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 */ 821e6e9d0fSGustavo A. R. Silva __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 */ 87de9cbbaaSRoman Gushchin __u32 attach_type; /* program 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 * 315f67c9cbfSJoe Stringer * Return 316f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 317f67c9cbfSJoe Stringer * is set appropriately. 318f67c9cbfSJoe Stringer * 3192a3fdca4SJoe Stringer * **ENOSPC** 3202a3fdca4SJoe Stringer * Either *data_size_out* or *ctx_size_out* is too small. 3212a3fdca4SJoe Stringer * **ENOTSUPP** 3222a3fdca4SJoe Stringer * This command is not supported by the program type of 3232a3fdca4SJoe Stringer * the program referred to by *prog_fd*. 3242a3fdca4SJoe Stringer * 325f67c9cbfSJoe Stringer * BPF_PROG_GET_NEXT_ID 326f67c9cbfSJoe Stringer * Description 327f67c9cbfSJoe Stringer * Fetch the next eBPF program currently loaded into the kernel. 328f67c9cbfSJoe Stringer * 329f67c9cbfSJoe Stringer * Looks for the eBPF program with an id greater than *start_id* 330f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF programs 331f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 332f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 333f67c9cbfSJoe Stringer * 334f67c9cbfSJoe Stringer * Return 335f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 336f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 337f67c9cbfSJoe Stringer * 338f67c9cbfSJoe Stringer * BPF_MAP_GET_NEXT_ID 339f67c9cbfSJoe Stringer * Description 340f67c9cbfSJoe Stringer * Fetch the next eBPF map currently loaded into the kernel. 341f67c9cbfSJoe Stringer * 342f67c9cbfSJoe Stringer * Looks for the eBPF map with an id greater than *start_id* 343f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF maps 344f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 345f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 346f67c9cbfSJoe Stringer * 347f67c9cbfSJoe Stringer * Return 348f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 349f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 350f67c9cbfSJoe Stringer * 351f67c9cbfSJoe Stringer * BPF_PROG_GET_FD_BY_ID 352f67c9cbfSJoe Stringer * Description 353f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF program corresponding to 354f67c9cbfSJoe Stringer * *prog_id*. 355f67c9cbfSJoe Stringer * 356f67c9cbfSJoe Stringer * Return 357f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 358f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 359f67c9cbfSJoe Stringer * 360f67c9cbfSJoe Stringer * BPF_MAP_GET_FD_BY_ID 361f67c9cbfSJoe Stringer * Description 362f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF map corresponding to 363f67c9cbfSJoe Stringer * *map_id*. 364f67c9cbfSJoe Stringer * 365f67c9cbfSJoe Stringer * Return 366f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 367f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 368f67c9cbfSJoe Stringer * 369f67c9cbfSJoe Stringer * BPF_OBJ_GET_INFO_BY_FD 370f67c9cbfSJoe Stringer * Description 371f67c9cbfSJoe Stringer * Obtain information about the eBPF object corresponding to 372f67c9cbfSJoe Stringer * *bpf_fd*. 373f67c9cbfSJoe Stringer * 374f67c9cbfSJoe Stringer * Populates up to *info_len* bytes of *info*, which will be in 375f67c9cbfSJoe Stringer * one of the following formats depending on the eBPF object type 376f67c9cbfSJoe Stringer * of *bpf_fd*: 377f67c9cbfSJoe Stringer * 378f67c9cbfSJoe Stringer * * **struct bpf_prog_info** 379f67c9cbfSJoe Stringer * * **struct bpf_map_info** 380f67c9cbfSJoe Stringer * * **struct bpf_btf_info** 381f67c9cbfSJoe Stringer * * **struct bpf_link_info** 382f67c9cbfSJoe Stringer * 383f67c9cbfSJoe Stringer * Return 384f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 385f67c9cbfSJoe Stringer * is set appropriately. 386f67c9cbfSJoe Stringer * 387f67c9cbfSJoe Stringer * BPF_PROG_QUERY 388f67c9cbfSJoe Stringer * Description 389f67c9cbfSJoe Stringer * Obtain information about eBPF programs associated with the 390f67c9cbfSJoe Stringer * specified *attach_type* hook. 391f67c9cbfSJoe Stringer * 3925d999994SJoe Stringer * The *target_fd* must be a valid file descriptor for a kernel 3935d999994SJoe Stringer * object which depends on the attach type of *attach_bpf_fd*: 3945d999994SJoe Stringer * 3955d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_DEVICE**, 3965d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SKB**, 3975d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK**, 3985d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 3995d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 4005d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 4015d999994SJoe Stringer * **BPF_PROG_TYPE_SOCK_OPS** 4025d999994SJoe Stringer * 4035d999994SJoe Stringer * Control Group v2 hierarchy with the eBPF controller 4045d999994SJoe Stringer * enabled. Requires the kernel to be compiled with 4055d999994SJoe Stringer * **CONFIG_CGROUP_BPF**. 