Lau * Note, kind-length must be 0 for regular option. 43670813a841SMartin KaFai Lau * 43680813a841SMartin KaFai Lau * Searching for No-Op (0) and End-of-Option-List (1) are 43690813a841SMartin KaFai Lau * not supported. 43700813a841SMartin KaFai Lau * 43710813a841SMartin KaFai Lau * *len* must be at least 2 bytes which is the minimal size 43720813a841SMartin KaFai Lau * of a header option. 43730813a841SMartin KaFai Lau * 43740813a841SMartin KaFai Lau * Supported flags: 4375938c3efdSQuentin Monnet * 43760813a841SMartin KaFai Lau * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 43770813a841SMartin KaFai Lau * saved_syn packet or the just-received syn packet. 43780813a841SMartin KaFai Lau * 43790813a841SMartin KaFai Lau * Return 43800813a841SMartin KaFai Lau * > 0 when found, the header option is copied to *searchby_res*. 4381938c3efdSQuentin Monnet * The return value is the total length copied. On failure, a 4382938c3efdSQuentin Monnet * negative error code is returned: 43830813a841SMartin KaFai Lau * 4384938c3efdSQuentin Monnet * **-EINVAL** if a parameter is invalid. 43850813a841SMartin KaFai Lau * 4386938c3efdSQuentin Monnet * **-ENOMSG** if the option is not found. 43870813a841SMartin KaFai Lau * 4388938c3efdSQuentin Monnet * **-ENOENT** if no syn packet is available when 4389938c3efdSQuentin Monnet * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 43900813a841SMartin KaFai Lau * 4391938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space. Only *len* number of 43920813a841SMartin KaFai Lau * bytes are copied. 43930813a841SMartin KaFai Lau * 4394938c3efdSQuentin Monnet * **-EFAULT** on failure to parse the header options in the 4395938c3efdSQuentin Monnet * packet. 43960813a841SMartin KaFai Lau * 4397938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4398938c3efdSQuentin Monnet * *skops*\ **->op**. 43990813a841SMartin KaFai Lau * 44000813a841SMartin KaFai Lau * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 44010813a841SMartin KaFai Lau * Description 44020813a841SMartin KaFai Lau * Store header option. The data will be copied 44030813a841SMartin KaFai Lau * from buffer *from* with length *len* to the TCP header. 44040813a841SMartin KaFai Lau * 44050813a841SMartin KaFai Lau * The buffer *from* should have the whole option that 44060813a841SMartin KaFai Lau * includes the kind, kind-length, and the actual 44070813a841SMartin KaFai Lau * option data. The *len* must be at least kind-length 44080813a841SMartin KaFai Lau * long. The kind-length does not have to be 4 byte 44090813a841SMartin KaFai Lau * aligned. The kernel will take care of the padding 44100813a841SMartin KaFai Lau * and setting the 4 bytes aligned value to th->doff. 44110813a841SMartin KaFai Lau * 44120813a841SMartin KaFai Lau * This helper will check for duplicated option 44130813a841SMartin KaFai Lau * by searching the same option in the outgoing skb. 44140813a841SMartin KaFai Lau * 44150813a841SMartin KaFai Lau * This helper can only be called during 4416938c3efdSQuentin Monnet * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 44170813a841SMartin KaFai Lau * 44180813a841SMartin KaFai Lau * Return 44190813a841SMartin KaFai Lau * 0 on success, or negative error in case of failure: 44200813a841SMartin KaFai Lau * 4421938c3efdSQuentin Monnet * **-EINVAL** If param is invalid. 44220813a841SMartin KaFai Lau * 4423938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space in the header. 44240813a841SMartin KaFai Lau * Nothing has been written 44250813a841SMartin KaFai Lau * 4426938c3efdSQuentin Monnet * **-EEXIST** if the option already exists. 44270813a841SMartin KaFai Lau * 4428938c3efdSQuentin Monnet * **-EFAULT** on failrue to parse the existing header options. 44290813a841SMartin KaFai Lau * 4430938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4431938c3efdSQuentin Monnet * *skops*\ **->op**. 44320813a841SMartin KaFai Lau * 44330813a841SMartin KaFai Lau * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 44340813a841SMartin KaFai Lau * Description 44350813a841SMartin KaFai Lau * Reserve *len* bytes for the bpf header option. The 4436938c3efdSQuentin Monnet * space will be used by **bpf_store_hdr_opt**\ () later in 4437938c3efdSQuentin Monnet * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 44380813a841SMartin KaFai Lau * 4439938c3efdSQuentin Monnet * If **bpf_reserve_hdr_opt**\ () is called multiple times, 44400813a841SMartin KaFai Lau * the total number of bytes will be reserved. 44410813a841SMartin KaFai Lau * 44420813a841SMartin KaFai Lau * This helper can only be called during 4443938c3efdSQuentin Monnet * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 44440813a841SMartin KaFai Lau * 44450813a841SMartin KaFai Lau * Return 44460813a841SMartin KaFai Lau * 0 on success, or negative error in case of failure: 44470813a841SMartin KaFai Lau * 4448938c3efdSQuentin Monnet * **-EINVAL** if a parameter is invalid. 44490813a841SMartin KaFai Lau * 4450938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space in the header. 44510813a841SMartin KaFai Lau * 4452938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4453938c3efdSQuentin Monnet * *skops*\ **->op**. 44546e22ab9dSJiri Olsa * 44558ea63684SKP Singh * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 44568ea63684SKP Singh * Description 44578ea63684SKP Singh * Get a bpf_local_storage from an *inode*. 44588ea63684SKP Singh * 44598ea63684SKP Singh * Logically, it could be thought of as getting the value from 44608ea63684SKP Singh * a *map* with *inode* as the **key**. From this 44618ea63684SKP Singh * perspective, the usage is not much different from 44628ea63684SKP Singh * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 44638ea63684SKP Singh * helper enforces the key must be an inode and the map must also 44648ea63684SKP Singh * be a **BPF_MAP_TYPE_INODE_STORAGE**. 44658ea63684SKP Singh * 44668ea63684SKP Singh * Underneath, the value is stored locally at *inode* instead of 44678ea63684SKP Singh * the *map*. The *map* is used as the bpf-local-storage 44688ea63684SKP Singh * "type". The bpf-local-storage "type" (i.e. the *map*) is 44698ea63684SKP Singh * searched against all bpf_local_storage residing at *inode*. 44708ea63684SKP Singh * 44718ea63684SKP Singh * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 44728ea63684SKP Singh * used such that a new bpf_local_storage will be 44738ea63684SKP Singh * created if one does not exist. *value* can be used 44748ea63684SKP Singh * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 44758ea63684SKP Singh * the initial value of a bpf_local_storage. If *value* is 44768ea63684SKP Singh * **NULL**, the new bpf_local_storage will be zero initialized. 44778ea63684SKP Singh * Return 44788ea63684SKP Singh * A bpf_local_storage pointer is returned on success. 44798ea63684SKP Singh * 44808ea63684SKP Singh * **NULL** if not found or there was an error in adding 44818ea63684SKP Singh * a new bpf_local_storage. 44828ea63684SKP Singh * 44838ea63684SKP Singh * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 44848ea63684SKP Singh * Description 44858ea63684SKP Singh * Delete a bpf_local_storage from an *inode*. 44868ea63684SKP Singh * Return 44878ea63684SKP Singh * 0 on success. 44888ea63684SKP Singh * 44898ea63684SKP Singh * **-ENOENT** if the bpf_local_storage cannot be found. 44906e22ab9dSJiri Olsa * 44916e22ab9dSJiri Olsa * long bpf_d_path(struct path *path, char *buf, u32 sz) 44926e22ab9dSJiri Olsa * Description 4493938c3efdSQuentin Monnet * Return full path for given **struct path** object, which 4494938c3efdSQuentin Monnet * needs to be the kernel BTF *path* object. The path is 4495938c3efdSQuentin Monnet * returned in the provided buffer *buf* of size *sz* and 44966e22ab9dSJiri Olsa * is zero terminated. 44976e22ab9dSJiri Olsa * 44986e22ab9dSJiri Olsa * Return 44996e22ab9dSJiri Olsa * On success, the strictly positive length of the string, 45006e22ab9dSJiri Olsa * including the trailing NUL character. On error, a negative 45016e22ab9dSJiri Olsa * value. 450207be4c4aSAlexei Starovoitov * 450307be4c4aSAlexei Starovoitov * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 450407be4c4aSAlexei Starovoitov * Description 450507be4c4aSAlexei Starovoitov * Read *size* bytes from user space address *user_ptr* and store 4506938c3efdSQuentin Monnet * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 450707be4c4aSAlexei Starovoitov * Return 450807be4c4aSAlexei Starovoitov * 0 on success, or a negative error in case of failure. 4509c4d0bfb4SAlan Maguire * 4510c4d0bfb4SAlan Maguire * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4511c4d0bfb4SAlan Maguire * Description 4512c4d0bfb4SAlan Maguire * Use BTF to store a string representation of *ptr*->ptr in *str*, 4513c4d0bfb4SAlan Maguire * using *ptr*->type_id. This value should specify the type 4514c4d0bfb4SAlan Maguire * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4515c4d0bfb4SAlan Maguire * can be used to look up vmlinux BTF type ids. Traversing the 4516c4d0bfb4SAlan Maguire * data structure using BTF, the type information and values are 4517c4d0bfb4SAlan Maguire * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4518c4d0bfb4SAlan Maguire * the pointer data is carried out to avoid kernel crashes during 4519c4d0bfb4SAlan Maguire * operation. Smaller types can use string space on the stack; 4520c4d0bfb4SAlan Maguire * larger programs can use map data to store the string 4521c4d0bfb4SAlan Maguire * representation. 4522c4d0bfb4SAlan Maguire * 4523c4d0bfb4SAlan Maguire * The string can be subsequently shared with userspace via 4524c4d0bfb4SAlan Maguire * bpf_perf_event_output() or ring buffer interfaces. 4525c4d0bfb4SAlan Maguire * bpf_trace_printk() is to be avoided as it places too small 4526c4d0bfb4SAlan Maguire * a limit on string size to be useful. 4527c4d0bfb4SAlan Maguire * 4528c4d0bfb4SAlan Maguire * *flags* is a combination of 4529c4d0bfb4SAlan Maguire * 4530c4d0bfb4SAlan Maguire * **BTF_F_COMPACT** 4531c4d0bfb4SAlan Maguire * no formatting around type information 4532c4d0bfb4SAlan Maguire * **BTF_F_NONAME** 4533c4d0bfb4SAlan Maguire * no struct/union member names/types 4534c4d0bfb4SAlan Maguire * **BTF_F_PTR_RAW** 4535c4d0bfb4SAlan Maguire * show raw (unobfuscated) pointer values; 4536c4d0bfb4SAlan Maguire * equivalent to printk specifier %px. 4537c4d0bfb4SAlan Maguire * **BTF_F_ZERO** 4538c4d0bfb4SAlan Maguire * show zero-valued struct/union members; they 4539c4d0bfb4SAlan Maguire * are not displayed by default 4540c4d0bfb4SAlan Maguire * 4541c4d0bfb4SAlan Maguire * Return 4542c4d0bfb4SAlan Maguire * The number of bytes that were written (or would have been 4543c4d0bfb4SAlan Maguire * written if output had to be truncated due to string size), 4544c4d0bfb4SAlan Maguire * or a negative error in cases of failure. 4545eb411377SAlan Maguire * 4546eb411377SAlan Maguire * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4547eb411377SAlan Maguire * Description 4548eb411377SAlan Maguire * Use BTF to write to seq_write a string representation of 4549eb411377SAlan Maguire * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4550eb411377SAlan Maguire * *flags* are identical to those used for bpf_snprintf_btf. 4551eb411377SAlan Maguire * Return 4552eb411377SAlan Maguire * 0 on success or a negative error in case of failure. 4553b426ce83SDaniel Borkmann * 4554b426ce83SDaniel Borkmann * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4555b426ce83SDaniel Borkmann * Description 4556b426ce83SDaniel Borkmann * See **bpf_get_cgroup_classid**\ () for the main description. 4557b426ce83SDaniel Borkmann * This helper differs from **bpf_get_cgroup_classid**\ () in that 4558b426ce83SDaniel Borkmann * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4559b426ce83SDaniel Borkmann * associated socket instead of the current process. 4560b426ce83SDaniel Borkmann * Return 4561b426ce83SDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 4562b4ab3141SDaniel Borkmann * 4563ba452c9eSToke Høiland-Jørgensen * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4564b4ab3141SDaniel Borkmann * Description 4565b4ab3141SDaniel Borkmann * Redirect the packet to another net device of index *ifindex* 4566b4ab3141SDaniel Borkmann * and fill in L2 addresses from neighboring subsystem. This helper 4567b4ab3141SDaniel Borkmann * is somewhat similar to **bpf_redirect**\ (), except that it 4568dd2ce6a5SDaniel Borkmann * populates L2 addresses as well, meaning, internally, the helper 4569ba452c9eSToke Høiland-Jørgensen * relies on the neighbor lookup for the L2 address of the nexthop. 4570ba452c9eSToke Høiland-Jørgensen * 4571ba452c9eSToke Høiland-Jørgensen * The helper will perform a FIB lookup based on the skb's 4572ba452c9eSToke Høiland-Jørgensen * networking header to get the address of the next hop, unless 4573ba452c9eSToke Høiland-Jørgensen * this is supplied by the caller in the *params* argument. The 4574ba452c9eSToke Høiland-Jørgensen * *plen* argument indicates the len of *params* and should be set 4575ba452c9eSToke Høiland-Jørgensen * to 0 if *params* is NULL. 4576dd2ce6a5SDaniel Borkmann * 4577b4ab3141SDaniel Borkmann * The *flags* argument is reserved and must be 0. The helper is 4578dd2ce6a5SDaniel Borkmann * currently only supported for tc BPF program types, and enabled 4579dd2ce6a5SDaniel Borkmann * for IPv4 and IPv6 protocols. 4580b4ab3141SDaniel Borkmann * Return 4581b4ab3141SDaniel Borkmann * The helper returns **TC_ACT_REDIRECT** on success or 4582b4ab3141SDaniel Borkmann * **TC_ACT_SHOT** on error. 4583eaa6bcb7SHao Luo * 4584eaa6bcb7SHao Luo * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4585eaa6bcb7SHao Luo * Description 4586eaa6bcb7SHao Luo * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4587eaa6bcb7SHao Luo * pointer to the percpu kernel variable on *cpu*. A ksym is an 4588eaa6bcb7SHao Luo * extern variable decorated with '__ksym'. For ksym, there is a 4589eaa6bcb7SHao Luo * global var (either static or global) defined of the same name 4590eaa6bcb7SHao Luo * in the kernel. The ksym is percpu if the global var is percpu. 4591eaa6bcb7SHao Luo * The returned pointer points to the global percpu var on *cpu*. 4592eaa6bcb7SHao Luo * 4593eaa6bcb7SHao Luo * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4594eaa6bcb7SHao Luo * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4595eaa6bcb7SHao Luo * happens if *cpu* is larger than nr_cpu_ids. The caller of 4596eaa6bcb7SHao Luo * bpf_per_cpu_ptr() must check the returned value. 4597eaa6bcb7SHao Luo * Return 4598eaa6bcb7SHao Luo * A pointer pointing to the kernel percpu variable on *cpu*, or 4599eaa6bcb7SHao Luo * NULL, if *cpu* is invalid. 460063d9b80dSHao Luo * 460163d9b80dSHao Luo * void *bpf_this_cpu_ptr(const void *percpu_ptr) 460263d9b80dSHao Luo * Description 460363d9b80dSHao Luo * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 460463d9b80dSHao Luo * pointer to the percpu kernel variable on this cpu. See the 460563d9b80dSHao Luo * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 460663d9b80dSHao Luo * 460763d9b80dSHao Luo * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 460863d9b80dSHao Luo * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 460963d9b80dSHao Luo * never return NULL. 461063d9b80dSHao Luo * Return 461163d9b80dSHao Luo * A pointer pointing to the kernel percpu variable on this cpu. 46129aa1206eSDaniel Borkmann * 46139aa1206eSDaniel Borkmann * long bpf_redirect_peer(u32 ifindex, u64 flags) 46149aa1206eSDaniel Borkmann * Description 46159aa1206eSDaniel Borkmann * Redirect the packet to another net device of index *ifindex*. 46169aa1206eSDaniel Borkmann * This helper is somewhat similar to **bpf_redirect**\ (), except 46179aa1206eSDaniel Borkmann * that the redirection happens to the *ifindex*' peer device and 46189aa1206eSDaniel Borkmann * the netns switch takes place from ingress to ingress without 46199aa1206eSDaniel Borkmann * going through the CPU's backlog queue. 46209aa1206eSDaniel Borkmann * 46219aa1206eSDaniel Borkmann * The *flags* argument is reserved and must be 0. The helper is 46229aa1206eSDaniel Borkmann * currently only supported for tc BPF program types at the ingress 46239aa1206eSDaniel Borkmann * hook and for veth device types. The peer device must reside in a 46249aa1206eSDaniel Borkmann * different network namespace. 46259aa1206eSDaniel Borkmann * Return 46269aa1206eSDaniel Borkmann * The helper returns **TC_ACT_REDIRECT** on success or 46279aa1206eSDaniel Borkmann * **TC_ACT_SHOT** on error. 46284cf1bc1fSKP Singh * 46294cf1bc1fSKP Singh * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 46304cf1bc1fSKP Singh * Description 46314cf1bc1fSKP Singh * Get a bpf_local_storage from the *task*. 46324cf1bc1fSKP Singh * 46334cf1bc1fSKP Singh * Logically, it could be thought of as getting the value from 46344cf1bc1fSKP Singh * a *map* with *task* as the **key**. From this 46354cf1bc1fSKP Singh * perspective, the usage is not much different from 46364cf1bc1fSKP Singh * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 46374cf1bc1fSKP Singh * helper enforces the key must be an task_struct and the map must also 46384cf1bc1fSKP Singh * be a **BPF_MAP_TYPE_TASK_STORAGE**. 46394cf1bc1fSKP Singh * 46404cf1bc1fSKP Singh * Underneath, the value is stored locally at *task* instead of 46414cf1bc1fSKP Singh * the *map*. The *map* is used as the bpf-local-storage 46424cf1bc1fSKP Singh * "type". The bpf-local-storage "type" (i.e. the *map*) is 46434cf1bc1fSKP Singh * searched against all bpf_local_storage residing at *task*. 46444cf1bc1fSKP Singh * 46454cf1bc1fSKP Singh * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 46464cf1bc1fSKP Singh * used such that a new bpf_local_storage will be 46474cf1bc1fSKP Singh * created if one does not exist. *value* can be used 46484cf1bc1fSKP Singh * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 46494cf1bc1fSKP Singh * the initial value of a bpf_local_storage. If *value* is 46504cf1bc1fSKP Singh * **NULL**, the new bpf_local_storage will be zero initialized. 46514cf1bc1fSKP Singh * Return 46524cf1bc1fSKP Singh * A bpf_local_storage pointer is returned on success. 