4065d999994SJoe Stringer * 4075d999994SJoe Stringer * **BPF_PROG_TYPE_FLOW_DISSECTOR** 4085d999994SJoe Stringer * 4095d999994SJoe Stringer * Network namespace (eg /proc/self/ns/net). 4105d999994SJoe Stringer * 4115d999994SJoe Stringer * **BPF_PROG_TYPE_LIRC_MODE2** 4125d999994SJoe Stringer * 4135d999994SJoe Stringer * LIRC device path (eg /dev/lircN). Requires the kernel 4145d999994SJoe Stringer * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 4155d999994SJoe Stringer * 4165d999994SJoe Stringer * **BPF_PROG_QUERY** always fetches the number of programs 4175d999994SJoe Stringer * attached and the *attach_flags* which were used to attach those 4185d999994SJoe Stringer * programs. Additionally, if *prog_ids* is nonzero and the number 4195d999994SJoe Stringer * of attached programs is less than *prog_cnt*, populates 4205d999994SJoe Stringer * *prog_ids* with the eBPF program ids of the programs attached 4215d999994SJoe Stringer * at *target_fd*. 4225d999994SJoe Stringer * 4235d999994SJoe Stringer * The following flags may alter the result: 4245d999994SJoe Stringer * 4255d999994SJoe Stringer * **BPF_F_QUERY_EFFECTIVE** 4265d999994SJoe Stringer * Only return information regarding programs which are 4275d999994SJoe Stringer * currently effective at the specified *target_fd*. 4285d999994SJoe Stringer * 429f67c9cbfSJoe Stringer * Return 430f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 431f67c9cbfSJoe Stringer * is set appropriately. 432f67c9cbfSJoe Stringer * 433f67c9cbfSJoe Stringer * BPF_RAW_TRACEPOINT_OPEN 434f67c9cbfSJoe Stringer * Description 435f67c9cbfSJoe Stringer * Attach an eBPF program to a tracepoint *name* to access kernel 436f67c9cbfSJoe Stringer * internal arguments of the tracepoint in their raw form. 437f67c9cbfSJoe Stringer * 438f67c9cbfSJoe Stringer * The *prog_fd* must be a valid file descriptor associated with 439f67c9cbfSJoe Stringer * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 440f67c9cbfSJoe Stringer * 441f67c9cbfSJoe Stringer * No ABI guarantees are made about the content of tracepoint 442f67c9cbfSJoe Stringer * arguments exposed to the corresponding eBPF program. 443f67c9cbfSJoe Stringer * 444f67c9cbfSJoe Stringer * Applying **close**\ (2) to the file descriptor returned by 445f67c9cbfSJoe Stringer * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 446f67c9cbfSJoe Stringer * 447f67c9cbfSJoe Stringer * Return 448f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 449f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 450f67c9cbfSJoe Stringer * 451f67c9cbfSJoe Stringer * BPF_BTF_LOAD 452f67c9cbfSJoe Stringer * Description 453f67c9cbfSJoe Stringer * Verify and load BPF Type Format (BTF) metadata into the kernel, 454f67c9cbfSJoe Stringer * returning a new file descriptor associated with the metadata. 455f67c9cbfSJoe Stringer * BTF is described in more detail at 456f67c9cbfSJoe Stringer * https://www.kernel.org/doc/html/latest/bpf/btf.html. 457f67c9cbfSJoe Stringer * 458f67c9cbfSJoe Stringer * The *btf* parameter must point to valid memory providing 459f67c9cbfSJoe Stringer * *btf_size* bytes of BTF binary metadata. 460f67c9cbfSJoe Stringer * 461f67c9cbfSJoe Stringer * The returned file descriptor can be passed to other **bpf**\ () 462f67c9cbfSJoe Stringer * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 463f67c9cbfSJoe Stringer * associate the BTF with those objects. 464f67c9cbfSJoe Stringer * 465f67c9cbfSJoe Stringer * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 466f67c9cbfSJoe Stringer * parameters to specify a *btf_log_buf*, *btf_log_size* and 467f67c9cbfSJoe Stringer * *btf_log_level* which allow the kernel to return freeform log 468f67c9cbfSJoe Stringer * output regarding the BTF verification process. 469f67c9cbfSJoe Stringer * 470f67c9cbfSJoe Stringer * Return 471f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 472f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 473f67c9cbfSJoe Stringer * 474f67c9cbfSJoe Stringer * BPF_BTF_GET_FD_BY_ID 475f67c9cbfSJoe Stringer * Description 476f67c9cbfSJoe Stringer * Open a file descriptor for the BPF Type Format (BTF) 477f67c9cbfSJoe Stringer * corresponding to *btf_id*. 478f67c9cbfSJoe Stringer * 479f67c9cbfSJoe Stringer * Return 480f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 481f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 482f67c9cbfSJoe Stringer * 483f67c9cbfSJoe Stringer * BPF_TASK_FD_QUERY 484f67c9cbfSJoe Stringer * Description 485f67c9cbfSJoe Stringer * Obtain information about eBPF programs associated with the 486f67c9cbfSJoe Stringer * target process identified by *pid* and *fd*. 