46534cf1bc1fSKP Singh * 46544cf1bc1fSKP Singh * **NULL** if not found or there was an error in adding 46554cf1bc1fSKP Singh * a new bpf_local_storage. 46564cf1bc1fSKP Singh * 46574cf1bc1fSKP Singh * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 46584cf1bc1fSKP Singh * Description 46594cf1bc1fSKP Singh * Delete a bpf_local_storage from a *task*. 46604cf1bc1fSKP Singh * Return 46614cf1bc1fSKP Singh * 0 on success. 46624cf1bc1fSKP Singh * 46634cf1bc1fSKP Singh * **-ENOENT** if the bpf_local_storage cannot be found. 46643ca1032aSKP Singh * 46653ca1032aSKP Singh * struct task_struct *bpf_get_current_task_btf(void) 46663ca1032aSKP Singh * Description 46673ca1032aSKP Singh * Return a BTF pointer to the "current" task. 46683ca1032aSKP Singh * This pointer can also be used in helpers that accept an 46693ca1032aSKP Singh * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 46703ca1032aSKP Singh * Return 46713ca1032aSKP Singh * Pointer to the current task. 46723f6719c7SKP Singh * 46733f6719c7SKP Singh * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 46743f6719c7SKP Singh * Description 46753f6719c7SKP Singh * Set or clear certain options on *bprm*: 46763f6719c7SKP Singh * 46773f6719c7SKP Singh * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 46783f6719c7SKP Singh * which sets the **AT_SECURE** auxv for glibc. The bit 46793f6719c7SKP Singh * is cleared if the flag is not specified. 46803f6719c7SKP Singh * Return 46813f6719c7SKP Singh * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4682d0551261SDmitrii Banshchikov * 4683d0551261SDmitrii Banshchikov * u64 bpf_ktime_get_coarse_ns(void) 4684d0551261SDmitrii Banshchikov * Description 4685d0551261SDmitrii Banshchikov * Return a coarse-grained version of the time elapsed since 4686d0551261SDmitrii Banshchikov * system boot, in nanoseconds. Does not include time the system 4687d0551261SDmitrii Banshchikov * was suspended. 4688d0551261SDmitrii Banshchikov * 4689d0551261SDmitrii Banshchikov * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4690d0551261SDmitrii Banshchikov * Return 4691d0551261SDmitrii Banshchikov * Current *ktime*. 469227672f0dSKP Singh * 469327672f0dSKP Singh * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 469427672f0dSKP Singh * Description 469527672f0dSKP Singh * Returns the stored IMA hash of the *inode* (if it's avaialable). 469627672f0dSKP Singh * If the hash is larger than *size*, then only *size* 469727672f0dSKP Singh * bytes will be copied to *dst* 469827672f0dSKP Singh * Return 469927672f0dSKP Singh * The **hash_algo** is returned on success, 470027672f0dSKP Singh * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 470127672f0dSKP Singh * invalid arguments are passed. 47024f19cab7SFlorent Revest * 47034f19cab7SFlorent Revest * struct socket *bpf_sock_from_file(struct file *file) 47044f19cab7SFlorent Revest * Description 47054f19cab7SFlorent Revest * If the given file represents a socket, returns the associated 47064f19cab7SFlorent Revest * socket. 47074f19cab7SFlorent Revest * Return 47084f19cab7SFlorent Revest * A pointer to a struct socket on success or NULL if the file is 47094f19cab7SFlorent Revest * not a socket. 471034b2021cSJesper Dangaard Brouer * 471134b2021cSJesper Dangaard Brouer * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 471234b2021cSJesper Dangaard Brouer * Description 4713e5e35e75SJesper Dangaard Brouer * Check packet size against exceeding MTU of net device (based 471434b2021cSJesper Dangaard Brouer * on *ifindex*). This helper will likely be used in combination 471534b2021cSJesper Dangaard Brouer * with helpers that adjust/change the packet size. 471634b2021cSJesper Dangaard Brouer * 471734b2021cSJesper Dangaard Brouer * The argument *len_diff* can be used for querying with a planned 471834b2021cSJesper Dangaard Brouer * size change. This allows to check MTU prior to changing packet 471934b2021cSJesper Dangaard Brouer * ctx. Providing an *len_diff* adjustment that is larger than the 472034b2021cSJesper Dangaard Brouer * actual packet size (resulting in negative packet size) will in 472134b2021cSJesper Dangaard Brouer * principle not exceed the MTU, why it is not considered a 472234b2021cSJesper Dangaard Brouer * failure. Other BPF-helpers are needed for performing the 472334b2021cSJesper Dangaard Brouer * planned size change, why the responsability for catch a negative 472434b2021cSJesper Dangaard Brouer * packet size belong in those helpers. 472534b2021cSJesper Dangaard Brouer * 472634b2021cSJesper Dangaard Brouer * Specifying *ifindex* zero means the MTU check is performed 472734b2021cSJesper Dangaard Brouer * against the current net device. This is practical if this isn't 472834b2021cSJesper Dangaard Brouer * used prior to redirect. 472934b2021cSJesper Dangaard Brouer * 4730e5e35e75SJesper Dangaard Brouer * On input *mtu_len* must be a valid pointer, else verifier will 4731e5e35e75SJesper Dangaard Brouer * reject BPF program. If the value *mtu_len* is initialized to 4732e5e35e75SJesper Dangaard Brouer * zero then the ctx packet size is use. When value *mtu_len* is 4733e5e35e75SJesper Dangaard Brouer * provided as input this specify the L3 length that the MTU check 4734e5e35e75SJesper Dangaard Brouer * is done against. Remember XDP and TC length operate at L2, but 4735e5e35e75SJesper Dangaard Brouer * this value is L3 as this correlate to MTU and IP-header tot_len 4736e5e35e75SJesper Dangaard Brouer * values which are L3 (similar behavior as bpf_fib_lookup). 4737e5e35e75SJesper Dangaard Brouer * 473834b2021cSJesper Dangaard Brouer * The Linux kernel route table can configure MTUs on a more 473934b2021cSJesper Dangaard Brouer * specific per route level, which is not provided by this helper. 474034b2021cSJesper Dangaard Brouer * For route level MTU checks use the **bpf_fib_lookup**\ () 474134b2021cSJesper Dangaard Brouer * helper. 474234b2021cSJesper Dangaard Brouer * 474334b2021cSJesper Dangaard Brouer * *ctx* is either **struct xdp_md** for XDP programs or 474434b2021cSJesper Dangaard Brouer * **struct sk_buff** for tc cls_act programs. 474534b2021cSJesper Dangaard Brouer * 474634b2021cSJesper Dangaard Brouer * The *flags* argument can be a combination of one or more of the 474734b2021cSJesper Dangaard Brouer * following values: 474834b2021cSJesper Dangaard Brouer * 474934b2021cSJesper Dangaard Brouer * **BPF_MTU_CHK_SEGS** 475034b2021cSJesper Dangaard Brouer * This flag will only works for *ctx* **struct sk_buff**. 475134b2021cSJesper Dangaard Brouer * If packet context contains extra packet segment buffers 475234b2021cSJesper Dangaard Brouer * (often knows as GSO skb), then MTU check is harder to 475334b2021cSJesper Dangaard Brouer * check at this point, because in transmit path it is 475434b2021cSJesper Dangaard Brouer * possible for the skb packet to get re-segmented 475534b2021cSJesper Dangaard Brouer * (depending on net device features). This could still be 475634b2021cSJesper Dangaard Brouer * a MTU violation, so this flag enables performing MTU 475734b2021cSJesper Dangaard Brouer * check against segments, with a different violation 475834b2021cSJesper Dangaard Brouer * return code to tell it apart. Check cannot use len_diff. 475934b2021cSJesper Dangaard Brouer * 476034b2021cSJesper Dangaard Brouer * On return *mtu_len* pointer contains the MTU value of the net 476134b2021cSJesper Dangaard Brouer * device. Remember the net device configured MTU is the L3 size, 4762e5e35e75SJesper Dangaard Brouer * which is returned here and XDP and TC length operate at L2. 476334b2021cSJesper Dangaard Brouer * Helper take this into account for you, but remember when using 4764e5e35e75SJesper Dangaard Brouer * MTU value in your BPF-code. 476534b2021cSJesper Dangaard Brouer * 476634b2021cSJesper Dangaard Brouer * Return 476734b2021cSJesper Dangaard Brouer * * 0 on success, and populate MTU value in *mtu_len* pointer. 476834b2021cSJesper Dangaard Brouer * 476934b2021cSJesper Dangaard Brouer * * < 0 if any input argument is invalid (*mtu_len* not updated) 477034b2021cSJesper Dangaard Brouer * 477134b2021cSJesper Dangaard Brouer * MTU violations return positive values, but also populate MTU 477234b2021cSJesper Dangaard Brouer * value in *mtu_len* pointer, as this can be needed for 477334b2021cSJesper Dangaard Brouer * implementing PMTU handing: 477434b2021cSJesper Dangaard Brouer * 477534b2021cSJesper Dangaard Brouer * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 477634b2021cSJesper Dangaard Brouer * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 477734b2021cSJesper Dangaard Brouer * 477869c087baSYonghong Song * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 477969c087baSYonghong Song * Description 478069c087baSYonghong Song * For each element in **map**, call **callback_fn** function with 478169c087baSYonghong Song * **map**, **callback_ctx** and other map-specific parameters. 478269c087baSYonghong Song * The **callback_fn** should be a static function and 478369c087baSYonghong Song * the **callback_ctx** should be a pointer to the stack. 478469c087baSYonghong Song * The **flags** is used to control certain aspects of the helper. 478569c087baSYonghong Song * Currently, the **flags** must be 0. 478669c087baSYonghong Song * 478769c087baSYonghong Song * The following are a list of supported map types and their 478869c087baSYonghong Song * respective expected callback signatures: 478969c087baSYonghong Song * 479069c087baSYonghong Song * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 479169c087baSYonghong Song * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 479269c087baSYonghong Song * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 479369c087baSYonghong Song * 479469c087baSYonghong Song * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 479569c087baSYonghong Song * 479669c087baSYonghong Song * For per_cpu maps, the map_value is the value on the cpu where the 479769c087baSYonghong Song * bpf_prog is running. 479869c087baSYonghong Song * 479969c087baSYonghong Song * If **callback_fn** return 0, the helper will continue to the next 480069c087baSYonghong Song * element. If return value is 1, the helper will skip the rest of 480169c087baSYonghong Song * elements and return. Other return values are not used now. 480269c087baSYonghong Song * 480369c087baSYonghong Song * Return 480469c087baSYonghong Song * The number of traversed map elements for success, **-EINVAL** for 480569c087baSYonghong Song * invalid **flags**. 48067b15523aSFlorent Revest * 48077b15523aSFlorent Revest * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 48087b15523aSFlorent Revest * Description 48097b15523aSFlorent Revest * Outputs a string into the **str** buffer of size **str_size** 48107b15523aSFlorent Revest * based on a format string stored in a read-only map pointed by 48117b15523aSFlorent Revest * **fmt**. 48127b15523aSFlorent Revest * 48137b15523aSFlorent Revest * Each format specifier in **fmt** corresponds to one u64 element 48147b15523aSFlorent Revest * in the **data** array. For strings and pointers where pointees 48157b15523aSFlorent Revest * are accessed, only the pointer values are stored in the *data* 4816a42effb0SDave Marchevsky * array. The *data_len* is the size of *data* in bytes - must be 4817a42effb0SDave Marchevsky * a multiple of 8. 48187b15523aSFlorent Revest * 48197b15523aSFlorent Revest * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 48207b15523aSFlorent Revest * memory. Reading kernel memory may fail due to either invalid 48217b15523aSFlorent Revest * address or valid address but requiring a major memory fault. If 48227b15523aSFlorent Revest * reading kernel memory fails, the string for **%s** will be an 48237b15523aSFlorent Revest * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 48247b15523aSFlorent Revest * Not returning error to bpf program is consistent with what 48257b15523aSFlorent Revest * **bpf_trace_printk**\ () does for now. 48267b15523aSFlorent Revest * 48277b15523aSFlorent Revest * Return 48287b15523aSFlorent Revest * The strictly positive length of the formatted string, including 48297b15523aSFlorent Revest * the trailing zero character. If the return value is greater than 48307b15523aSFlorent Revest * **str_size**, **str** contains a truncated string, guaranteed to 48317b15523aSFlorent Revest * be zero-terminated except when **str_size** is 0. 48327b15523aSFlorent Revest * 48337b15523aSFlorent Revest * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 483479a7f8bdSAlexei Starovoitov * 483579a7f8bdSAlexei Starovoitov * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 483679a7f8bdSAlexei Starovoitov * Description 483779a7f8bdSAlexei Starovoitov * Execute bpf syscall with given arguments. 483879a7f8bdSAlexei Starovoitov * Return 483979a7f8bdSAlexei Starovoitov * A syscall result. 48403d78417bSAlexei Starovoitov * 48413d78417bSAlexei Starovoitov * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 48423d78417bSAlexei Starovoitov * Description 48433d78417bSAlexei Starovoitov * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 48443d78417bSAlexei Starovoitov * Return 48453d78417bSAlexei Starovoitov * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 48463abea089SAlexei Starovoitov * 48473abea089SAlexei Starovoitov * long bpf_sys_close(u32 fd) 48483abea089SAlexei Starovoitov * Description 48493abea089SAlexei Starovoitov * Execute close syscall for given FD. 48503abea089SAlexei Starovoitov * Return 48513abea089SAlexei Starovoitov * A syscall result. 4852b00628b1SAlexei Starovoitov * 4853b00628b1SAlexei Starovoitov * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 4854b00628b1SAlexei Starovoitov * Description 4855b00628b1SAlexei Starovoitov * Initialize the timer. 4856b00628b1SAlexei Starovoitov * First 4 bits of *flags* specify clockid. 4857b00628b1SAlexei Starovoitov * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 4858b00628b1SAlexei Starovoitov * All other bits of *flags* are reserved. 4859b00628b1SAlexei Starovoitov * The verifier will reject the program if *timer* is not from 4860b00628b1SAlexei Starovoitov * the same *map*. 4861b00628b1SAlexei Starovoitov * Return 4862b00628b1SAlexei Starovoitov * 0 on success. 4863b00628b1SAlexei Starovoitov * **-EBUSY** if *timer* is already initialized. 4864b00628b1SAlexei Starovoitov * **-EINVAL** if invalid *flags* are passed. 4865b00628b1SAlexei Starovoitov * **-EPERM** if *timer* is in a map that doesn't have any user references. 4866b00628b1SAlexei Starovoitov * The user space should either hold a file descriptor to a map with timers 4867b00628b1SAlexei Starovoitov * or pin such map in bpffs. When map is unpinned or file descriptor is 4868b00628b1SAlexei Starovoitov * closed all timers in the map will be cancelled and freed. 4869b00628b1SAlexei Starovoitov * 4870b00628b1SAlexei Starovoitov * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 4871b00628b1SAlexei Starovoitov * Description 4872b00628b1SAlexei Starovoitov * Configure the timer to call *callback_fn* static function. 4873b00628b1SAlexei Starovoitov * Return 4874b00628b1SAlexei Starovoitov * 0 on success. 4875b00628b1SAlexei Starovoitov * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4876b00628b1SAlexei Starovoitov * **-EPERM** if *timer* is in a map that doesn't have any user references. 4877b00628b1SAlexei Starovoitov * The user space should either hold a file descriptor to a map with timers 4878b00628b1SAlexei Starovoitov * or pin such map in bpffs. When map is unpinned or file descriptor is 4879b00628b1SAlexei Starovoitov * closed all timers in the map will be cancelled and freed. 4880b00628b1SAlexei Starovoitov * 4881b00628b1SAlexei Starovoitov * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 4882b00628b1SAlexei Starovoitov * Description 4883b00628b1SAlexei Starovoitov * Set timer expiration N nanoseconds from the current time. The 4884b00628b1SAlexei Starovoitov * configured callback will be invoked in soft irq context on some cpu 4885b00628b1SAlexei Starovoitov * and will not repeat unless another bpf_timer_start() is made. 4886b00628b1SAlexei Starovoitov * In such case the next invocation can migrate to a different cpu. 4887b00628b1SAlexei Starovoitov * Since struct bpf_timer is a field inside map element the map 4888b00628b1SAlexei Starovoitov * owns the timer. The bpf_timer_set_callback() will increment refcnt 4889b00628b1SAlexei Starovoitov * of BPF program to make sure that callback_fn code stays valid. 4890b00628b1SAlexei Starovoitov * When user space reference to a map reaches zero all timers 4891b00628b1SAlexei Starovoitov * in a map are cancelled and corresponding program's refcnts are 4892b00628b1SAlexei Starovoitov * decremented. This is done to make sure that Ctrl-C of a user 4893b00628b1SAlexei Starovoitov * process doesn't leave any timers running. If map is pinned in 4894b00628b1SAlexei Starovoitov * bpffs the callback_fn can re-arm itself indefinitely. 4895b00628b1SAlexei Starovoitov * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 4896b00628b1SAlexei Starovoitov * cancel and free the timer in the given map element. 4897b00628b1SAlexei Starovoitov * The map can contain timers that invoke callback_fn-s from different 4898b00628b1SAlexei Starovoitov * programs. The same callback_fn can serve different timers from 4899b00628b1SAlexei Starovoitov * different maps if key/value layout matches across maps. 4900b00628b1SAlexei Starovoitov * Every bpf_timer_set_callback() can have different callback_fn. 4901b00628b1SAlexei Starovoitov * 4902b00628b1SAlexei Starovoitov * Return 4903b00628b1SAlexei Starovoitov * 0 on success. 4904b00628b1SAlexei Starovoitov * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 4905b00628b1SAlexei Starovoitov * or invalid *flags* are passed. 4906b00628b1SAlexei Starovoitov * 4907b00628b1SAlexei Starovoitov * long bpf_timer_cancel(struct bpf_timer *timer) 4908b00628b1SAlexei Starovoitov * Description 4909b00628b1SAlexei Starovoitov * Cancel the timer and wait for callback_fn to finish if it was running. 4910b00628b1SAlexei Starovoitov * Return 4911b00628b1SAlexei Starovoitov * 0 if the timer was not active. 4912b00628b1SAlexei Starovoitov * 1 if the timer was active. 4913b00628b1SAlexei Starovoitov * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4914b00628b1SAlexei Starovoitov * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 4915b00628b1SAlexei Starovoitov * own timer which would have led to a deadlock otherwise. 49169b99edcaSJiri Olsa * 49179b99edcaSJiri Olsa * u64 bpf_get_func_ip(void *ctx) 49189b99edcaSJiri Olsa * Description 49199ffd9f3fSJiri Olsa * Get address of the traced function (for tracing and kprobe programs). 49209b99edcaSJiri Olsa * Return 49219b99edcaSJiri Olsa * Address of the traced function. 49227adfc6c9SAndrii Nakryiko * 49237adfc6c9SAndrii Nakryiko * u64 bpf_get_attach_cookie(void *ctx) 49247adfc6c9SAndrii Nakryiko * Description 49257adfc6c9SAndrii Nakryiko * Get bpf_cookie value provided (optionally) during the program 49267adfc6c9SAndrii Nakryiko * attachment. It might be different for each individual 49277adfc6c9SAndrii Nakryiko * attachment, even if BPF program itself is the same. 49287adfc6c9SAndrii Nakryiko * Expects BPF program context *ctx* as a first argument. 49297adfc6c9SAndrii Nakryiko * 49307adfc6c9SAndrii Nakryiko * Supported for the following program types: 49317adfc6c9SAndrii Nakryiko * - kprobe/uprobe; 49327adfc6c9SAndrii Nakryiko * - tracepoint; 49337adfc6c9SAndrii Nakryiko * - perf_event. 49347adfc6c9SAndrii Nakryiko * Return 49357adfc6c9SAndrii Nakryiko * Value specified by user at BPF link creation/attachment time 49367adfc6c9SAndrii Nakryiko * or 0, if it was not specified. 4937dd6e10fbSDaniel Xu * 4938dd6e10fbSDaniel Xu * long bpf_task_pt_regs(struct task_struct *task) 4939dd6e10fbSDaniel Xu * Description 4940dd6e10fbSDaniel Xu * Get the struct pt_regs associated with **task**. 4941dd6e10fbSDaniel Xu * Return 4942dd6e10fbSDaniel Xu * A pointer to struct pt_regs. 4943856c02dbSSong Liu * 4944856c02dbSSong Liu * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 4945856c02dbSSong Liu * Description 4946856c02dbSSong Liu * Get branch trace from hardware engines like Intel LBR. The 4947856c02dbSSong Liu * hardware engine is stopped shortly after the helper is 4948856c02dbSSong Liu * called. Therefore, the user need to filter branch entries 4949856c02dbSSong Liu * based on the actual use case. To capture branch trace 4950856c02dbSSong Liu * before the trigger point of the BPF program, the helper 4951856c02dbSSong Liu * should be called at the beginning of the BPF program. 4952856c02dbSSong Liu * 4953856c02dbSSong Liu * The data is stored as struct perf_branch_entry into output 4954856c02dbSSong Liu * buffer *entries*. *size* is the size of *entries* in bytes. 4955856c02dbSSong Liu * *flags* is reserved for now and must be zero. 4956856c02dbSSong Liu * 4957856c02dbSSong Liu * Return 4958856c02dbSSong Liu * On success, number of bytes written to *buf*. On error, a 4959856c02dbSSong Liu * negative value. 4960856c02dbSSong Liu * 4961856c02dbSSong Liu * **-EINVAL** if *flags* is not zero. 4962856c02dbSSong Liu * 4963856c02dbSSong Liu * **-ENOENT** if architecture does not support branch records. 496410aceb62SDave Marchevsky * 496510aceb62SDave Marchevsky * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 496610aceb62SDave Marchevsky * Description 496710aceb62SDave Marchevsky * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 496810aceb62SDave Marchevsky * to format and can handle more format args as a result. 496910aceb62SDave Marchevsky * 497010aceb62SDave Marchevsky * Arguments are to be used as in **bpf_seq_printf**\ () helper. 497110aceb62SDave Marchevsky * Return 497210aceb62SDave Marchevsky * The number of bytes written to the buffer, or a negative error 497310aceb62SDave Marchevsky * in case of failure. 49749eeb3aa3SHengqi Chen * 49759eeb3aa3SHengqi Chen * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 49769eeb3aa3SHengqi Chen * Description 49779eeb3aa3SHengqi Chen * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 49789eeb3aa3SHengqi Chen * Return 49799eeb3aa3SHengqi Chen * *sk* if casting is valid, or **NULL** otherwise. 4980d6aef08aSKumar Kartikeya Dwivedi * 4981d6aef08aSKumar Kartikeya Dwivedi * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 4982d6aef08aSKumar Kartikeya Dwivedi * Description 4983d6aef08aSKumar Kartikeya Dwivedi * Get the address of a kernel symbol, returned in *res*. *res* is 4984d6aef08aSKumar Kartikeya Dwivedi * set to 0 if the symbol is not found. 4985d6aef08aSKumar Kartikeya Dwivedi * Return 4986d6aef08aSKumar Kartikeya Dwivedi * On success, zero. On error, a negative value. 4987d6aef08aSKumar Kartikeya Dwivedi * 4988d6aef08aSKumar Kartikeya Dwivedi * **-EINVAL** if *flags* is not zero. 4989d6aef08aSKumar Kartikeya Dwivedi * 4990d6aef08aSKumar Kartikeya Dwivedi * **-EINVAL** if string *name* is not the same size as *name_sz*. 4991d6aef08aSKumar Kartikeya Dwivedi * 4992d6aef08aSKumar Kartikeya Dwivedi * **-ENOENT** if symbol is not found. 4993d6aef08aSKumar Kartikeya Dwivedi * 4994d6aef08aSKumar Kartikeya Dwivedi * **-EPERM** if caller does not have permission to obtain kernel address. 49957c7e3d31SSong Liu * 49967c7e3d31SSong Liu * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 49977c7e3d31SSong Liu * Description 49987c7e3d31SSong Liu * Find vma of *task* that contains *addr*, call *callback_fn* 49997c7e3d31SSong Liu * function with *task*, *vma*, and *callback_ctx*. 50007c7e3d31SSong Liu * The *callback_fn* should be a static function and 50017c7e3d31SSong Liu * the *callback_ctx* should be a pointer to the stack. 50027c7e3d31SSong Liu * The *flags* is used to control certain aspects of the helper. 50037c7e3d31SSong Liu * Currently, the *flags* must be 0. 50047c7e3d31SSong Liu * 50057c7e3d31SSong Liu * The expected callback signature is 50067c7e3d31SSong Liu * 50077c7e3d31SSong Liu * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 50087c7e3d31SSong Liu * 50097c7e3d31SSong Liu * Return 50107c7e3d31SSong Liu * 0 on success. 50117c7e3d31SSong Liu * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 50127c7e3d31SSong Liu * **-EBUSY** if failed to try lock mmap_lock. 50137c7e3d31SSong Liu * **-EINVAL** for invalid **flags**. 5014e6f2dd0fSJoanne Koong * 5015e6f2dd0fSJoanne Koong * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 5016e6f2dd0fSJoanne Koong * Description 5017e6f2dd0fSJoanne Koong * For **nr_loops**, call **callback_fn** function 5018e6f2dd0fSJoanne Koong * with **callback_ctx** as the context parameter. 5019e6f2dd0fSJoanne Koong * The **callback_fn** should be a static function and 5020e6f2dd0fSJoanne Koong * the **callback_ctx** should be a pointer to the stack. 5021e6f2dd0fSJoanne Koong * The **flags** is used to control certain aspects of the helper. 5022e6f2dd0fSJoanne Koong * Currently, the **flags** must be 0. Currently, nr_loops is 5023e6f2dd0fSJoanne Koong * limited to 1 << 23 (~8 million) loops. 5024e6f2dd0fSJoanne Koong * 5025e6f2dd0fSJoanne Koong * long (\*callback_fn)(u32 index, void \*ctx); 5026e6f2dd0fSJoanne Koong * 5027e6f2dd0fSJoanne Koong * where **index** is the current index in the loop. The index 5028e6f2dd0fSJoanne Koong * is zero-indexed. 5029e6f2dd0fSJoanne Koong * 5030e6f2dd0fSJoanne Koong * If **callback_fn** returns 0, the helper will continue to the next 5031e6f2dd0fSJoanne Koong * loop. If return value is 1, the helper will skip the rest of 5032e6f2dd0fSJoanne Koong * the loops and return. Other return values are not used now, 5033e6f2dd0fSJoanne Koong * and will be rejected by the verifier. 5034e6f2dd0fSJoanne Koong * 5035e6f2dd0fSJoanne Koong * Return 5036e6f2dd0fSJoanne Koong * The number of loops performed, **-EINVAL** for invalid **flags**, 5037e6f2dd0fSJoanne Koong * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5038c5fb1993SHou Tao * 5039c5fb1993SHou Tao * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5040c5fb1993SHou Tao * Description 5041c5fb1993SHou Tao * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5042c5fb1993SHou Tao * to be null-terminated and **s1_sz** is the maximum storage 5043c5fb1993SHou Tao * size of **s1**. **s2** must be a read-only string. 5044c5fb1993SHou Tao * Return 5045c5fb1993SHou Tao * An integer less than, equal to, or greater than zero 5046c5fb1993SHou Tao * if the first **s1_sz** bytes of **s1** is found to be 5047c5fb1993SHou Tao * less than, to match, or be greater than **s2**. 5048f92c1e18SJiri Olsa * 5049f92c1e18SJiri Olsa * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5050f92c1e18SJiri Olsa * Description 5051f92c1e18SJiri Olsa * Get **n**-th argument (zero based) of the traced function (for tracing programs) 5052f92c1e18SJiri Olsa * returned in **value**. 5053f92c1e18SJiri Olsa * 5054f92c1e18SJiri Olsa * Return 5055f92c1e18SJiri Olsa * 0 on success. 5056f92c1e18SJiri Olsa * **-EINVAL** if n >= arguments count of traced function. 5057f92c1e18SJiri Olsa * 5058f92c1e18SJiri Olsa * long bpf_get_func_ret(void *ctx, u64 *value) 5059f92c1e18SJiri Olsa * Description 5060f92c1e18SJiri Olsa * Get return value of the traced function (for tracing programs) 5061f92c1e18SJiri Olsa * in **value**. 5062f92c1e18SJiri Olsa * 5063f92c1e18SJiri Olsa * Return 5064f92c1e18SJiri Olsa * 0 on success. 5065f92c1e18SJiri Olsa * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5066f92c1e18SJiri Olsa * 5067f92c1e18SJiri Olsa * long bpf_get_func_arg_cnt(void *ctx) 5068f92c1e18SJiri Olsa * Description 5069f92c1e18SJiri Olsa * Get number of arguments of the traced function (for tracing programs). 5070f92c1e18SJiri Olsa * 5071f92c1e18SJiri Olsa * Return 5072f92c1e18SJiri Olsa * The number of arguments of the traced function. 5073b44123b4SYiFei Zhu * 5074b44123b4SYiFei Zhu * int bpf_get_retval(void) 5075b44123b4SYiFei Zhu * Description 5076b44123b4SYiFei Zhu * Get the syscall's return value that will be returned to userspace. 5077b44123b4SYiFei Zhu * 5078b44123b4SYiFei Zhu * This helper is currently supported by cgroup programs only. 5079b44123b4SYiFei Zhu * Return 5080b44123b4SYiFei Zhu * The syscall's return value. 5081b44123b4SYiFei Zhu * 5082b44123b4SYiFei Zhu * int bpf_set_retval(int retval) 5083b44123b4SYiFei Zhu * Description 5084b44123b4SYiFei Zhu * Set the syscall's return value that will be returned to userspace. 5085b44123b4SYiFei Zhu * 5086b44123b4SYiFei Zhu * This helper is currently supported by cgroup programs only. 5087b44123b4SYiFei Zhu * Return 5088b44123b4SYiFei Zhu * 0 on success, or a negative error in case of failure. 50890165cc81SLorenzo Bianconi * 50900165cc81SLorenzo Bianconi * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 50910165cc81SLorenzo Bianconi * Description 50920165cc81SLorenzo Bianconi * Get the total size of a given xdp buff (linear and paged area) 50930165cc81SLorenzo Bianconi * Return 50940165cc81SLorenzo Bianconi * The total size of a given xdp buffer. 50953f364222SLorenzo Bianconi * 50963f364222SLorenzo Bianconi * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 50973f364222SLorenzo Bianconi * Description 50983f364222SLorenzo Bianconi * This helper is provided as an easy way to load data from a 50993f364222SLorenzo Bianconi * xdp buffer. It can be used to load *len* bytes from *offset* from 51003f364222SLorenzo Bianconi * the frame associated to *xdp_md*, into the buffer pointed by 51013f364222SLorenzo Bianconi * *buf*. 51023f364222SLorenzo Bianconi * Return 51033f364222SLorenzo Bianconi * 0 on success, or a negative error in case of failure. 51043f364222SLorenzo Bianconi * 51053f364222SLorenzo Bianconi * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 51063f364222SLorenzo Bianconi * Description 51073f364222SLorenzo Bianconi * Store *len* bytes from buffer *buf* into the frame 51083f364222SLorenzo Bianconi * associated to *xdp_md*, at *offset*. 51093f364222SLorenzo Bianconi * Return 51103f364222SLorenzo Bianconi * 0 on success, or a negative error in case of failure. 5111376040e4SKenny Yu * 5112376040e4SKenny Yu * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5113376040e4SKenny Yu * Description 5114376040e4SKenny Yu * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5115376040e4SKenny Yu * address space, and stores the data in *dst*. *flags* is not 5116376040e4SKenny Yu * used yet and is provided for future extensibility. This helper 5117376040e4SKenny Yu * can only be used by sleepable programs. 5118376040e4SKenny Yu * Return 5119376040e4SKenny Yu * 0 on success, or a negative error in case of failure. On error 5120376040e4SKenny Yu * *dst* buffer is zeroed out. 51218d21ec0eSMartin KaFai Lau * 51229bb984f2SMartin KaFai Lau * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 51238d21ec0eSMartin KaFai Lau * Description 51249bb984f2SMartin KaFai Lau * Change the __sk_buff->tstamp_type to *tstamp_type* 51259bb984f2SMartin KaFai Lau * and set *tstamp* to the __sk_buff->tstamp together. 51268d21ec0eSMartin KaFai Lau * 51279bb984f2SMartin KaFai Lau * If there is no need to change the __sk_buff->tstamp_type, 51289bb984f2SMartin KaFai Lau * the tstamp value can be directly written to __sk_buff->tstamp 51298d21ec0eSMartin KaFai Lau * instead. 51308d21ec0eSMartin KaFai Lau * 51319bb984f2SMartin KaFai Lau * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 51329bb984f2SMartin KaFai Lau * will be kept during bpf_redirect_*(). A non zero 51339bb984f2SMartin KaFai Lau * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 51349bb984f2SMartin KaFai Lau * *tstamp_type*. 51359bb984f2SMartin KaFai Lau * 51369bb984f2SMartin KaFai Lau * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 51379bb984f2SMartin KaFai Lau * with a zero *tstamp*. 51388d21ec0eSMartin KaFai Lau * 51398d21ec0eSMartin KaFai Lau * Only IPv4 and IPv6 skb->protocol are supported. 51408d21ec0eSMartin KaFai Lau * 51418d21ec0eSMartin KaFai Lau * This function is most useful when it needs to set a 51428d21ec0eSMartin KaFai Lau * mono delivery time to __sk_buff->tstamp and then 51438d21ec0eSMartin KaFai Lau * bpf_redirect_*() to the egress of an iface. For example, 51448d21ec0eSMartin KaFai Lau * changing the (rcv) timestamp in __sk_buff->tstamp at 51458d21ec0eSMartin KaFai Lau * ingress to a mono delivery time and then bpf_redirect_*() 51468d21ec0eSMartin KaFai Lau * to sch_fq@phy-dev. 51478d21ec0eSMartin KaFai Lau * Return 51488d21ec0eSMartin KaFai Lau * 0 on success. 51498d21ec0eSMartin KaFai Lau * **-EINVAL** for invalid input 51509bb984f2SMartin KaFai Lau * **-EOPNOTSUPP** for unsupported protocol 5151174b1694SRoberto Sassu * 5152174b1694SRoberto Sassu * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5153174b1694SRoberto Sassu * Description 5154174b1694SRoberto Sassu * Returns a calculated IMA hash of the *file*. 5155174b1694SRoberto Sassu * If the hash is larger than *size*, then only *size* 5156174b1694SRoberto Sassu * bytes will be copied to *dst* 5157174b1694SRoberto Sassu * Return 5158174b1694SRoberto Sassu * The **hash_algo** is returned on success, 5159174b1694SRoberto Sassu * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if 5160174b1694SRoberto Sassu * invalid arguments are passed. 5161c0a5a21cSKumar Kartikeya Dwivedi * 5162c0a5a21cSKumar Kartikeya Dwivedi * void *bpf_kptr_xchg(void *map_value, void *ptr) 5163c0a5a21cSKumar Kartikeya Dwivedi * Description 5164c0a5a21cSKumar Kartikeya Dwivedi * Exchange kptr at pointer *map_value* with *ptr*, and return the 5165c0a5a21cSKumar Kartikeya Dwivedi * old value. *ptr* can be NULL, otherwise it must be a referenced 5166c0a5a21cSKumar Kartikeya Dwivedi * pointer which will be released when this helper is called. 5167c0a5a21cSKumar Kartikeya Dwivedi * Return 5168c0a5a21cSKumar Kartikeya Dwivedi * The old value of kptr (which can be NULL). The returned pointer 5169c0a5a21cSKumar Kartikeya Dwivedi * if not NULL, is a reference which must be released using its 5170c0a5a21cSKumar Kartikeya Dwivedi * corresponding release function, or moved into a BPF map before 5171c0a5a21cSKumar Kartikeya Dwivedi * program exit. 517207343110SFeng Zhou * 517307343110SFeng Zhou * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu) 517407343110SFeng Zhou * Description 517507343110SFeng Zhou * Perform a lookup in *percpu map* for an entry associated to 517607343110SFeng Zhou * *key* on *cpu*. 517707343110SFeng Zhou * Return 517807343110SFeng Zhou * Map value associated to *key* on *cpu*, or **NULL** if no entry 517907343110SFeng Zhou * was found or *cpu* is invalid. 51803bc253c2SGeliang Tang * 51813bc253c2SGeliang Tang * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk) 51823bc253c2SGeliang Tang * Description 51833bc253c2SGeliang Tang * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 51843bc253c2SGeliang Tang * Return 51853bc253c2SGeliang Tang * *sk* if casting is valid, or **NULL** otherwise. 5186263ae152SJoanne Koong * 5187263ae152SJoanne Koong * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr) 5188263ae152SJoanne Koong * Description 5189263ae152SJoanne Koong * Get a dynptr to local memory *data*. 5190263ae152SJoanne Koong * 5191263ae152SJoanne Koong * *data* must be a ptr to a map value. 5192263ae152SJoanne Koong * The maximum *size* supported is DYNPTR_MAX_SIZE. 5193263ae152SJoanne Koong * *flags* is currently unused. 5194263ae152SJoanne Koong * Return 5195263ae152SJoanne Koong * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 5196263ae152SJoanne Koong * -EINVAL if flags is not 0. 5197bc34dee6SJoanne Koong * 5198bc34dee6SJoanne Koong * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr) 5199bc34dee6SJoanne Koong * Description 5200bc34dee6SJoanne Koong * Reserve *size* bytes of payload in a ring buffer *ringbuf* 5201bc34dee6SJoanne Koong * through the dynptr interface. *flags* must be 0. 5202bc34dee6SJoanne Koong * 5203bc34dee6SJoanne Koong * Please note that a corresponding bpf_ringbuf_submit_dynptr or 5204bc34dee6SJoanne Koong * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 5205bc34dee6SJoanne Koong * reservation fails. This is enforced by the verifier. 5206bc34dee6SJoanne Koong * Return 5207bc34dee6SJoanne Koong * 0 on success, or a negative error in case of failure. 5208bc34dee6SJoanne Koong * 5209bc34dee6SJoanne Koong * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags) 5210bc34dee6SJoanne Koong * Description 5211bc34dee6SJoanne Koong * Submit reserved ring buffer sample, pointed to by *data*, 5212bc34dee6SJoanne Koong * through the dynptr interface. This is a no-op if the dynptr is 5213bc34dee6SJoanne Koong * invalid/null. 5214bc34dee6SJoanne Koong * 5215bc34dee6SJoanne Koong * For more information on *flags*, please see 5216bc34dee6SJoanne Koong * 'bpf_ringbuf_submit'. 5217bc34dee6SJoanne Koong * Return 5218bc34dee6SJoanne Koong * Nothing. Always succeeds. 5219bc34dee6SJoanne Koong * 5220bc34dee6SJoanne Koong * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags) 5221bc34dee6SJoanne Koong * Description 5222bc34dee6SJoanne Koong * Discard reserved ring buffer sample through the dynptr 5223bc34dee6SJoanne Koong * interface. This is a no-op if the dynptr is invalid/null. 5224bc34dee6SJoanne Koong * 5225bc34dee6SJoanne Koong * For more information on *flags*, please see 5226bc34dee6SJoanne Koong * 'bpf_ringbuf_discard'. 5227bc34dee6SJoanne Koong * Return 5228bc34dee6SJoanne Koong * Nothing. Always succeeds. 522913bbbfbeSJoanne Koong * 5230f8d3da4eSJoanne Koong * long bpf_dynptr_read(void *dst, u32 len, struct bpf_dynptr *src, u32 offset, u64 flags) 523113bbbfbeSJoanne Koong * Description 523213bbbfbeSJoanne Koong * Read *len* bytes from *src* into *dst*, starting from *offset* 523313bbbfbeSJoanne Koong * into *src*. 5234f8d3da4eSJoanne Koong * *flags* is currently unused. 523513bbbfbeSJoanne Koong * Return 523613bbbfbeSJoanne Koong * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5237f8d3da4eSJoanne Koong * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 5238f8d3da4eSJoanne Koong * *flags* is not 0. 523913bbbfbeSJoanne Koong * 5240f8d3da4eSJoanne Koong * long bpf_dynptr_write(struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags) 524113bbbfbeSJoanne Koong * Description 524213bbbfbeSJoanne Koong * Write *len* bytes from *src* into *dst*, starting from *offset* 524313bbbfbeSJoanne Koong * into *dst*. 5244f8d3da4eSJoanne Koong * *flags* is currently unused. 524513bbbfbeSJoanne Koong * Return 524613bbbfbeSJoanne Koong * 0 on success, -E2BIG if *offset* + *len* exceeds the length 524713bbbfbeSJoanne Koong * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 5248f8d3da4eSJoanne Koong * is a read-only dynptr or if *flags* is not 0. 524934d4ef57SJoanne Koong * 525034d4ef57SJoanne Koong * void *bpf_dynptr_data(struct bpf_dynptr *ptr, u32 offset, u32 len) 525134d4ef57SJoanne Koong * Description 525234d4ef57SJoanne Koong * Get a pointer to the underlying dynptr data. 525334d4ef57SJoanne Koong * 525434d4ef57SJoanne Koong * *len* must be a statically known value. The returned data slice 525534d4ef57SJoanne Koong * is invalidated whenever the dynptr is invalidated. 525634d4ef57SJoanne Koong * Return 525734d4ef57SJoanne Koong * Pointer to the underlying dynptr data, NULL if the dynptr is 525834d4ef57SJoanne Koong * read-only, if the dynptr is invalid, or if the offset and length 525934d4ef57SJoanne Koong * is out of bounds. 526033bf9885SMaxim Mikityanskiy * 526133bf9885SMaxim Mikityanskiy * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len) 526233bf9885SMaxim Mikityanskiy * Description 526333bf9885SMaxim Mikityanskiy * Try to issue a SYN cookie for the packet with corresponding 526433bf9885SMaxim Mikityanskiy * IPv4/TCP headers, *iph* and *th*, without depending on a 526533bf9885SMaxim Mikityanskiy * listening socket. 526633bf9885SMaxim Mikityanskiy * 526733bf9885SMaxim Mikityanskiy * *iph* points to the IPv4 header. 526833bf9885SMaxim Mikityanskiy * 526933bf9885SMaxim Mikityanskiy * *th* points to the start of the TCP header, while *th_len* 527033bf9885SMaxim Mikityanskiy * contains the length of the TCP header (at least 527133bf9885SMaxim Mikityanskiy * **sizeof**\ (**struct tcphdr**)). 527233bf9885SMaxim Mikityanskiy * Return 527333bf9885SMaxim Mikityanskiy * On success, lower 32 bits hold the generated SYN cookie in 527433bf9885SMaxim Mikityanskiy * followed by 16 bits which hold the MSS value for that cookie, 527533bf9885SMaxim Mikityanskiy * and the top 16 bits are unused. 527633bf9885SMaxim Mikityanskiy * 527733bf9885SMaxim Mikityanskiy * On failure, the returned value is one of the following: 527833bf9885SMaxim Mikityanskiy * 527933bf9885SMaxim Mikityanskiy * **-EINVAL** if *th_len* is invalid. 528033bf9885SMaxim Mikityanskiy * 528133bf9885SMaxim Mikityanskiy * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len) 528233bf9885SMaxim Mikityanskiy * Description 528333bf9885SMaxim Mikityanskiy * Try to issue a SYN cookie for the packet with corresponding 528433bf9885SMaxim Mikityanskiy * IPv6/TCP headers, *iph* and *th*, without depending on a 528533bf9885SMaxim Mikityanskiy * listening socket. 528633bf9885SMaxim Mikityanskiy * 528733bf9885SMaxim Mikityanskiy * *iph* points to the IPv6 header. 528833bf9885SMaxim Mikityanskiy * 528933bf9885SMaxim Mikityanskiy * *th* points to the start of the TCP header, while *th_len* 529033bf9885SMaxim Mikityanskiy * contains the length of the TCP header (at least 529133bf9885SMaxim Mikityanskiy * **sizeof**\ (**struct tcphdr**)). 529233bf9885SMaxim Mikityanskiy * Return 529333bf9885SMaxim Mikityanskiy * On success, lower 32 bits hold the generated SYN cookie in 529433bf9885SMaxim Mikityanskiy * followed by 16 bits which hold the MSS value for that cookie, 529533bf9885SMaxim Mikityanskiy * and the top 16 bits are unused. 529633bf9885SMaxim Mikityanskiy * 529733bf9885SMaxim Mikityanskiy * On failure, the returned value is one of the following: 529833bf9885SMaxim Mikityanskiy * 529933bf9885SMaxim Mikityanskiy * **-EINVAL** if *th_len* is invalid. 530033bf9885SMaxim Mikityanskiy * 530133bf9885SMaxim Mikityanskiy * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 530233bf9885SMaxim Mikityanskiy * 530333bf9885SMaxim Mikityanskiy * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th) 530433bf9885SMaxim Mikityanskiy * Description 530533bf9885SMaxim Mikityanskiy * Check whether *iph* and *th* contain a valid SYN cookie ACK 530633bf9885SMaxim Mikityanskiy * without depending on a listening socket. 530733bf9885SMaxim Mikityanskiy * 530833bf9885SMaxim Mikityanskiy * *iph* points to the IPv4 header. 530933bf9885SMaxim Mikityanskiy * 531033bf9885SMaxim Mikityanskiy * *th* points to the TCP header. 531133bf9885SMaxim Mikityanskiy * Return 531233bf9885SMaxim Mikityanskiy * 0 if *iph* and *th* are a valid SYN cookie ACK. 531333bf9885SMaxim Mikityanskiy * 531433bf9885SMaxim Mikityanskiy * On failure, the returned value is one of the following: 531533bf9885SMaxim Mikityanskiy * 531633bf9885SMaxim Mikityanskiy * **-EACCES** if the SYN cookie is not valid. 531733bf9885SMaxim Mikityanskiy * 531833bf9885SMaxim Mikityanskiy * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th) 531933bf9885SMaxim Mikityanskiy * Description 532033bf9885SMaxim Mikityanskiy * Check whether *iph* and *th* contain a valid SYN cookie ACK 532133bf9885SMaxim Mikityanskiy * without depending on a listening socket. 532233bf9885SMaxim Mikityanskiy * 532333bf9885SMaxim Mikityanskiy * *iph* points to the IPv6 header. 532433bf9885SMaxim Mikityanskiy * 532533bf9885SMaxim Mikityanskiy * *th* points to the TCP header. 532633bf9885SMaxim Mikityanskiy * Return 532733bf9885SMaxim Mikityanskiy * 0 if *iph* and *th* are a valid SYN cookie ACK. 532833bf9885SMaxim Mikityanskiy * 532933bf9885SMaxim Mikityanskiy * On failure, the returned value is one of the following: 533033bf9885SMaxim Mikityanskiy * 533133bf9885SMaxim Mikityanskiy * **-EACCES** if the SYN cookie is not valid. 533233bf9885SMaxim Mikityanskiy * 533333bf9885SMaxim Mikityanskiy * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 53347a4b28c6SDaniel Borkmann */ 5335ebb676daSThomas Graf #define __BPF_FUNC_MAPPER(FN) \ 5336ebb676daSThomas Graf FN(unspec), \ 5337ebb676daSThomas Graf FN(map_lookup_elem), \ 5338ebb676daSThomas Graf FN(map_update_elem), \ 5339ebb676daSThomas Graf FN(map_delete_elem), \ 5340ebb676daSThomas Graf FN(probe_read), \ 5341ebb676daSThomas Graf FN(ktime_get_ns), \ 5342ebb676daSThomas Graf FN(trace_printk), \ 5343ebb676daSThomas Graf FN(get_prandom_u32), \ 5344ebb676daSThomas Graf FN(get_smp_processor_id), \ 5345ebb676daSThomas Graf FN(skb_store_bytes), \ 5346ebb676daSThomas Graf FN(l3_csum_replace), \ 5347ebb676daSThomas Graf FN(l4_csum_replace), \ 5348ebb676daSThomas Graf FN(tail_call), \ 5349ebb676daSThomas Graf FN(clone_redirect), \ 5350ebb676daSThomas Graf FN(get_current_pid_tgid), \ 5351ebb676daSThomas Graf FN(get_current_uid_gid), \ 5352ebb676daSThomas Graf FN(get_current_comm), \ 5353ebb676daSThomas Graf FN(get_cgroup_classid), \ 5354ebb676daSThomas Graf FN(skb_vlan_push), \ 5355ebb676daSThomas Graf FN(skb_vlan_pop), \ 5356ebb676daSThomas Graf FN(skb_get_tunnel_key), \ 5357ebb676daSThomas Graf FN(skb_set_tunnel_key), \ 5358ebb676daSThomas Graf FN(perf_event_read), \ 5359ebb676daSThomas Graf FN(redirect), \ 5360ebb676daSThomas Graf FN(get_route_realm), \ 5361ebb676daSThomas Graf FN(perf_event_output), \ 5362ebb676daSThomas Graf FN(skb_load_bytes), \ 5363ebb676daSThomas Graf FN(get_stackid), \ 5364ebb676daSThomas Graf FN(csum_diff), \ 5365ebb676daSThomas Graf FN(skb_get_tunnel_opt), \ 5366ebb676daSThomas Graf FN(skb_set_tunnel_opt), \ 5367ebb676daSThomas Graf FN(skb_change_proto), \ 5368ebb676daSThomas Graf FN(skb_change_type), \ 5369ebb676daSThomas Graf FN(skb_under_cgroup), \ 5370ebb676daSThomas Graf FN(get_hash_recalc), \ 5371ebb676daSThomas Graf FN(get_current_task), \ 5372ebb676daSThomas Graf FN(probe_write_user), \ 5373ebb676daSThomas Graf FN(current_task_under_cgroup), \ 5374ebb676daSThomas Graf FN(skb_change_tail), \ 5375ebb676daSThomas Graf FN(skb_pull_data), \ 5376ebb676daSThomas Graf FN(csum_update), \ 5377ebb676daSThomas Graf FN(set_hash_invalid), \ 53783a0af8fdSThomas Graf FN(get_numa_node_id), \ 537917bedab2SMartin KaFai Lau FN(skb_change_head), \ 5380a5e8c070SGianluca Borello FN(xdp_adjust_head), \ 538191b8270fSChenbo Feng FN(probe_read_str), \ 53826acc5c29SChenbo Feng FN(get_socket_cookie), \ 5383ded092cdSDaniel Borkmann FN(get_socket_uid), \ 53848c4b4c7eSLawrence Brakmo FN(set_hash), \ 53852be7e212SDaniel Borkmann FN(setsockopt), \ 538697f91a7cSJohn Fastabend FN(skb_adjust_room), \ 5387174a79ffSJohn Fastabend FN(redirect_map), \ 5388174a79ffSJohn Fastabend FN(sk_redirect_map), \ 5389174a79ffSJohn Fastabend FN(sock_map_update), \ 5390908432caSYonghong Song FN(xdp_adjust_meta), \ 53914bebdc7aSYonghong Song FN(perf_event_read_value), \ 5392cd86d1fdSLawrence Brakmo FN(perf_prog_read_value), \ 53939802d865SJosef Bacik FN(getsockopt), \ 5394b13d8807SLawrence Brakmo FN(override_return), \ 53954f738adbSJohn Fastabend FN(sock_ops_cb_flags_set), \ 53962a100317SJohn Fastabend FN(msg_redirect_map), \ 539791843d54SJohn Fastabend FN(msg_apply_bytes), \ 5398015632bbSJohn Fastabend FN(msg_cork_bytes), \ 5399d74bad4eSAndrey Ignatov FN(msg_pull_data), \ 5400b32cc5b9SNikita V. Shirokov FN(bind), \ 540112bed760SEyal Birger FN(xdp_adjust_tail), \ 5402c195651eSYonghong Song FN(skb_get_xfrm_state), \ 54034e1ec56cSDaniel Borkmann FN(get_stack), \ 540487f5fc7eSDavid Ahern FN(skb_load_bytes_relative), \ 540581110384SJohn Fastabend FN(fib_lookup), \ 540681110384SJohn Fastabend FN(sock_hash_update), \ 540781110384SJohn Fastabend FN(msg_redirect_hash), \ 5408fe94cc29SMathieu Xhonneux FN(sk_redirect_hash), \ 5409fe94cc29SMathieu Xhonneux FN(lwt_push_encap), \ 5410fe94cc29SMathieu Xhonneux FN(lwt_seg6_store_bytes), \ 5411fe94cc29SMathieu Xhonneux FN(lwt_seg6_adjust_srh), \ 5412f4364dcfSSean Young FN(lwt_seg6_action), \ 5413f4364dcfSSean Young FN(rc_repeat), \ 5414cb20b08eSDaniel Borkmann FN(rc_keydown), \ 5415bf6fa2c8SYonghong Song FN(skb_cgroup_id), \ 5416cd339431SRoman Gushchin FN(get_current_cgroup_id), \ 54172dbb9b9eSMartin KaFai Lau FN(get_local_storage), \ 541877236281SAndrey Ignatov FN(sk_select_reuseport), \ 54196acc9b43SJoe Stringer FN(skb_ancestor_cgroup_id), \ 54206acc9b43SJoe Stringer FN(sk_lookup_tcp), \ 54216acc9b43SJoe Stringer FN(sk_lookup_udp), \ 5422f1a2e44aSMauricio Vasquez B FN(sk_release), \ 5423f1a2e44aSMauricio Vasquez B FN(map_push_elem), \ 5424f1a2e44aSMauricio Vasquez B FN(map_pop_elem), \ 54256fff607eSJohn Fastabend FN(map_peek_elem), \ 54267246d8edSJohn Fastabend FN(msg_push_data), \ 542701d3240aSSean Young FN(msg_pop_data), \ 5428d83525caSAlexei Starovoitov FN(rc_pointer_rel), \ 5429d83525caSAlexei Starovoitov FN(spin_lock), \ 543046f8bc92SMartin KaFai Lau FN(spin_unlock), \ 5431655a51e5SMartin KaFai Lau FN(sk_fullsock), \ 5432f7c917baSbrakmo FN(tcp_sock), \ 5433dbafd7ddSMartin KaFai Lau FN(skb_ecn_set_ce), \ 5434edbf8c01SLorenz Bauer FN(get_listener_sock), \ 543539904084SLorenz Bauer FN(skc_lookup_tcp), \ 5436808649fbSAndrey Ignatov FN(tcp_check_syncookie), \ 54371d11b301SAndrey Ignatov FN(sysctl_get_name), \ 54384e63acdfSAndrey Ignatov FN(sysctl_get_current_value), \ 54394e63acdfSAndrey Ignatov FN(sysctl_get_new_value), \ 5440d7a4cb9bSAndrey Ignatov FN(sysctl_set_new_value), \ 5441d7a4cb9bSAndrey Ignatov FN(strtol), \ 54426ac99e8fSMartin KaFai Lau FN(strtoul), \ 54436ac99e8fSMartin KaFai Lau FN(sk_storage_get), \ 54448b401f9eSYonghong Song FN(sk_storage_delete), \ 544570d66244SPetar Penkov FN(send_signal), \ 5446a7658e1aSAlexei Starovoitov FN(tcp_gen_syncookie), \ 54476ae08ae3SDaniel Borkmann FN(skb_output), \ 54486ae08ae3SDaniel Borkmann FN(probe_read_user), \ 54496ae08ae3SDaniel Borkmann FN(probe_read_kernel), \ 54506ae08ae3SDaniel Borkmann FN(probe_read_user_str), \ 5451206057feSMartin KaFai Lau FN(probe_read_kernel_str), \ 54528482941fSYonghong Song FN(tcp_send_ack), \ 54535576b991SMartin KaFai Lau FN(send_signal_thread), \ 5454fff7b643SDaniel Xu FN(jiffies64), \ 5455b4490c5cSCarlos Neira FN(read_branch_records), \ 5456d831ee84SEelco Chaudron FN(get_ns_current_pid_tgid), \ 5457f318903cSDaniel Borkmann FN(xdp_output), \ 54580f09abd1SDaniel Borkmann FN(get_netns_cookie), \ 5459cf7fbe66SJoe Stringer FN(get_current_ancestor_cgroup_id), \ 546071d19214SMaciej Żenczykowski FN(sk_assign), \ 5461492e639fSYonghong Song FN(ktime_get_boot_ns), \ 5462492e639fSYonghong Song FN(seq_printf), \ 5463f307fa2cSAndrey Ignatov FN(seq_write), \ 5464f307fa2cSAndrey Ignatov FN(sk_cgroup_id), \ 5465457f4436SAndrii Nakryiko FN(sk_ancestor_cgroup_id), \ 5466457f4436SAndrii Nakryiko FN(ringbuf_output), \ 5467457f4436SAndrii Nakryiko FN(ringbuf_reserve), \ 5468457f4436SAndrii Nakryiko FN(ringbuf_submit), \ 5469457f4436SAndrii Nakryiko FN(ringbuf_discard), \ 54707cdec54fSDaniel Borkmann FN(ringbuf_query), \ 5471af7ec138SYonghong Song FN(csum_level), \ 5472478cfbdfSYonghong Song FN(skc_to_tcp6_sock), \ 5473478cfbdfSYonghong Song FN(skc_to_tcp_sock), \ 5474478cfbdfSYonghong Song FN(skc_to_tcp_timewait_sock), \ 54750d4fad3eSYonghong Song FN(skc_to_tcp_request_sock), \ 5476fa28dcb8SSong Liu FN(skc_to_udp6_sock), \ 5477fa28dcb8SSong Liu FN(get_task_stack), \ 54780813a841SMartin KaFai Lau FN(load_hdr_opt), \ 54790813a841SMartin KaFai Lau FN(store_hdr_opt), \ 54808ea63684SKP Singh FN(reserve_hdr_opt), \ 54818ea63684SKP Singh FN(inode_storage_get), \ 54828ea63684SKP Singh FN(inode_storage_delete), \ 54836e22ab9dSJiri Olsa FN(d_path), \ 548407be4c4aSAlexei Starovoitov FN(copy_from_user), \ 5485c4d0bfb4SAlan Maguire FN(snprintf_btf), \ 5486eb411377SAlan Maguire FN(seq_printf_btf), \ 5487b426ce83SDaniel Borkmann FN(skb_cgroup_classid), \ 5488b4ab3141SDaniel Borkmann FN(redirect_neigh), \ 5489b7906b70SAndrii Nakryiko FN(per_cpu_ptr), \ 5490b7906b70SAndrii Nakryiko FN(this_cpu_ptr), \ 54919aa1206eSDaniel Borkmann FN(redirect_peer), \ 54924cf1bc1fSKP Singh FN(task_storage_get), \ 54934cf1bc1fSKP Singh FN(task_storage_delete), \ 54943ca1032aSKP Singh FN(get_current_task_btf), \ 54953f6719c7SKP Singh FN(bprm_opts_set), \ 5496d0551261SDmitrii Banshchikov FN(ktime_get_coarse_ns), \ 549727672f0dSKP Singh FN(ima_inode_hash), \ 54984f19cab7SFlorent Revest FN(sock_from_file), \ 549934b2021cSJesper Dangaard Brouer FN(check_mtu), \ 550069c087baSYonghong Song FN(for_each_map_elem), \ 55017b15523aSFlorent Revest FN(snprintf), \ 550279a7f8bdSAlexei Starovoitov FN(sys_bpf), \ 55033d78417bSAlexei Starovoitov FN(btf_find_by_name_kind), \ 55043abea089SAlexei Starovoitov FN(sys_close), \ 5505b00628b1SAlexei Starovoitov FN(timer_init), \ 5506b00628b1SAlexei Starovoitov FN(timer_set_callback), \ 5507b00628b1SAlexei Starovoitov FN(timer_start), \ 5508b00628b1SAlexei Starovoitov FN(timer_cancel), \ 55099b99edcaSJiri Olsa FN(get_func_ip), \ 55107adfc6c9SAndrii Nakryiko FN(get_attach_cookie), \ 5511dd6e10fbSDaniel Xu FN(task_pt_regs), \ 5512856c02dbSSong Liu FN(get_branch_snapshot), \ 551310aceb62SDave Marchevsky FN(trace_vprintk), \ 55149eeb3aa3SHengqi Chen FN(skc_to_unix_sock), \ 5515d6aef08aSKumar Kartikeya Dwivedi FN(kallsyms_lookup_name), \ 55167c7e3d31SSong Liu FN(find_vma), \ 5517e6f2dd0fSJoanne Koong FN(loop), \ 5518c5fb1993SHou Tao FN(strncmp), \ 5519f92c1e18SJiri Olsa FN(get_func_arg), \ 5520f92c1e18SJiri Olsa FN(get_func_ret), \ 5521f92c1e18SJiri Olsa FN(get_func_arg_cnt), \ 5522b44123b4SYiFei Zhu FN(get_retval), \ 5523b44123b4SYiFei Zhu FN(set_retval), \ 55240165cc81SLorenzo Bianconi FN(xdp_get_buff_len), \ 55253f364222SLorenzo Bianconi FN(xdp_load_bytes), \ 55263f364222SLorenzo Bianconi FN(xdp_store_bytes), \ 5527376040e4SKenny Yu FN(copy_from_user_task), \ 55289bb984f2SMartin KaFai Lau FN(skb_set_tstamp), \ 5529174b1694SRoberto Sassu FN(ima_file_hash), \ 5530c0a5a21cSKumar Kartikeya Dwivedi FN(kptr_xchg), \ 553107343110SFeng Zhou FN(map_lookup_percpu_elem), \ 55323bc253c2SGeliang Tang FN(skc_to_mptcp_sock), \ 5533263ae152SJoanne Koong FN(dynptr_from_mem), \ 5534bc34dee6SJoanne Koong FN(ringbuf_reserve_dynptr), \ 5535bc34dee6SJoanne Koong FN(ringbuf_submit_dynptr), \ 5536bc34dee6SJoanne Koong FN(ringbuf_discard_dynptr), \ 553713bbbfbeSJoanne Koong FN(dynptr_read), \ 553813bbbfbeSJoanne Koong FN(dynptr_write), \ 553934d4ef57SJoanne Koong FN(dynptr_data), \ 554033bf9885SMaxim Mikityanskiy FN(tcp_raw_gen_syncookie_ipv4), \ 554133bf9885SMaxim Mikityanskiy FN(tcp_raw_gen_syncookie_ipv6), \ 554233bf9885SMaxim Mikityanskiy FN(tcp_raw_check_syncookie_ipv4), \ 554333bf9885SMaxim Mikityanskiy FN(tcp_raw_check_syncookie_ipv6), \ 5544fa28dcb8SSong Liu /* */ 55457a4b28c6SDaniel Borkmann 5546ebb676daSThomas Graf /* integer value in 'imm' field of BPF_CALL instruction selects which helper 5547ebb676daSThomas Graf * function eBPF program intends to call 55482d0e30c3SDaniel Borkmann */ 5549ebb676daSThomas Graf #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 5550ebb676daSThomas Graf enum bpf_func_id { 5551ebb676daSThomas Graf __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 555209756af4SAlexei Starovoitov __BPF_FUNC_MAX_ID, 555309756af4SAlexei Starovoitov }; 5554ebb676daSThomas Graf #undef __BPF_ENUM_FN 555509756af4SAlexei Starovoitov 5556781c53bcSDaniel Borkmann /* All flags used by eBPF helper functions, placed here. */ 5557781c53bcSDaniel Borkmann 5558781c53bcSDaniel Borkmann /* BPF_FUNC_skb_store_bytes flags. */ 55591aae4bddSAndrii Nakryiko enum { 55601aae4bddSAndrii Nakryiko BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 55611aae4bddSAndrii Nakryiko BPF_F_INVALIDATE_HASH = (1ULL << 1), 55621aae4bddSAndrii Nakryiko }; 5563781c53bcSDaniel Borkmann 5564781c53bcSDaniel Borkmann /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 5565781c53bcSDaniel Borkmann * First 4 bits are for passing the header field size. 5566781c53bcSDaniel Borkmann */ 55671aae4bddSAndrii Nakryiko enum { 55681aae4bddSAndrii Nakryiko BPF_F_HDR_FIELD_MASK = 0xfULL, 55691aae4bddSAndrii Nakryiko }; 5570781c53bcSDaniel Borkmann 5571781c53bcSDaniel Borkmann /* BPF_FUNC_l4_csum_replace flags. */ 55721aae4bddSAndrii Nakryiko enum { 55731aae4bddSAndrii Nakryiko BPF_F_PSEUDO_HDR = (1ULL << 4), 55741aae4bddSAndrii Nakryiko BPF_F_MARK_MANGLED_0 = (1ULL << 5), 55751aae4bddSAndrii Nakryiko BPF_F_MARK_ENFORCE = (1ULL << 6), 55761aae4bddSAndrii Nakryiko }; 5577781c53bcSDaniel Borkmann 5578781c53bcSDaniel Borkmann /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 55791aae4bddSAndrii Nakryiko enum { 55801aae4bddSAndrii Nakryiko BPF_F_INGRESS = (1ULL << 0), 55811aae4bddSAndrii Nakryiko }; 5582781c53bcSDaniel Borkmann 5583c6c33454SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 55841aae4bddSAndrii Nakryiko enum { 55851aae4bddSAndrii Nakryiko BPF_F_TUNINFO_IPV6 = (1ULL << 0), 55861aae4bddSAndrii Nakryiko }; 5587c6c33454SDaniel Borkmann 5588c195651eSYonghong Song /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 55891aae4bddSAndrii Nakryiko enum { 55901aae4bddSAndrii Nakryiko BPF_F_SKIP_FIELD_MASK = 0xffULL, 55911aae4bddSAndrii Nakryiko BPF_F_USER_STACK = (1ULL << 8), 5592c195651eSYonghong Song /* flags used by BPF_FUNC_get_stackid only. */ 55931aae4bddSAndrii Nakryiko BPF_F_FAST_STACK_CMP = (1ULL << 9), 55941aae4bddSAndrii Nakryiko BPF_F_REUSE_STACKID = (1ULL << 10), 5595c195651eSYonghong Song /* flags used by BPF_FUNC_get_stack only. */ 55961aae4bddSAndrii Nakryiko BPF_F_USER_BUILD_ID = (1ULL << 11), 55971aae4bddSAndrii Nakryiko }; 5598d5a3b1f6SAlexei Starovoitov 55992da897e5SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key flags. */ 56001aae4bddSAndrii Nakryiko enum { 56011aae4bddSAndrii Nakryiko BPF_F_ZERO_CSUM_TX = (1ULL << 1), 56021aae4bddSAndrii Nakryiko BPF_F_DONT_FRAGMENT = (1ULL << 2), 56031aae4bddSAndrii Nakryiko BPF_F_SEQ_NUMBER = (1ULL << 3), 56041aae4bddSAndrii Nakryiko }; 56052da897e5SDaniel Borkmann 5606908432caSYonghong Song /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 5607908432caSYonghong Song * BPF_FUNC_perf_event_read_value flags. 5608908432caSYonghong Song */ 56091aae4bddSAndrii Nakryiko enum { 56101aae4bddSAndrii Nakryiko BPF_F_INDEX_MASK = 0xffffffffULL, 56111aae4bddSAndrii Nakryiko BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 5612555c8a86SDaniel Borkmann /* BPF_FUNC_perf_event_output for sk_buff input context. */ 56131aae4bddSAndrii Nakryiko BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 56141aae4bddSAndrii Nakryiko }; 56151e33759cSDaniel Borkmann 5616f71c6143SJoe Stringer /* Current network namespace */ 56171aae4bddSAndrii Nakryiko enum { 56181aae4bddSAndrii Nakryiko BPF_F_CURRENT_NETNS = (-1L), 56191aae4bddSAndrii Nakryiko }; 5620f71c6143SJoe Stringer 56217cdec54fSDaniel Borkmann /* BPF_FUNC_csum_level level values. */ 56227cdec54fSDaniel Borkmann enum { 56237cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_QUERY, 56247cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_INC, 56257cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_DEC, 56267cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_RESET, 56277cdec54fSDaniel Borkmann }; 56287cdec54fSDaniel Borkmann 56292278f6ccSWillem de Bruijn /* BPF_FUNC_skb_adjust_room flags. */ 56301aae4bddSAndrii Nakryiko enum { 56311aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 56321aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 56331aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 56341aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 56351aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 5636836e66c2SDaniel Borkmann BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 5637d01b59c9SXuesen Huang BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 56381aae4bddSAndrii Nakryiko }; 56392278f6ccSWillem de Bruijn 56401aae4bddSAndrii Nakryiko enum { 56411aae4bddSAndrii Nakryiko BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 56421aae4bddSAndrii Nakryiko BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 56431aae4bddSAndrii Nakryiko }; 564458dfc900SAlan Maguire 564558dfc900SAlan Maguire #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 564658dfc900SAlan Maguire BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 564758dfc900SAlan Maguire << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 5648868d5235SWillem de Bruijn 5649808649fbSAndrey Ignatov /* BPF_FUNC_sysctl_get_name flags. */ 56501aae4bddSAndrii Nakryiko enum { 56511aae4bddSAndrii Nakryiko BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 56521aae4bddSAndrii Nakryiko }; 5653808649fbSAndrey Ignatov 5654f836a56eSKP Singh /* BPF_FUNC_<kernel_obj>_storage_get flags */ 56551aae4bddSAndrii Nakryiko enum { 5656f836a56eSKP Singh BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 5657f836a56eSKP Singh /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 5658f836a56eSKP Singh * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 5659f836a56eSKP Singh */ 5660f836a56eSKP Singh BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 56611aae4bddSAndrii Nakryiko }; 56626ac99e8fSMartin KaFai Lau 5663fff7b643SDaniel Xu /* BPF_FUNC_read_branch_records flags. */ 56641aae4bddSAndrii Nakryiko enum { 56651aae4bddSAndrii Nakryiko BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 56661aae4bddSAndrii Nakryiko }; 5667fff7b643SDaniel Xu 5668457f4436SAndrii Nakryiko /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 5669457f4436SAndrii Nakryiko * BPF_FUNC_bpf_ringbuf_output flags. 5670457f4436SAndrii Nakryiko */ 5671457f4436SAndrii Nakryiko enum { 5672457f4436SAndrii Nakryiko BPF_RB_NO_WAKEUP = (1ULL << 0), 5673457f4436SAndrii Nakryiko BPF_RB_FORCE_WAKEUP = (1ULL << 1), 5674457f4436SAndrii Nakryiko }; 5675457f4436SAndrii Nakryiko 5676457f4436SAndrii Nakryiko /* BPF_FUNC_bpf_ringbuf_query flags */ 5677457f4436SAndrii Nakryiko enum { 5678457f4436SAndrii Nakryiko BPF_RB_AVAIL_DATA = 0, 5679457f4436SAndrii Nakryiko BPF_RB_RING_SIZE = 1, 5680457f4436SAndrii Nakryiko BPF_RB_CONS_POS = 2, 5681457f4436SAndrii Nakryiko BPF_RB_PROD_POS = 3, 5682457f4436SAndrii Nakryiko }; 5683457f4436SAndrii Nakryiko 5684457f4436SAndrii Nakryiko /* BPF ring buffer constants */ 5685457f4436SAndrii Nakryiko enum { 5686457f4436SAndrii Nakryiko BPF_RINGBUF_BUSY_BIT = (1U << 31), 5687457f4436SAndrii Nakryiko BPF_RINGBUF_DISCARD_BIT = (1U << 30), 5688457f4436SAndrii Nakryiko BPF_RINGBUF_HDR_SZ = 8, 5689457f4436SAndrii Nakryiko }; 5690457f4436SAndrii Nakryiko 5691e9ddbb77SJakub Sitnicki /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 5692e9ddbb77SJakub Sitnicki enum { 5693e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 5694e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 5695e9ddbb77SJakub Sitnicki }; 5696e9ddbb77SJakub Sitnicki 56972be7e212SDaniel Borkmann /* Mode for BPF_FUNC_skb_adjust_room helper. */ 56982be7e212SDaniel Borkmann enum bpf_adj_room_mode { 56992be7e212SDaniel Borkmann BPF_ADJ_ROOM_NET, 570014aa3192SWillem de Bruijn BPF_ADJ_ROOM_MAC, 57012be7e212SDaniel Borkmann }; 57022be7e212SDaniel Borkmann 57034e1ec56cSDaniel Borkmann /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 57044e1ec56cSDaniel Borkmann enum bpf_hdr_start_off { 57054e1ec56cSDaniel Borkmann BPF_HDR_START_MAC, 57064e1ec56cSDaniel Borkmann BPF_HDR_START_NET, 57074e1ec56cSDaniel Borkmann }; 57084e1ec56cSDaniel Borkmann 5709fe94cc29SMathieu Xhonneux /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 5710fe94cc29SMathieu Xhonneux enum bpf_lwt_encap_mode { 5711fe94cc29SMathieu Xhonneux BPF_LWT_ENCAP_SEG6, 57123e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_SEG6_INLINE, 57133e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_IP, 5714fe94cc29SMathieu Xhonneux }; 5715fe94cc29SMathieu Xhonneux 57163f6719c7SKP Singh /* Flags for bpf_bprm_opts_set helper */ 57173f6719c7SKP Singh enum { 57183f6719c7SKP Singh BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 57193f6719c7SKP Singh }; 57203f6719c7SKP Singh 5721e624d4edSHangbin Liu /* Flags for bpf_redirect_map helper */ 5722e624d4edSHangbin Liu enum { 5723e624d4edSHangbin Liu BPF_F_BROADCAST = (1ULL << 3), 5724e624d4edSHangbin Liu BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 5725e624d4edSHangbin Liu }; 5726e624d4edSHangbin Liu 5727b7df9adaSDaniel Borkmann #define __bpf_md_ptr(type, name) \ 5728b7df9adaSDaniel Borkmann union { \ 5729b7df9adaSDaniel Borkmann type name; \ 5730b7df9adaSDaniel Borkmann __u64 :64; \ 5731b7df9adaSDaniel Borkmann } __attribute__((aligned(8))) 5732b7df9adaSDaniel Borkmann 57338d21ec0eSMartin KaFai Lau enum { 57349bb984f2SMartin KaFai Lau BPF_SKB_TSTAMP_UNSPEC, 57359bb984f2SMartin KaFai Lau BPF_SKB_TSTAMP_DELIVERY_MONO, /* tstamp has mono delivery time */ 57369bb984f2SMartin KaFai Lau /* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle, 57379bb984f2SMartin KaFai Lau * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC 57389bb984f2SMartin KaFai Lau * and try to deduce it by ingress, egress or skb->sk->sk_clockid. 57399bb984f2SMartin KaFai Lau */ 57408d21ec0eSMartin KaFai Lau }; 57418d21ec0eSMartin KaFai Lau 57429bac3d6dSAlexei Starovoitov /* user accessible mirror of in-kernel sk_buff. 57439bac3d6dSAlexei Starovoitov * new fields can only be added to the end of this structure 57449bac3d6dSAlexei Starovoitov */ 57459bac3d6dSAlexei Starovoitov struct __sk_buff { 57469bac3d6dSAlexei Starovoitov __u32 len; 57479bac3d6dSAlexei Starovoitov __u32 pkt_type; 57489bac3d6dSAlexei Starovoitov __u32 mark; 57499bac3d6dSAlexei Starovoitov __u32 queue_mapping; 5750c2497395SAlexei Starovoitov __u32 protocol; 5751c2497395SAlexei Starovoitov __u32 vlan_present; 5752c2497395SAlexei Starovoitov __u32 vlan_tci; 575327cd5452SMichal Sekletar __u32 vlan_proto; 5754bcad5718SDaniel Borkmann __u32 priority; 575537e82c2fSAlexei Starovoitov __u32 ingress_ifindex; 575637e82c2fSAlexei Starovoitov __u32 ifindex; 5757d691f9e8SAlexei Starovoitov __u32 tc_index; 5758d691f9e8SAlexei Starovoitov __u32 cb[5]; 5759ba7591d8SDaniel Borkmann __u32 hash; 5760045efa82SDaniel Borkmann __u32 tc_classid; 5761969bf05eSAlexei Starovoitov __u32 data; 5762969bf05eSAlexei Starovoitov __u32 data_end; 5763b1d9fc41SDaniel Borkmann __u32 napi_id; 57648a31db56SJohn Fastabend 5765de8f3a83SDaniel Borkmann /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 57668a31db56SJohn Fastabend __u32 family; 57678a31db56SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 57688a31db56SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 57698a31db56SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 57708a31db56SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 57718a31db56SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 57728a31db56SJohn Fastabend __u32 local_port; /* stored in host byte order */ 5773de8f3a83SDaniel Borkmann /* ... here. */ 5774de8f3a83SDaniel Borkmann 5775de8f3a83SDaniel Borkmann __u32 data_meta; 5776b7df9adaSDaniel Borkmann __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 5777f11216b2SVlad Dumitrescu __u64 tstamp; 5778e3da08d0SPetar Penkov __u32 wire_len; 5779d9ff286aSEric Dumazet __u32 gso_segs; 578046f8bc92SMartin KaFai Lau __bpf_md_ptr(struct bpf_sock *, sk); 5781cf62089bSWillem de Bruijn __u32 gso_size; 57829bb984f2SMartin KaFai Lau __u8 tstamp_type; 57838d21ec0eSMartin KaFai Lau __u32 :24; /* Padding, future use. */ 5784f64c4aceSVadim Fedorenko __u64 hwtstamp; 57859bac3d6dSAlexei Starovoitov }; 57869bac3d6dSAlexei Starovoitov 5787d3aa45ceSAlexei Starovoitov struct bpf_tunnel_key { 5788d3aa45ceSAlexei Starovoitov __u32 tunnel_id; 5789c6c33454SDaniel Borkmann union { 5790d3aa45ceSAlexei Starovoitov __u32 remote_ipv4; 5791c6c33454SDaniel Borkmann __u32 remote_ipv6[4]; 5792c6c33454SDaniel Borkmann }; 5793c6c33454SDaniel Borkmann __u8 tunnel_tos; 5794c6c33454SDaniel Borkmann __u8 tunnel_ttl; 57951fbc2e0cSDaniel Borkmann __u16 tunnel_ext; /* Padding, future use. */ 57964018ab18SDaniel Borkmann __u32 tunnel_label; 579726101f5aSKaixi Fan union { 579826101f5aSKaixi Fan __u32 local_ipv4; 579926101f5aSKaixi Fan __u32 local_ipv6[4]; 580026101f5aSKaixi Fan }; 5801d3aa45ceSAlexei Starovoitov }; 5802d3aa45ceSAlexei Starovoitov 580312bed760SEyal Birger /* user accessible mirror of in-kernel xfrm_state. 580412bed760SEyal Birger * new fields can only be added to the end of this structure 580512bed760SEyal Birger */ 580612bed760SEyal Birger struct bpf_xfrm_state { 580712bed760SEyal Birger __u32 reqid; 580812bed760SEyal Birger __u32 spi; /* Stored in network byte order */ 580912bed760SEyal Birger __u16 family; 58101fbc2e0cSDaniel Borkmann __u16 ext; /* Padding, future use. */ 581112bed760SEyal Birger union { 581212bed760SEyal Birger __u32 remote_ipv4; /* Stored in network byte order */ 581312bed760SEyal Birger __u32 remote_ipv6[4]; /* Stored in network byte order */ 581412bed760SEyal Birger }; 581512bed760SEyal Birger }; 581612bed760SEyal Birger 58173a0af8fdSThomas Graf /* Generic BPF return codes which all BPF program types may support. 58183a0af8fdSThomas Graf * The values are binary compatible with their TC_ACT_* counter-part to 58193a0af8fdSThomas Graf * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 58203a0af8fdSThomas Graf * programs. 58213a0af8fdSThomas Graf * 58223a0af8fdSThomas Graf * XDP is handled seprately, see XDP_*. 58233a0af8fdSThomas Graf */ 58243a0af8fdSThomas Graf enum bpf_ret_code { 58253a0af8fdSThomas Graf BPF_OK = 0, 58263a0af8fdSThomas Graf /* 1 reserved */ 58273a0af8fdSThomas Graf BPF_DROP = 2, 58283a0af8fdSThomas Graf /* 3-6 reserved */ 58293a0af8fdSThomas Graf BPF_REDIRECT = 7, 58303e0bd37cSPeter Oskolkov /* >127 are reserved for prog type specific return codes. 58313e0bd37cSPeter Oskolkov * 58323e0bd37cSPeter Oskolkov * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 58333e0bd37cSPeter Oskolkov * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 58343e0bd37cSPeter Oskolkov * changed and should be routed based on its new L3 header. 58353e0bd37cSPeter Oskolkov * (This is an L3 redirect, as opposed to L2 redirect 58363e0bd37cSPeter Oskolkov * represented by BPF_REDIRECT above). 58373e0bd37cSPeter Oskolkov */ 58383e0bd37cSPeter Oskolkov BPF_LWT_REROUTE = 128, 58393a0af8fdSThomas Graf }; 58403a0af8fdSThomas Graf 584161023658SDavid Ahern struct bpf_sock { 584261023658SDavid Ahern __u32 bound_dev_if; 5843aa4c1037SDavid Ahern __u32 family; 5844aa4c1037SDavid Ahern __u32 type; 5845aa4c1037SDavid Ahern __u32 protocol; 5846482dca93SDavid Ahern __u32 mark; 5847482dca93SDavid Ahern __u32 priority; 5848aa65d696SMartin KaFai Lau /* IP address also allows 1 and 2 bytes access */ 5849aa65d696SMartin KaFai Lau __u32 src_ip4; 5850aa65d696SMartin KaFai Lau __u32 src_ip6[4]; 5851aa65d696SMartin KaFai Lau __u32 src_port; /* host byte order */ 58524421a582SJakub Sitnicki __be16 dst_port; /* network byte order */ 58534421a582SJakub Sitnicki __u16 :16; /* zero padding */ 5854aa65d696SMartin KaFai Lau __u32 dst_ip4; 5855aa65d696SMartin KaFai Lau __u32 dst_ip6[4]; 5856aa65d696SMartin KaFai Lau __u32 state; 5857c3c16f2eSAmritha Nambiar __s32 rx_queue_mapping; 585861023658SDavid Ahern }; 585961023658SDavid Ahern 5860655a51e5SMartin KaFai Lau struct bpf_tcp_sock { 5861655a51e5SMartin KaFai Lau __u32 snd_cwnd; /* Sending congestion window */ 5862655a51e5SMartin KaFai Lau __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 5863655a51e5SMartin KaFai Lau __u32 rtt_min; 5864655a51e5SMartin KaFai Lau __u32 snd_ssthresh; /* Slow start size threshold */ 5865655a51e5SMartin KaFai Lau __u32 rcv_nxt; /* What we want to receive next */ 5866655a51e5SMartin KaFai Lau __u32 snd_nxt; /* Next sequence we send */ 5867655a51e5SMartin KaFai Lau __u32 snd_una; /* First byte we want an ack for */ 5868655a51e5SMartin KaFai Lau __u32 mss_cache; /* Cached effective mss, not including SACKS */ 5869655a51e5SMartin KaFai Lau __u32 ecn_flags; /* ECN status bits. */ 5870655a51e5SMartin KaFai Lau __u32 rate_delivered; /* saved rate sample: packets delivered */ 5871655a51e5SMartin KaFai Lau __u32 rate_interval_us; /* saved rate sample: time elapsed */ 5872655a51e5SMartin KaFai Lau __u32 packets_out; /* Packets which are "in flight" */ 5873655a51e5SMartin KaFai Lau __u32 retrans_out; /* Retransmitted packets out */ 5874655a51e5SMartin KaFai Lau __u32 total_retrans; /* Total retransmits for entire connection */ 5875655a51e5SMartin KaFai Lau __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 5876655a51e5SMartin KaFai Lau * total number of segments in. 5877655a51e5SMartin KaFai Lau */ 5878655a51e5SMartin KaFai Lau __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 5879655a51e5SMartin KaFai Lau * total number of data segments in. 5880655a51e5SMartin KaFai Lau */ 5881655a51e5SMartin KaFai Lau __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 5882655a51e5SMartin KaFai Lau * The total number of segments sent. 5883655a51e5SMartin KaFai Lau */ 5884655a51e5SMartin KaFai Lau __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 5885655a51e5SMartin KaFai Lau * total number of data segments sent. 5886655a51e5SMartin KaFai Lau */ 5887655a51e5SMartin KaFai Lau __u32 lost_out; /* Lost packets */ 5888655a51e5SMartin KaFai Lau __u32 sacked_out; /* SACK'd packets */ 5889655a51e5SMartin KaFai Lau __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 5890655a51e5SMartin KaFai Lau * sum(delta(rcv_nxt)), or how many bytes 5891655a51e5SMartin KaFai Lau * were acked. 5892655a51e5SMartin KaFai Lau */ 5893655a51e5SMartin KaFai Lau __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 5894655a51e5SMartin KaFai Lau * sum(delta(snd_una)), or how many bytes 5895655a51e5SMartin KaFai Lau * were acked. 5896655a51e5SMartin KaFai Lau */ 58970357746dSStanislav Fomichev __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 58980357746dSStanislav Fomichev * total number of DSACK blocks received 58990357746dSStanislav Fomichev */ 59000357746dSStanislav Fomichev __u32 delivered; /* Total data packets delivered incl. rexmits */ 59010357746dSStanislav Fomichev __u32 delivered_ce; /* Like the above but only ECE marked packets */ 5902c2cb5e82SStanislav Fomichev __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 5903655a51e5SMartin KaFai Lau }; 5904655a51e5SMartin KaFai Lau 59056acc9b43SJoe Stringer struct bpf_sock_tuple { 59066acc9b43SJoe Stringer union { 59076acc9b43SJoe Stringer struct { 59086acc9b43SJoe Stringer __be32 saddr; 59096acc9b43SJoe Stringer __be32 daddr; 59106acc9b43SJoe Stringer __be16 sport; 59116acc9b43SJoe Stringer __be16 dport; 59126acc9b43SJoe Stringer } ipv4; 59136acc9b43SJoe Stringer struct { 59146acc9b43SJoe Stringer __be32 saddr[4]; 59156acc9b43SJoe Stringer __be32 daddr[4]; 59166acc9b43SJoe Stringer __be16 sport; 59176acc9b43SJoe Stringer __be16 dport; 59186acc9b43SJoe Stringer } ipv6; 59196acc9b43SJoe Stringer }; 59206acc9b43SJoe Stringer }; 59216acc9b43SJoe Stringer 5922fada7fdcSJonathan Lemon struct bpf_xdp_sock { 5923fada7fdcSJonathan Lemon __u32 queue_id; 5924fada7fdcSJonathan Lemon }; 5925fada7fdcSJonathan Lemon 592617bedab2SMartin KaFai Lau #define XDP_PACKET_HEADROOM 256 592717bedab2SMartin KaFai Lau 59286a773a15SBrenden Blanco /* User return codes for XDP prog type. 59296a773a15SBrenden Blanco * A valid XDP program must return one of these defined values. All other 59309beb8bedSDaniel Borkmann * return codes are reserved for future use. Unknown return codes will 59319beb8bedSDaniel Borkmann * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 59326a773a15SBrenden Blanco */ 59336a773a15SBrenden Blanco enum xdp_action { 59346a773a15SBrenden Blanco XDP_ABORTED = 0, 59356a773a15SBrenden Blanco XDP_DROP, 59366a773a15SBrenden Blanco XDP_PASS, 59376ce96ca3SBrenden Blanco XDP_TX, 5938814abfabSJohn Fastabend XDP_REDIRECT, 59396a773a15SBrenden Blanco }; 59406a773a15SBrenden Blanco 59416a773a15SBrenden Blanco /* user accessible metadata for XDP packet hook 59426a773a15SBrenden Blanco * new fields must be added to the end of this structure 59436a773a15SBrenden Blanco */ 59446a773a15SBrenden Blanco struct xdp_md { 59456a773a15SBrenden Blanco __u32 data; 59466a773a15SBrenden Blanco __u32 data_end; 5947de8f3a83SDaniel Borkmann __u32 data_meta; 5948daaf24c6SJesper Dangaard Brouer /* Below access go through struct xdp_rxq_info */ 594902dd3291SJesper Dangaard Brouer __u32 ingress_ifindex; /* rxq->dev->ifindex */ 595002dd3291SJesper Dangaard Brouer __u32 rx_queue_index; /* rxq->queue_index */ 595164b59025SDavid Ahern 595264b59025SDavid Ahern __u32 egress_ifindex; /* txq->dev->ifindex */ 59536a773a15SBrenden Blanco }; 59546a773a15SBrenden Blanco 5955281920b7SJesper Dangaard Brouer /* DEVMAP map-value layout 5956281920b7SJesper Dangaard Brouer * 5957281920b7SJesper Dangaard Brouer * The struct data-layout of map-value is a configuration interface. 5958281920b7SJesper Dangaard Brouer * New members can only be added to the end of this structure. 5959281920b7SJesper Dangaard Brouer */ 5960281920b7SJesper Dangaard Brouer struct bpf_devmap_val { 5961281920b7SJesper Dangaard Brouer __u32 ifindex; /* device index */ 5962281920b7SJesper Dangaard Brouer union { 5963281920b7SJesper Dangaard Brouer int fd; /* prog fd on map write */ 5964281920b7SJesper Dangaard Brouer __u32 id; /* prog id on map read */ 5965281920b7SJesper Dangaard Brouer } bpf_prog; 5966281920b7SJesper Dangaard Brouer }; 5967281920b7SJesper Dangaard Brouer 5968644bfe51SLorenzo Bianconi /* CPUMAP map-value layout 5969644bfe51SLorenzo Bianconi * 5970644bfe51SLorenzo Bianconi * The struct data-layout of map-value is a configuration interface. 5971644bfe51SLorenzo Bianconi * New members can only be added to the end of this structure. 5972644bfe51SLorenzo Bianconi */ 5973644bfe51SLorenzo Bianconi struct bpf_cpumap_val { 5974644bfe51SLorenzo Bianconi __u32 qsize; /* queue size to remote target CPU */ 597592164774SLorenzo Bianconi union { 597692164774SLorenzo Bianconi int fd; /* prog fd on map write */ 597792164774SLorenzo Bianconi __u32 id; /* prog id on map read */ 597892164774SLorenzo Bianconi } bpf_prog; 5979644bfe51SLorenzo Bianconi }; 5980644bfe51SLorenzo Bianconi 5981174a79ffSJohn Fastabend enum sk_action { 5982bfa64075SJohn Fastabend SK_DROP = 0, 5983bfa64075SJohn Fastabend SK_PASS, 5984174a79ffSJohn Fastabend }; 5985174a79ffSJohn Fastabend 59864f738adbSJohn Fastabend /* user accessible metadata for SK_MSG packet hook, new fields must 59874f738adbSJohn Fastabend * be added to the end of this structure 59884f738adbSJohn Fastabend */ 59894f738adbSJohn Fastabend struct sk_msg_md { 5990b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 5991b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 5992303def35SJohn Fastabend 5993303def35SJohn Fastabend __u32 family; 5994303def35SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 5995303def35SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 5996303def35SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 5997303def35SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 5998303def35SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 5999303def35SJohn Fastabend __u32 local_port; /* stored in host byte order */ 60003bdbd022SJohn Fastabend __u32 size; /* Total size of sk_msg */ 600113d70f5aSJohn Fastabend 600213d70f5aSJohn Fastabend __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 60034f738adbSJohn Fastabend }; 60044f738adbSJohn Fastabend 60052dbb9b9eSMartin KaFai Lau struct sk_reuseport_md { 60062dbb9b9eSMartin KaFai Lau /* 60072dbb9b9eSMartin KaFai Lau * Start of directly accessible data. It begins from 60082dbb9b9eSMartin KaFai Lau * the tcp/udp header. 60092dbb9b9eSMartin KaFai Lau */ 6010b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 6011b7df9adaSDaniel Borkmann /* End of directly accessible data */ 6012b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 60132dbb9b9eSMartin KaFai Lau /* 60142dbb9b9eSMartin KaFai Lau * Total length of packet (starting from the tcp/udp header). 60152dbb9b9eSMartin KaFai Lau * Note that the directly accessible bytes (data_end - data) 60162dbb9b9eSMartin KaFai Lau * could be less than this "len". Those bytes could be 60172dbb9b9eSMartin KaFai Lau * indirectly read by a helper "bpf_skb_load_bytes()". 60182dbb9b9eSMartin KaFai Lau */ 60192dbb9b9eSMartin KaFai Lau __u32 len; 60202dbb9b9eSMartin KaFai Lau /* 60212dbb9b9eSMartin KaFai Lau * Eth protocol in the mac header (network byte order). e.g. 60222dbb9b9eSMartin KaFai Lau * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 60232dbb9b9eSMartin KaFai Lau */ 60242dbb9b9eSMartin KaFai Lau __u32 eth_protocol; 60252dbb9b9eSMartin KaFai Lau __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 60262dbb9b9eSMartin KaFai Lau __u32 bind_inany; /* Is sock bound to an INANY address? */ 60272dbb9b9eSMartin KaFai Lau __u32 hash; /* A hash of the packet 4 tuples */ 6028d5e4ddaeSKuniyuki Iwashima /* When reuse->migrating_sk is NULL, it is selecting a sk for the 6029d5e4ddaeSKuniyuki Iwashima * new incoming connection request (e.g. selecting a listen sk for 6030d5e4ddaeSKuniyuki Iwashima * the received SYN in the TCP case). reuse->sk is one of the sk 6031d5e4ddaeSKuniyuki Iwashima * in the reuseport group. The bpf prog can use reuse->sk to learn 6032d5e4ddaeSKuniyuki Iwashima * the local listening ip/port without looking into the skb. 6033d5e4ddaeSKuniyuki Iwashima * 6034d5e4ddaeSKuniyuki Iwashima * When reuse->migrating_sk is not NULL, reuse->sk is closed and 6035d5e4ddaeSKuniyuki Iwashima * reuse->migrating_sk is the socket that needs to be migrated 6036d5e4ddaeSKuniyuki Iwashima * to another listening socket. migrating_sk could be a fullsock 6037d5e4ddaeSKuniyuki Iwashima * sk that is fully established or a reqsk that is in-the-middle 6038d5e4ddaeSKuniyuki Iwashima * of 3-way handshake. 6039d5e4ddaeSKuniyuki Iwashima */ 6040e0610476SKuniyuki Iwashima __bpf_md_ptr(struct bpf_sock *, sk); 6041d5e4ddaeSKuniyuki Iwashima __bpf_md_ptr(struct bpf_sock *, migrating_sk); 60422dbb9b9eSMartin KaFai Lau }; 60432dbb9b9eSMartin KaFai Lau 60441e270976SMartin KaFai Lau #define BPF_TAG_SIZE 8 60451e270976SMartin KaFai Lau 60461e270976SMartin KaFai Lau struct bpf_prog_info { 60471e270976SMartin KaFai Lau __u32 type; 60481e270976SMartin KaFai Lau __u32 id; 60491e270976SMartin KaFai Lau __u8 tag[BPF_TAG_SIZE]; 60501e270976SMartin KaFai Lau __u32 jited_prog_len; 60511e270976SMartin KaFai Lau __u32 xlated_prog_len; 60521e270976SMartin KaFai Lau __aligned_u64 jited_prog_insns; 60531e270976SMartin KaFai Lau __aligned_u64 xlated_prog_insns; 6054cb4d2b3fSMartin KaFai Lau __u64 load_time; /* ns since boottime */ 6055cb4d2b3fSMartin KaFai Lau __u32 created_by_uid; 6056cb4d2b3fSMartin KaFai Lau __u32 nr_map_ids; 6057cb4d2b3fSMartin KaFai Lau __aligned_u64 map_ids; 6058067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 6059675fc275SJakub Kicinski __u32 ifindex; 6060b85fab0eSJiri Olsa __u32 gpl_compatible:1; 60610472301aSBaruch Siach __u32 :31; /* alignment pad */ 6062675fc275SJakub Kicinski __u64 netns_dev; 6063675fc275SJakub Kicinski __u64 netns_ino; 6064dbecd738SSandipan Das __u32 nr_jited_ksyms; 6065815581c1SSandipan Das __u32 nr_jited_func_lens; 6066dbecd738SSandipan Das __aligned_u64 jited_ksyms; 6067815581c1SSandipan Das __aligned_u64 jited_func_lens; 6068838e9690SYonghong Song __u32 btf_id; 6069838e9690SYonghong Song __u32 func_info_rec_size; 6070838e9690SYonghong Song __aligned_u64 func_info; 607111d8b82dSYonghong Song __u32 nr_func_info; 607211d8b82dSYonghong Song __u32 nr_line_info; 6073c454a46bSMartin KaFai Lau __aligned_u64 line_info; 6074c454a46bSMartin KaFai Lau __aligned_u64 jited_line_info; 607511d8b82dSYonghong Song __u32 nr_jited_line_info; 6076c454a46bSMartin KaFai Lau __u32 line_info_rec_size; 6077c454a46bSMartin KaFai Lau __u32 jited_line_info_rec_size; 6078c872bdb3SSong Liu __u32 nr_prog_tags; 6079c872bdb3SSong Liu __aligned_u64 prog_tags; 60805f8f8b93SAlexei Starovoitov __u64 run_time_ns; 60815f8f8b93SAlexei Starovoitov __u64 run_cnt; 60829ed9e9baSAlexei Starovoitov __u64 recursion_misses; 6083aba64c7dSDave Marchevsky __u32 verified_insns; 6084b79c9fc9SStanislav Fomichev __u32 attach_btf_obj_id; 6085b79c9fc9SStanislav Fomichev __u32 attach_btf_id; 60861e270976SMartin KaFai Lau } __attribute__((aligned(8))); 60871e270976SMartin KaFai Lau 60881e270976SMartin KaFai Lau struct bpf_map_info { 60891e270976SMartin KaFai Lau __u32 type; 60901e270976SMartin KaFai Lau __u32 id; 60911e270976SMartin KaFai Lau __u32 key_size; 60921e270976SMartin KaFai Lau __u32 value_size; 60931e270976SMartin KaFai Lau __u32 max_entries; 60941e270976SMartin KaFai Lau __u32 map_flags; 6095067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 609652775b33SJakub Kicinski __u32 ifindex; 609785d33df3SMartin KaFai Lau __u32 btf_vmlinux_value_type_id; 609852775b33SJakub Kicinski __u64 netns_dev; 609952775b33SJakub Kicinski __u64 netns_ino; 610078958fcaSMartin KaFai Lau __u32 btf_id; 61019b2cf328SMartin KaFai Lau __u32 btf_key_type_id; 61029b2cf328SMartin KaFai Lau __u32 btf_value_type_id; 61038845b468SJoanne Koong __u32 :32; /* alignment pad */ 61049330986cSJoanne Koong __u64 map_extra; 61051e270976SMartin KaFai Lau } __attribute__((aligned(8))); 61061e270976SMartin KaFai Lau 610762dab84cSMartin KaFai Lau struct bpf_btf_info { 610862dab84cSMartin KaFai Lau __aligned_u64 btf; 610962dab84cSMartin KaFai Lau __u32 btf_size; 611062dab84cSMartin KaFai Lau __u32 id; 611153297220SAndrii Nakryiko __aligned_u64 name; 611253297220SAndrii Nakryiko __u32 name_len; 611353297220SAndrii Nakryiko __u32 kernel_btf; 611462dab84cSMartin KaFai Lau } __attribute__((aligned(8))); 611562dab84cSMartin KaFai Lau 6116f2e10bffSAndrii Nakryiko struct bpf_link_info { 6117f2e10bffSAndrii Nakryiko __u32 type; 6118f2e10bffSAndrii Nakryiko __u32 id; 6119f2e10bffSAndrii Nakryiko __u32 prog_id; 6120f2e10bffSAndrii Nakryiko union { 6121f2e10bffSAndrii Nakryiko struct { 6122f2e10bffSAndrii Nakryiko __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 6123f2e10bffSAndrii Nakryiko __u32 tp_name_len; /* in/out: tp_name buffer len */ 6124f2e10bffSAndrii Nakryiko } raw_tracepoint; 6125f2e10bffSAndrii Nakryiko struct { 6126f2e10bffSAndrii Nakryiko __u32 attach_type; 6127441e8c66SToke Høiland-Jørgensen __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 6128441e8c66SToke Høiland-Jørgensen __u32 target_btf_id; /* BTF type id inside the object */ 6129f2e10bffSAndrii Nakryiko } tracing; 6130f2e10bffSAndrii Nakryiko struct { 6131f2e10bffSAndrii Nakryiko __u64 cgroup_id; 6132f2e10bffSAndrii Nakryiko __u32 attach_type; 6133f2e10bffSAndrii Nakryiko } cgroup; 61347f045a49SJakub Sitnicki struct { 61356b0a249aSYonghong Song __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 61366b0a249aSYonghong Song __u32 target_name_len; /* in/out: target_name buffer len */ 61376b0a249aSYonghong Song union { 6138b0c9eb37SYonghong Song struct { 61396b0a249aSYonghong Song __u32 map_id; 61406b0a249aSYonghong Song } map; 6141b0c9eb37SYonghong Song }; 61426b0a249aSYonghong Song } iter; 61436b0a249aSYonghong Song struct { 61447f045a49SJakub Sitnicki __u32 netns_ino; 61457f045a49SJakub Sitnicki __u32 attach_type; 61467f045a49SJakub Sitnicki } netns; 6147c1931c97SAndrii Nakryiko struct { 6148c1931c97SAndrii Nakryiko __u32 ifindex; 6149c1931c97SAndrii Nakryiko } xdp; 6150f2e10bffSAndrii Nakryiko }; 6151f2e10bffSAndrii Nakryiko } __attribute__((aligned(8))); 6152f2e10bffSAndrii Nakryiko 61534fbac77dSAndrey Ignatov /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 61544fbac77dSAndrey Ignatov * by user and intended to be used by socket (e.g. to bind to, depends on 6155bfdfa517SRandy Dunlap * attach type). 61564fbac77dSAndrey Ignatov */ 61574fbac77dSAndrey Ignatov struct bpf_sock_addr { 61584fbac77dSAndrey Ignatov __u32 user_family; /* Allows 4-byte read, but no write. */ 61594fbac77dSAndrey Ignatov __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 61604fbac77dSAndrey Ignatov * Stored in network byte order. 61614fbac77dSAndrey Ignatov */ 6162d4ecfeb1SStanislav Fomichev __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 61634fbac77dSAndrey Ignatov * Stored in network byte order. 61644fbac77dSAndrey Ignatov */ 61657aebfa1bSAndrey Ignatov __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 61664fbac77dSAndrey Ignatov * Stored in network byte order 61674fbac77dSAndrey Ignatov */ 61684fbac77dSAndrey Ignatov __u32 family; /* Allows 4-byte read, but no write */ 61694fbac77dSAndrey Ignatov __u32 type; /* Allows 4-byte read, but no write */ 61704fbac77dSAndrey Ignatov __u32 protocol; /* Allows 4-byte read, but no write */ 6171600c70baSStanislav Fomichev __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 61721cedee13SAndrey Ignatov * Stored in network byte order. 61731cedee13SAndrey Ignatov */ 6174d4ecfeb1SStanislav Fomichev __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 61751cedee13SAndrey Ignatov * Stored in network byte order. 61761cedee13SAndrey Ignatov */ 6177fb85c4a7SStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 61784fbac77dSAndrey Ignatov }; 61794fbac77dSAndrey Ignatov 618040304b2aSLawrence Brakmo /* User bpf_sock_ops struct to access socket values and specify request ops 618140304b2aSLawrence Brakmo * and their replies. 618240304b2aSLawrence Brakmo * Some of this fields are in network (bigendian) byte order and may need 618340304b2aSLawrence Brakmo * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 618440304b2aSLawrence Brakmo * New fields can only be added at the end of this structure 618540304b2aSLawrence Brakmo */ 618640304b2aSLawrence Brakmo struct bpf_sock_ops { 618740304b2aSLawrence Brakmo __u32 op; 618840304b2aSLawrence Brakmo union { 6189de525be2SLawrence Brakmo __u32 args[4]; /* Optionally passed to bpf program */ 6190de525be2SLawrence Brakmo __u32 reply; /* Returned by bpf program */ 6191de525be2SLawrence Brakmo __u32 replylong[4]; /* Optionally returned by bpf prog */ 619240304b2aSLawrence Brakmo }; 619340304b2aSLawrence Brakmo __u32 family; 619440304b2aSLawrence Brakmo __u32 remote_ip4; /* Stored in network byte order */ 619540304b2aSLawrence Brakmo __u32 local_ip4; /* Stored in network byte order */ 619640304b2aSLawrence Brakmo __u32 remote_ip6[4]; /* Stored in network byte order */ 619740304b2aSLawrence Brakmo __u32 local_ip6[4]; /* Stored in network byte order */ 619840304b2aSLawrence Brakmo __u32 remote_port; /* Stored in network byte order */ 619940304b2aSLawrence Brakmo __u32 local_port; /* stored in host byte order */ 6200f19397a5SLawrence Brakmo __u32 is_fullsock; /* Some TCP fields are only valid if 6201f19397a5SLawrence Brakmo * there is a full socket. If not, the 6202f19397a5SLawrence Brakmo * fields read as zero. 6203f19397a5SLawrence Brakmo */ 6204f19397a5SLawrence Brakmo __u32 snd_cwnd; 6205f19397a5SLawrence Brakmo __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 6206b13d8807SLawrence Brakmo __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 620744f0e430SLawrence Brakmo __u32 state; 620844f0e430SLawrence Brakmo __u32 rtt_min; 620944f0e430SLawrence Brakmo __u32 snd_ssthresh; 621044f0e430SLawrence Brakmo __u32 rcv_nxt; 621144f0e430SLawrence Brakmo __u32 snd_nxt; 621244f0e430SLawrence Brakmo __u32 snd_una; 621344f0e430SLawrence Brakmo __u32 mss_cache; 621444f0e430SLawrence Brakmo __u32 ecn_flags; 621544f0e430SLawrence Brakmo __u32 rate_delivered; 621644f0e430SLawrence Brakmo __u32 rate_interval_us; 621744f0e430SLawrence Brakmo __u32 packets_out; 621844f0e430SLawrence Brakmo __u32 retrans_out; 621944f0e430SLawrence Brakmo __u32 total_retrans; 622044f0e430SLawrence Brakmo __u32 segs_in; 622144f0e430SLawrence Brakmo __u32 data_segs_in; 622244f0e430SLawrence Brakmo __u32 segs_out; 622344f0e430SLawrence Brakmo __u32 data_segs_out; 622444f0e430SLawrence Brakmo __u32 lost_out; 622544f0e430SLawrence Brakmo __u32 sacked_out; 622644f0e430SLawrence Brakmo __u32 sk_txhash; 622744f0e430SLawrence Brakmo __u64 bytes_received; 622844f0e430SLawrence Brakmo __u64 bytes_acked; 62291314ef56SStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 62300813a841SMartin KaFai Lau /* [skb_data, skb_data_end) covers the whole TCP header. 62310813a841SMartin KaFai Lau * 62320813a841SMartin KaFai Lau * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 62330813a841SMartin KaFai Lau * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 62340813a841SMartin KaFai Lau * header has not been written. 62350813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 62360813a841SMartin KaFai Lau * been written so far. 62370813a841SMartin KaFai Lau * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 62380813a841SMartin KaFai Lau * the 3WHS. 62390813a841SMartin KaFai Lau * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 62400813a841SMartin KaFai Lau * the 3WHS. 62410813a841SMartin KaFai Lau * 62420813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also be used to read a particular option. 62430813a841SMartin KaFai Lau */ 62440813a841SMartin KaFai Lau __bpf_md_ptr(void *, skb_data); 62450813a841SMartin KaFai Lau __bpf_md_ptr(void *, skb_data_end); 62460813a841SMartin KaFai Lau __u32 skb_len; /* The total length of a packet. 62470813a841SMartin KaFai Lau * It includes the header, options, 62480813a841SMartin KaFai Lau * and payload. 62490813a841SMartin KaFai Lau */ 62500813a841SMartin KaFai Lau __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 62510813a841SMartin KaFai Lau * an easy way to check for tcp_flags 62520813a841SMartin KaFai Lau * without parsing skb_data. 62530813a841SMartin KaFai Lau * 62540813a841SMartin KaFai Lau * In particular, the skb_tcp_flags 62550813a841SMartin KaFai Lau * will still be available in 62560813a841SMartin KaFai Lau * BPF_SOCK_OPS_HDR_OPT_LEN even though 62570813a841SMartin KaFai Lau * the outgoing header has not 62580813a841SMartin KaFai Lau * been written yet. 62590813a841SMartin KaFai Lau */ 626040304b2aSLawrence Brakmo }; 626140304b2aSLawrence Brakmo 6262b13d8807SLawrence Brakmo /* Definitions for bpf_sock_ops_cb_flags */ 62631aae4bddSAndrii Nakryiko enum { 62641aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 62651aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 62661aae4bddSAndrii Nakryiko BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 62671aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 62680813a841SMartin KaFai Lau /* Call bpf for all received TCP headers. The bpf prog will be 62690813a841SMartin KaFai Lau * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 62700813a841SMartin KaFai Lau * 62710813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 62720813a841SMartin KaFai Lau * for the header option related helpers that will be useful 62730813a841SMartin KaFai Lau * to the bpf programs. 62740813a841SMartin KaFai Lau * 62750813a841SMartin KaFai Lau * It could be used at the client/active side (i.e. connect() side) 62760813a841SMartin KaFai Lau * when the server told it that the server was in syncookie 62770813a841SMartin KaFai Lau * mode and required the active side to resend the bpf-written 62780813a841SMartin KaFai Lau * options. The active side can keep writing the bpf-options until 62790813a841SMartin KaFai Lau * it received a valid packet from the server side to confirm 62800813a841SMartin KaFai Lau * the earlier packet (and options) has been received. The later 62810813a841SMartin KaFai Lau * example patch is using it like this at the active side when the 62820813a841SMartin KaFai Lau * server is in syncookie mode. 62830813a841SMartin KaFai Lau * 62840813a841SMartin KaFai Lau * The bpf prog will usually turn this off in the common cases. 62850813a841SMartin KaFai Lau */ 628600d211a4SMartin KaFai Lau BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 62870813a841SMartin KaFai Lau /* Call bpf when kernel has received a header option that 62880813a841SMartin KaFai Lau * the kernel cannot handle. The bpf prog will be called under 62890813a841SMartin KaFai Lau * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 62900813a841SMartin KaFai Lau * 62910813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 62920813a841SMartin KaFai Lau * for the header option related helpers that will be useful 62930813a841SMartin KaFai Lau * to the bpf programs. 62940813a841SMartin KaFai Lau */ 629500d211a4SMartin KaFai Lau BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 62960813a841SMartin KaFai Lau /* Call bpf when the kernel is writing header options for the 62970813a841SMartin KaFai Lau * outgoing packet. The bpf prog will first be called 62980813a841SMartin KaFai Lau * to reserve space in a skb under 62990813a841SMartin KaFai Lau * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 63000813a841SMartin KaFai Lau * the bpf prog will be called to write the header option(s) 63010813a841SMartin KaFai Lau * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 63020813a841SMartin KaFai Lau * 63030813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 63040813a841SMartin KaFai Lau * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 63050813a841SMartin KaFai Lau * related helpers that will be useful to the bpf programs. 63060813a841SMartin KaFai Lau * 63070813a841SMartin KaFai Lau * The kernel gets its chance to reserve space and write 63080813a841SMartin KaFai Lau * options first before the BPF program does. 63090813a841SMartin KaFai Lau */ 6310331fca43SMartin KaFai Lau BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 63111aae4bddSAndrii Nakryiko /* Mask of all currently supported cb flags */ 6312331fca43SMartin KaFai Lau BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 63131aae4bddSAndrii Nakryiko }; 6314b13d8807SLawrence Brakmo 631540304b2aSLawrence Brakmo /* List of known BPF sock_ops operators. 631640304b2aSLawrence Brakmo * New entries can only be added at the end 631740304b2aSLawrence Brakmo */ 631840304b2aSLawrence Brakmo enum { 631940304b2aSLawrence Brakmo BPF_SOCK_OPS_VOID, 63208550f328SLawrence Brakmo BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 63218550f328SLawrence Brakmo * -1 if default value should be used 63228550f328SLawrence Brakmo */ 632313d3b1ebSLawrence Brakmo BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 632413d3b1ebSLawrence Brakmo * window (in packets) or -1 if default 632513d3b1ebSLawrence Brakmo * value should be used 632613d3b1ebSLawrence Brakmo */ 63279872a4bdSLawrence Brakmo BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 63289872a4bdSLawrence Brakmo * active connection is initialized 63299872a4bdSLawrence Brakmo */ 63309872a4bdSLawrence Brakmo BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 63319872a4bdSLawrence Brakmo * active connection is 63329872a4bdSLawrence Brakmo * established 63339872a4bdSLawrence Brakmo */ 63349872a4bdSLawrence Brakmo BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 63359872a4bdSLawrence Brakmo * passive connection is 63369872a4bdSLawrence Brakmo * established 63379872a4bdSLawrence Brakmo */ 633891b5b21cSLawrence Brakmo BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 633991b5b21cSLawrence Brakmo * needs ECN 634091b5b21cSLawrence Brakmo */ 6341e6546ef6SLawrence Brakmo BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 6342e6546ef6SLawrence Brakmo * based on the path and may be 6343e6546ef6SLawrence Brakmo * dependent on the congestion control 6344e6546ef6SLawrence Brakmo * algorithm. In general it indicates 6345e6546ef6SLawrence Brakmo * a congestion threshold. RTTs above 6346e6546ef6SLawrence Brakmo * this indicate congestion 6347e6546ef6SLawrence Brakmo */ 6348f89013f6SLawrence Brakmo BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 6349f89013f6SLawrence Brakmo * Arg1: value of icsk_retransmits 6350f89013f6SLawrence Brakmo * Arg2: value of icsk_rto 6351f89013f6SLawrence Brakmo * Arg3: whether RTO has expired 6352f89013f6SLawrence Brakmo */ 6353a31ad29eSLawrence Brakmo BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6354a31ad29eSLawrence Brakmo * Arg1: sequence number of 1st byte 6355a31ad29eSLawrence Brakmo * Arg2: # segments 6356a31ad29eSLawrence Brakmo * Arg3: return value of 6357a31ad29eSLawrence Brakmo * tcp_transmit_skb (0 => success) 6358a31ad29eSLawrence Brakmo */ 6359d4487491SLawrence Brakmo BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6360d4487491SLawrence Brakmo * Arg1: old_state 6361d4487491SLawrence Brakmo * Arg2: new_state 6362d4487491SLawrence Brakmo */ 6363f333ee0cSAndrey Ignatov BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6364f333ee0cSAndrey Ignatov * socket transition to LISTEN state. 6365f333ee0cSAndrey Ignatov */ 636623729ff2SStanislav Fomichev BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 636723729ff2SStanislav Fomichev */ 63680813a841SMartin KaFai Lau BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 63690813a841SMartin KaFai Lau * It will be called to handle 63700813a841SMartin KaFai Lau * the packets received at 63710813a841SMartin KaFai Lau * an already established 63720813a841SMartin KaFai Lau * connection. 63730813a841SMartin KaFai Lau * 63740813a841SMartin KaFai Lau * sock_ops->skb_data: 63750813a841SMartin KaFai Lau * Referring to the received skb. 63760813a841SMartin KaFai Lau * It covers the TCP header only. 63770813a841SMartin KaFai Lau * 63780813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also 63790813a841SMartin KaFai Lau * be used to search for a 63800813a841SMartin KaFai Lau * particular option. 63810813a841SMartin KaFai Lau */ 63820813a841SMartin KaFai Lau BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 63830813a841SMartin KaFai Lau * header option later in 63840813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 63850813a841SMartin KaFai Lau * Arg1: bool want_cookie. (in 63860813a841SMartin KaFai Lau * writing SYNACK only) 63870813a841SMartin KaFai Lau * 63880813a841SMartin KaFai Lau * sock_ops->skb_data: 63890813a841SMartin KaFai Lau * Not available because no header has 63900813a841SMartin KaFai Lau * been written yet. 63910813a841SMartin KaFai Lau * 63920813a841SMartin KaFai Lau * sock_ops->skb_tcp_flags: 63930813a841SMartin KaFai Lau * The tcp_flags of the 63940813a841SMartin KaFai Lau * outgoing skb. (e.g. SYN, ACK, FIN). 63950813a841SMartin KaFai Lau * 63960813a841SMartin KaFai Lau * bpf_reserve_hdr_opt() should 63970813a841SMartin KaFai Lau * be used to reserve space. 63980813a841SMartin KaFai Lau */ 63990813a841SMartin KaFai Lau BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 64000813a841SMartin KaFai Lau * Arg1: bool want_cookie. (in 64010813a841SMartin KaFai Lau * writing SYNACK only) 64020813a841SMartin KaFai Lau * 64030813a841SMartin KaFai Lau * sock_ops->skb_data: 64040813a841SMartin KaFai Lau * Referring to the outgoing skb. 64050813a841SMartin KaFai Lau * It covers the TCP header 64060813a841SMartin KaFai Lau * that has already been written 64070813a841SMartin KaFai Lau * by the kernel and the 64080813a841SMartin KaFai Lau * earlier bpf-progs. 64090813a841SMartin KaFai Lau * 64100813a841SMartin KaFai Lau * sock_ops->skb_tcp_flags: 64110813a841SMartin KaFai Lau * The tcp_flags of the outgoing 64120813a841SMartin KaFai Lau * skb. (e.g. SYN, ACK, FIN). 64130813a841SMartin KaFai Lau * 64140813a841SMartin KaFai Lau * bpf_store_hdr_opt() should 64150813a841SMartin KaFai Lau * be used to write the 64160813a841SMartin KaFai Lau * option. 64170813a841SMartin KaFai Lau * 64180813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also 64190813a841SMartin KaFai Lau * be used to search for a 64200813a841SMartin KaFai Lau * particular option that 64210813a841SMartin KaFai Lau * has already been written 64220813a841SMartin KaFai Lau * by the kernel or the 64230813a841SMartin KaFai Lau * earlier bpf-progs. 64240813a841SMartin KaFai Lau */ 6425d4487491SLawrence Brakmo }; 6426d4487491SLawrence Brakmo 6427d4487491SLawrence Brakmo /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 6428d4487491SLawrence Brakmo * changes between the TCP and BPF versions. Ideally this should never happen. 6429d4487491SLawrence Brakmo * If it does, we need to add code to convert them before calling 6430d4487491SLawrence Brakmo * the BPF sock_ops function. 6431d4487491SLawrence Brakmo */ 6432d4487491SLawrence Brakmo enum { 6433d4487491SLawrence Brakmo BPF_TCP_ESTABLISHED = 1, 6434d4487491SLawrence Brakmo BPF_TCP_SYN_SENT, 6435d4487491SLawrence Brakmo BPF_TCP_SYN_RECV, 6436d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT1, 6437d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT2, 6438d4487491SLawrence Brakmo BPF_TCP_TIME_WAIT, 6439d4487491SLawrence Brakmo BPF_TCP_CLOSE, 6440d4487491SLawrence Brakmo BPF_TCP_CLOSE_WAIT, 6441d4487491SLawrence Brakmo BPF_TCP_LAST_ACK, 6442d4487491SLawrence Brakmo BPF_TCP_LISTEN, 6443d4487491SLawrence Brakmo BPF_TCP_CLOSING, /* Now a valid state */ 6444d4487491SLawrence Brakmo BPF_TCP_NEW_SYN_RECV, 6445d4487491SLawrence Brakmo 6446d4487491SLawrence Brakmo BPF_TCP_MAX_STATES /* Leave at the end! */ 644740304b2aSLawrence Brakmo }; 644840304b2aSLawrence Brakmo 64491aae4bddSAndrii Nakryiko enum { 64501aae4bddSAndrii Nakryiko TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 64511aae4bddSAndrii Nakryiko TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 64522b8ee4f0SMartin KaFai Lau TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 6453ca584ba0SMartin KaFai Lau TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 64540813a841SMartin KaFai Lau /* Copy the SYN pkt to optval 64550813a841SMartin KaFai Lau * 64560813a841SMartin KaFai Lau * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 64570813a841SMartin KaFai Lau * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 64580813a841SMartin KaFai Lau * to only getting from the saved_syn. It can either get the 64590813a841SMartin KaFai Lau * syn packet from: 64600813a841SMartin KaFai Lau * 64610813a841SMartin KaFai Lau * 1. the just-received SYN packet (only available when writing the 64620813a841SMartin KaFai Lau * SYNACK). It will be useful when it is not necessary to 64630813a841SMartin KaFai Lau * save the SYN packet for latter use. It is also the only way 64640813a841SMartin KaFai Lau * to get the SYN during syncookie mode because the syn 64650813a841SMartin KaFai Lau * packet cannot be saved during syncookie. 64660813a841SMartin KaFai Lau * 64670813a841SMartin KaFai Lau * OR 64680813a841SMartin KaFai Lau * 64690813a841SMartin KaFai Lau * 2. the earlier saved syn which was done by 64700813a841SMartin KaFai Lau * bpf_setsockopt(TCP_SAVE_SYN). 64710813a841SMartin KaFai Lau * 64720813a841SMartin KaFai Lau * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 64730813a841SMartin KaFai Lau * SYN packet is obtained. 64740813a841SMartin KaFai Lau * 64750813a841SMartin KaFai Lau * If the bpf-prog does not need the IP[46] header, the 64760813a841SMartin KaFai Lau * bpf-prog can avoid parsing the IP header by using 64770813a841SMartin KaFai Lau * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 64780813a841SMartin KaFai Lau * IP[46] and TCP header by using TCP_BPF_SYN_IP. 64790813a841SMartin KaFai Lau * 64800813a841SMartin KaFai Lau * >0: Total number of bytes copied 64810813a841SMartin KaFai Lau * -ENOSPC: Not enough space in optval. Only optlen number of 64820813a841SMartin KaFai Lau * bytes is copied. 64830813a841SMartin KaFai Lau * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 64840813a841SMartin KaFai Lau * is not saved by setsockopt(TCP_SAVE_SYN). 64850813a841SMartin KaFai Lau */ 64860813a841SMartin KaFai Lau TCP_BPF_SYN = 1005, /* Copy the TCP header */ 64870813a841SMartin KaFai Lau TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 6488267cf9faSMartin KaFai Lau TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 64890813a841SMartin KaFai Lau }; 64900813a841SMartin KaFai Lau 64910813a841SMartin KaFai Lau enum { 64920813a841SMartin KaFai Lau BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 64930813a841SMartin KaFai Lau }; 64940813a841SMartin KaFai Lau 64950813a841SMartin KaFai Lau /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 64960813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 64970813a841SMartin KaFai Lau */ 64980813a841SMartin KaFai Lau enum { 64990813a841SMartin KaFai Lau BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 65000813a841SMartin KaFai Lau * total option spaces 65010813a841SMartin KaFai Lau * required for an established 65020813a841SMartin KaFai Lau * sk in order to calculate the 65030813a841SMartin KaFai Lau * MSS. No skb is actually 65040813a841SMartin KaFai Lau * sent. 65050813a841SMartin KaFai Lau */ 65060813a841SMartin KaFai Lau BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 65070813a841SMartin KaFai Lau * when sending a SYN. 65080813a841SMartin KaFai Lau */ 65091aae4bddSAndrii Nakryiko }; 6510fc747810SLawrence Brakmo 6511908432caSYonghong Song struct bpf_perf_event_value { 6512908432caSYonghong Song __u64 counter; 6513908432caSYonghong Song __u64 enabled; 6514908432caSYonghong Song __u64 running; 6515908432caSYonghong Song }; 6516908432caSYonghong Song 65171aae4bddSAndrii Nakryiko enum { 65181aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 65191aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_READ = (1ULL << 1), 65201aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_WRITE = (1ULL << 2), 65211aae4bddSAndrii Nakryiko }; 6522ebc614f6SRoman Gushchin 65231aae4bddSAndrii Nakryiko enum { 65241aae4bddSAndrii Nakryiko BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 65251aae4bddSAndrii Nakryiko BPF_DEVCG_DEV_CHAR = (1ULL << 1), 65261aae4bddSAndrii Nakryiko }; 6527ebc614f6SRoman Gushchin 6528ebc614f6SRoman Gushchin struct bpf_cgroup_dev_ctx { 652906ef0ccbSYonghong Song /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 653006ef0ccbSYonghong Song __u32 access_type; 6531ebc614f6SRoman Gushchin __u32 major; 6532ebc614f6SRoman Gushchin __u32 minor; 6533ebc614f6SRoman Gushchin }; 6534ebc614f6SRoman Gushchin 6535c4f6699dSAlexei Starovoitov struct bpf_raw_tracepoint_args { 6536c4f6699dSAlexei Starovoitov __u64 args[0]; 6537c4f6699dSAlexei Starovoitov }; 6538c4f6699dSAlexei Starovoitov 653987f5fc7eSDavid Ahern /* DIRECT: Skip the FIB rules and go to FIB table associated with device 654087f5fc7eSDavid Ahern * OUTPUT: Do lookup from egress perspective; default is ingress 654187f5fc7eSDavid Ahern */ 65421aae4bddSAndrii Nakryiko enum { 65431aae4bddSAndrii Nakryiko BPF_FIB_LOOKUP_DIRECT = (1U << 0), 65441aae4bddSAndrii Nakryiko BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 65451aae4bddSAndrii Nakryiko }; 654687f5fc7eSDavid Ahern 65474c79579bSDavid Ahern enum { 65484c79579bSDavid Ahern BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 65494c79579bSDavid Ahern BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 65504c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 65514c79579bSDavid Ahern BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 65524c79579bSDavid Ahern BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 65534c79579bSDavid Ahern BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 65544c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 65554c79579bSDavid Ahern BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 65564c79579bSDavid Ahern BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 65574c79579bSDavid Ahern }; 65584c79579bSDavid Ahern 655987f5fc7eSDavid Ahern struct bpf_fib_lookup { 6560fa898d76SDavid Ahern /* input: network family for lookup (AF_INET, AF_INET6) 6561fa898d76SDavid Ahern * output: network family of egress nexthop 6562fa898d76SDavid Ahern */ 6563fa898d76SDavid Ahern __u8 family; 656487f5fc7eSDavid Ahern 656587f5fc7eSDavid Ahern /* set if lookup is to consider L4 data - e.g., FIB rules */ 656687f5fc7eSDavid Ahern __u8 l4_protocol; 656787f5fc7eSDavid Ahern __be16 sport; 656887f5fc7eSDavid Ahern __be16 dport; 656987f5fc7eSDavid Ahern 6570e1850ea9SJesper Dangaard Brouer union { /* used for MTU check */ 6571e1850ea9SJesper Dangaard Brouer /* input to lookup */ 6572e1850ea9SJesper Dangaard Brouer __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 65734c79579bSDavid Ahern 6574e1850ea9SJesper Dangaard Brouer /* output: MTU value */ 6575e1850ea9SJesper Dangaard Brouer __u16 mtu_result; 6576e1850ea9SJesper Dangaard Brouer }; 65774c79579bSDavid Ahern /* input: L3 device index for lookup 65784c79579bSDavid Ahern * output: device index from FIB lookup 65794c79579bSDavid Ahern */ 65804c79579bSDavid Ahern __u32 ifindex; 658187f5fc7eSDavid Ahern 658287f5fc7eSDavid Ahern union { 658387f5fc7eSDavid Ahern /* inputs to lookup */ 658487f5fc7eSDavid Ahern __u8 tos; /* AF_INET */ 6585bd3a08aaSDavid Ahern __be32 flowinfo; /* AF_INET6, flow_label + priority */ 658687f5fc7eSDavid Ahern 6587fa898d76SDavid Ahern /* output: metric of fib result (IPv4/IPv6 only) */ 658887f5fc7eSDavid Ahern __u32 rt_metric; 658987f5fc7eSDavid Ahern }; 659087f5fc7eSDavid Ahern 659187f5fc7eSDavid Ahern union { 659287f5fc7eSDavid Ahern __be32 ipv4_src; 659387f5fc7eSDavid Ahern __u32 ipv6_src[4]; /* in6_addr; network order */ 659487f5fc7eSDavid Ahern }; 659587f5fc7eSDavid Ahern 6596fa898d76SDavid Ahern /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 6597fa898d76SDavid Ahern * network header. output: bpf_fib_lookup sets to gateway address 6598fa898d76SDavid Ahern * if FIB lookup returns gateway route 659987f5fc7eSDavid Ahern */ 660087f5fc7eSDavid Ahern union { 660187f5fc7eSDavid Ahern __be32 ipv4_dst; 660287f5fc7eSDavid Ahern __u32 ipv6_dst[4]; /* in6_addr; network order */ 660387f5fc7eSDavid Ahern }; 660487f5fc7eSDavid Ahern 660587f5fc7eSDavid Ahern /* output */ 660687f5fc7eSDavid Ahern __be16 h_vlan_proto; 660787f5fc7eSDavid Ahern __be16 h_vlan_TCI; 660887f5fc7eSDavid Ahern __u8 smac[6]; /* ETH_ALEN */ 660987f5fc7eSDavid Ahern __u8 dmac[6]; /* ETH_ALEN */ 661087f5fc7eSDavid Ahern }; 661187f5fc7eSDavid Ahern 6612ba452c9eSToke Høiland-Jørgensen struct bpf_redir_neigh { 6613ba452c9eSToke Høiland-Jørgensen /* network family for lookup (AF_INET, AF_INET6) */ 6614ba452c9eSToke Høiland-Jørgensen __u32 nh_family; 6615ba452c9eSToke Høiland-Jørgensen /* network address of nexthop; skips fib lookup to find gateway */ 6616ba452c9eSToke Høiland-Jørgensen union { 6617ba452c9eSToke Høiland-Jørgensen __be32 ipv4_nh; 6618ba452c9eSToke Høiland-Jørgensen __u32 ipv6_nh[4]; /* in6_addr; network order */ 6619ba452c9eSToke Høiland-Jørgensen }; 6620ba452c9eSToke Høiland-Jørgensen }; 6621ba452c9eSToke Høiland-Jørgensen 662234b2021cSJesper Dangaard Brouer /* bpf_check_mtu flags*/ 662334b2021cSJesper Dangaard Brouer enum bpf_check_mtu_flags { 662434b2021cSJesper Dangaard Brouer BPF_MTU_CHK_SEGS = (1U << 0), 662534b2021cSJesper Dangaard Brouer }; 662634b2021cSJesper Dangaard Brouer 662734b2021cSJesper Dangaard Brouer enum bpf_check_mtu_ret { 662834b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 662934b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 663034b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 663134b2021cSJesper Dangaard Brouer }; 663234b2021cSJesper Dangaard Brouer 663341bdc4b4SYonghong Song enum bpf_task_fd_type { 663441bdc4b4SYonghong Song BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 663541bdc4b4SYonghong Song BPF_FD_TYPE_TRACEPOINT, /* tp name */ 663641bdc4b4SYonghong Song BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 663741bdc4b4SYonghong Song BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 663841bdc4b4SYonghong Song BPF_FD_TYPE_UPROBE, /* filename + offset */ 663941bdc4b4SYonghong Song BPF_FD_TYPE_URETPROBE, /* filename + offset */ 664041bdc4b4SYonghong Song }; 664141bdc4b4SYonghong Song 66421aae4bddSAndrii Nakryiko enum { 66431aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 66441aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 66451aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 66461aae4bddSAndrii Nakryiko }; 6647086f9568SStanislav Fomichev 6648d58e468bSPetar Penkov struct bpf_flow_keys { 6649d58e468bSPetar Penkov __u16 nhoff; 6650d58e468bSPetar Penkov __u16 thoff; 6651d58e468bSPetar Penkov __u16 addr_proto; /* ETH_P_* of valid addrs */ 6652d58e468bSPetar Penkov __u8 is_frag; 6653d58e468bSPetar Penkov __u8 is_first_frag; 6654d58e468bSPetar Penkov __u8 is_encap; 6655d58e468bSPetar Penkov __u8 ip_proto; 6656d58e468bSPetar Penkov __be16 n_proto; 6657d58e468bSPetar Penkov __be16 sport; 6658d58e468bSPetar Penkov __be16 dport; 6659d58e468bSPetar Penkov union { 6660d58e468bSPetar Penkov struct { 6661d58e468bSPetar Penkov __be32 ipv4_src; 6662d58e468bSPetar Penkov __be32 ipv4_dst; 6663d58e468bSPetar Penkov }; 6664d58e468bSPetar Penkov struct { 6665d58e468bSPetar Penkov __u32 ipv6_src[4]; /* in6_addr; network order */ 6666d58e468bSPetar Penkov __u32 ipv6_dst[4]; /* in6_addr; network order */ 6667d58e468bSPetar Penkov }; 6668d58e468bSPetar Penkov }; 6669086f9568SStanislav Fomichev __u32 flags; 667071c99e32SStanislav Fomichev __be32 flow_label; 6671d58e468bSPetar Penkov }; 6672d58e468bSPetar Penkov 6673838e9690SYonghong Song struct bpf_func_info { 6674d30d42e0SMartin KaFai Lau __u32 insn_off; 6675838e9690SYonghong Song __u32 type_id; 6676838e9690SYonghong Song }; 6677838e9690SYonghong Song 6678c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 6679c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 6680c454a46bSMartin KaFai Lau 6681c454a46bSMartin KaFai Lau struct bpf_line_info { 6682c454a46bSMartin KaFai Lau __u32 insn_off; 6683c454a46bSMartin KaFai Lau __u32 file_name_off; 6684c454a46bSMartin KaFai Lau __u32 line_off; 6685c454a46bSMartin KaFai Lau __u32 line_col; 6686c454a46bSMartin KaFai Lau }; 6687c454a46bSMartin KaFai Lau 6688d83525caSAlexei Starovoitov struct bpf_spin_lock { 6689d83525caSAlexei Starovoitov __u32 val; 6690d83525caSAlexei Starovoitov }; 66917b146cebSAndrey Ignatov 6692b00628b1SAlexei Starovoitov struct bpf_timer { 6693b00628b1SAlexei Starovoitov __u64 :64; 6694b00628b1SAlexei Starovoitov __u64 :64; 6695b00628b1SAlexei Starovoitov } __attribute__((aligned(8))); 6696b00628b1SAlexei Starovoitov 669797e03f52SJoanne Koong struct bpf_dynptr { 669897e03f52SJoanne Koong __u64 :64; 669997e03f52SJoanne Koong __u64 :64; 670097e03f52SJoanne Koong } __attribute__((aligned(8))); 670197e03f52SJoanne Koong 67027b146cebSAndrey Ignatov struct bpf_sysctl { 67037b146cebSAndrey Ignatov __u32 write; /* Sysctl is being read (= 0) or written (= 1). 67047b146cebSAndrey Ignatov * Allows 1,2,4-byte read, but no write. 67057b146cebSAndrey Ignatov */ 6706e1550bfeSAndrey Ignatov __u32 file_pos; /* Sysctl file position to read from, write to. 6707e1550bfeSAndrey Ignatov * Allows 1,2,4-byte read an 4-byte write. 6708e1550bfeSAndrey Ignatov */ 67097b146cebSAndrey Ignatov }; 67107b146cebSAndrey Ignatov 67110d01da6aSStanislav Fomichev struct bpf_sockopt { 67120d01da6aSStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 67130d01da6aSStanislav Fomichev __bpf_md_ptr(void *, optval); 67140d01da6aSStanislav Fomichev __bpf_md_ptr(void *, optval_end); 67150d01da6aSStanislav Fomichev 67160d01da6aSStanislav Fomichev __s32 level; 67170d01da6aSStanislav Fomichev __s32 optname; 67180d01da6aSStanislav Fomichev __s32 optlen; 67190d01da6aSStanislav Fomichev __s32 retval; 67200d01da6aSStanislav Fomichev }; 67210d01da6aSStanislav Fomichev 6722b4490c5cSCarlos Neira struct bpf_pidns_info { 6723b4490c5cSCarlos Neira __u32 pid; 6724b4490c5cSCarlos Neira __u32 tgid; 6725b4490c5cSCarlos Neira }; 6726e9ddbb77SJakub Sitnicki 6727e9ddbb77SJakub Sitnicki /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 6728e9ddbb77SJakub Sitnicki struct bpf_sk_lookup { 67297c32e8f8SLorenz Bauer union { 6730e9ddbb77SJakub Sitnicki __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 67317c32e8f8SLorenz Bauer __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 67327c32e8f8SLorenz Bauer }; 6733e9ddbb77SJakub Sitnicki 6734e9ddbb77SJakub Sitnicki __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 6735e9ddbb77SJakub Sitnicki __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 6736e9ddbb77SJakub Sitnicki __u32 remote_ip4; /* Network byte order */ 6737e9ddbb77SJakub Sitnicki __u32 remote_ip6[4]; /* Network byte order */ 67389a69e2b3SJakub Sitnicki __be16 remote_port; /* Network byte order */ 67399a69e2b3SJakub Sitnicki __u16 :16; /* Zero padding */ 6740e9ddbb77SJakub Sitnicki __u32 local_ip4; /* Network byte order */ 6741e9ddbb77SJakub Sitnicki __u32 local_ip6[4]; /* Network byte order */ 6742e9ddbb77SJakub Sitnicki __u32 local_port; /* Host byte order */ 6743f8931565SMark Pashmfouroush __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 6744e9ddbb77SJakub Sitnicki }; 6745e9ddbb77SJakub Sitnicki 6746c4d0bfb4SAlan Maguire /* 6747c4d0bfb4SAlan Maguire * struct btf_ptr is used for typed pointer representation; the 6748c4d0bfb4SAlan Maguire * type id is used to render the pointer data as the appropriate type 6749c4d0bfb4SAlan Maguire * via the bpf_snprintf_btf() helper described above. A flags field - 6750c4d0bfb4SAlan Maguire * potentially to specify additional details about the BTF pointer 6751c4d0bfb4SAlan Maguire * (rather than its mode of display) - is included for future use. 6752c4d0bfb4SAlan Maguire * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 6753c4d0bfb4SAlan Maguire */ 6754c4d0bfb4SAlan Maguire struct btf_ptr { 6755c4d0bfb4SAlan Maguire void *ptr; 6756c4d0bfb4SAlan Maguire __u32 type_id; 6757c4d0bfb4SAlan Maguire __u32 flags; /* BTF ptr flags; unused at present. */ 6758c4d0bfb4SAlan Maguire }; 6759c4d0bfb4SAlan Maguire 6760c4d0bfb4SAlan Maguire /* 6761c4d0bfb4SAlan Maguire * Flags to control bpf_snprintf_btf() behaviour. 6762c4d0bfb4SAlan Maguire * - BTF_F_COMPACT: no formatting around type information 6763c4d0bfb4SAlan Maguire * - BTF_F_NONAME: no struct/union member names/types 6764c4d0bfb4SAlan Maguire * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 6765c4d0bfb4SAlan Maguire * equivalent to %px. 6766c4d0bfb4SAlan Maguire * - BTF_F_ZERO: show zero-valued struct/union members; they 6767c4d0bfb4SAlan Maguire * are not displayed by default 6768c4d0bfb4SAlan Maguire */ 6769c4d0bfb4SAlan Maguire enum { 6770c4d0bfb4SAlan Maguire BTF_F_COMPACT = (1ULL << 0), 6771c4d0bfb4SAlan Maguire BTF_F_NONAME = (1ULL << 1), 6772c4d0bfb4SAlan Maguire BTF_F_PTR_RAW = (1ULL << 2), 6773c4d0bfb4SAlan Maguire BTF_F_ZERO = (1ULL << 3), 6774c4d0bfb4SAlan Maguire }; 6775c4d0bfb4SAlan Maguire 677646334a0cSAlexei Starovoitov /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 677746334a0cSAlexei Starovoitov * has to be adjusted by relocations. It is emitted by llvm and passed to 677846334a0cSAlexei Starovoitov * libbpf and later to the kernel. 677946334a0cSAlexei Starovoitov */ 678046334a0cSAlexei Starovoitov enum bpf_core_relo_kind { 678146334a0cSAlexei Starovoitov BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 678246334a0cSAlexei Starovoitov BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 678346334a0cSAlexei Starovoitov BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 678446334a0cSAlexei Starovoitov BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 678546334a0cSAlexei Starovoitov BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 678646334a0cSAlexei Starovoitov BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 678746334a0cSAlexei Starovoitov BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 678846334a0cSAlexei Starovoitov BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 678946334a0cSAlexei Starovoitov BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 679046334a0cSAlexei Starovoitov BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 679146334a0cSAlexei Starovoitov BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 679246334a0cSAlexei Starovoitov BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 67933c660a5dSDaniel Müller BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */ 679446334a0cSAlexei Starovoitov }; 679546334a0cSAlexei Starovoitov 6796fbd94c7aSAlexei Starovoitov /* 6797fbd94c7aSAlexei Starovoitov * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 6798fbd94c7aSAlexei Starovoitov * and from libbpf to the kernel. 6799fbd94c7aSAlexei Starovoitov * 6800fbd94c7aSAlexei Starovoitov * CO-RE relocation captures the following data: 6801fbd94c7aSAlexei Starovoitov * - insn_off - instruction offset (in bytes) within a BPF program that needs 6802fbd94c7aSAlexei Starovoitov * its insn->imm field to be relocated with actual field info; 6803fbd94c7aSAlexei Starovoitov * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 6804fbd94c7aSAlexei Starovoitov * type or field; 6805fbd94c7aSAlexei Starovoitov * - access_str_off - offset into corresponding .BTF string section. String 6806fbd94c7aSAlexei Starovoitov * interpretation depends on specific relocation kind: 6807fbd94c7aSAlexei Starovoitov * - for field-based relocations, string encodes an accessed field using 6808fbd94c7aSAlexei Starovoitov * a sequence of field and array indices, separated by colon (:). It's 6809fbd94c7aSAlexei Starovoitov * conceptually very close to LLVM's getelementptr ([0]) instruction's 6810fbd94c7aSAlexei Starovoitov * arguments for identifying offset to a field. 6811fbd94c7aSAlexei Starovoitov * - for type-based relocations, strings is expected to be just "0"; 6812fbd94c7aSAlexei Starovoitov * - for enum value-based relocations, string contains an index of enum 6813fbd94c7aSAlexei Starovoitov * value within its enum type; 6814fbd94c7aSAlexei Starovoitov * - kind - one of enum bpf_core_relo_kind; 6815fbd94c7aSAlexei Starovoitov * 6816fbd94c7aSAlexei Starovoitov * Example: 6817fbd94c7aSAlexei Starovoitov * struct sample { 6818fbd94c7aSAlexei Starovoitov * int a; 6819fbd94c7aSAlexei Starovoitov * struct { 6820fbd94c7aSAlexei Starovoitov * int b[10]; 6821fbd94c7aSAlexei Starovoitov * }; 6822fbd94c7aSAlexei Starovoitov * }; 6823fbd94c7aSAlexei Starovoitov * 6824fbd94c7aSAlexei Starovoitov * struct sample *s = ...; 6825fbd94c7aSAlexei Starovoitov * int *x = &s->a; // encoded as "0:0" (a is field #0) 6826fbd94c7aSAlexei Starovoitov * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 6827fbd94c7aSAlexei Starovoitov * // b is field #0 inside anon struct, accessing elem #5) 6828fbd94c7aSAlexei Starovoitov * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 6829fbd94c7aSAlexei Starovoitov * 6830fbd94c7aSAlexei Starovoitov * type_id for all relocs in this example will capture BTF type id of 6831fbd94c7aSAlexei Starovoitov * `struct sample`. 6832fbd94c7aSAlexei Starovoitov * 6833fbd94c7aSAlexei Starovoitov * Such relocation is emitted when using __builtin_preserve_access_index() 6834fbd94c7aSAlexei Starovoitov * Clang built-in, passing expression that captures field address, e.g.: 6835fbd94c7aSAlexei Starovoitov * 6836fbd94c7aSAlexei Starovoitov * bpf_probe_read(&dst, sizeof(dst), 6837fbd94c7aSAlexei Starovoitov * __builtin_preserve_access_index(&src->a.b.c)); 6838fbd94c7aSAlexei Starovoitov * 6839fbd94c7aSAlexei Starovoitov * In this case Clang will emit field relocation recording necessary data to 6840fbd94c7aSAlexei Starovoitov * be able to find offset of embedded `a.b.c` field within `src` struct. 6841fbd94c7aSAlexei Starovoitov * 6842fbd94c7aSAlexei Starovoitov * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 6843fbd94c7aSAlexei Starovoitov */ 6844fbd94c7aSAlexei Starovoitov struct bpf_core_relo { 6845fbd94c7aSAlexei Starovoitov __u32 insn_off; 6846fbd94c7aSAlexei Starovoitov __u32 type_id; 6847fbd94c7aSAlexei Starovoitov __u32 access_str_off; 6848fbd94c7aSAlexei Starovoitov enum bpf_core_relo_kind kind; 6849fbd94c7aSAlexei Starovoitov }; 6850fbd94c7aSAlexei Starovoitov 6851daedfb22SAlexei Starovoitov #endif /* _UAPI__LINUX_BPF_H__ */ 6852