487f67c9cbfSJoe Stringer * 488f67c9cbfSJoe Stringer * If the *pid* and *fd* are associated with a tracepoint, kprobe 489f67c9cbfSJoe Stringer * or uprobe perf event, then the *prog_id* and *fd_type* will 490f67c9cbfSJoe Stringer * be populated with the eBPF program id and file descriptor type 491f67c9cbfSJoe Stringer * of type **bpf_task_fd_type**. If associated with a kprobe or 492f67c9cbfSJoe Stringer * uprobe, the *probe_offset* and *probe_addr* will also be 493f67c9cbfSJoe Stringer * populated. Optionally, if *buf* is provided, then up to 494f67c9cbfSJoe Stringer * *buf_len* bytes of *buf* will be populated with the name of 495f67c9cbfSJoe Stringer * the tracepoint, kprobe or uprobe. 496f67c9cbfSJoe Stringer * 497f67c9cbfSJoe Stringer * The resulting *prog_id* may be introspected in deeper detail 498f67c9cbfSJoe Stringer * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 499f67c9cbfSJoe Stringer * 500f67c9cbfSJoe Stringer * Return 501f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 502f67c9cbfSJoe Stringer * is set appropriately. 503f67c9cbfSJoe Stringer * 504f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_AND_DELETE_ELEM 505f67c9cbfSJoe Stringer * Description 506f67c9cbfSJoe Stringer * Look up an element with the given *key* in the map referred to 507f67c9cbfSJoe Stringer * by the file descriptor *fd*, and if found, delete the element. 508f67c9cbfSJoe Stringer * 509f67c9cbfSJoe Stringer * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 510f67c9cbfSJoe Stringer * implement this command as a "pop" operation, deleting the top 511f67c9cbfSJoe Stringer * element rather than one corresponding to *key*. 512f67c9cbfSJoe Stringer * The *key* and *key_len* parameters should be zeroed when 513f67c9cbfSJoe Stringer * issuing this operation for these map types. 514f67c9cbfSJoe Stringer * 515f67c9cbfSJoe Stringer * This command is only valid for the following map types: 516f67c9cbfSJoe Stringer * * **BPF_MAP_TYPE_QUEUE** 517f67c9cbfSJoe Stringer * * **BPF_MAP_TYPE_STACK** 518f67c9cbfSJoe Stringer * 519f67c9cbfSJoe Stringer * Return 520f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 521f67c9cbfSJoe Stringer * is set appropriately. 522f67c9cbfSJoe Stringer * 523f67c9cbfSJoe Stringer * BPF_MAP_FREEZE 524f67c9cbfSJoe Stringer * Description 525f67c9cbfSJoe Stringer * Freeze the permissions of the specified map. 526f67c9cbfSJoe Stringer * 527f67c9cbfSJoe Stringer * Write permissions may be frozen by passing zero *flags*. 528f67c9cbfSJoe Stringer * Upon success, no future syscall invocations may alter the 529f67c9cbfSJoe Stringer * map state of *map_fd*. Write operations from eBPF programs 530f67c9cbfSJoe Stringer * are still possible for a frozen map. 531f67c9cbfSJoe Stringer * 532f67c9cbfSJoe Stringer * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 533f67c9cbfSJoe Stringer * 534f67c9cbfSJoe Stringer * Return 535f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 536f67c9cbfSJoe Stringer * is set appropriately. 537f67c9cbfSJoe Stringer * 538f67c9cbfSJoe Stringer * BPF_BTF_GET_NEXT_ID 539f67c9cbfSJoe Stringer * Description 540f67c9cbfSJoe Stringer * Fetch the next BPF Type Format (BTF) object currently loaded 541f67c9cbfSJoe Stringer * into the kernel. 542f67c9cbfSJoe Stringer * 543f67c9cbfSJoe Stringer * Looks for the BTF object with an id greater than *start_id* 544f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other BTF objects 545f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 546f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 547f67c9cbfSJoe Stringer * 548f67c9cbfSJoe Stringer * Return 549f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 550f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 551f67c9cbfSJoe Stringer * 552f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_BATCH 553f67c9cbfSJoe Stringer * Description 554f67c9cbfSJoe Stringer * Iterate and fetch multiple elements in a map. 555f67c9cbfSJoe Stringer * 556*0cb80454SJoe Stringer * Two opaque values are used to manage batch operations, 557*0cb80454SJoe Stringer * *in_batch* and *out_batch*. Initially, *in_batch* must be set 558*0cb80454SJoe Stringer * to NULL to begin the batched operation. After each subsequent 559*0cb80454SJoe Stringer * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 560*0cb80454SJoe Stringer * *out_batch* as the *in_batch* for the next operation to 561*0cb80454SJoe Stringer * continue iteration from the current point. 562*0cb80454SJoe Stringer * 563*0cb80454SJoe Stringer * The *keys* and *values* are output parameters which must point 564*0cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 565*0cb80454SJoe Stringer * and value size of the map *map_fd*. The *keys* buffer must be 566*0cb80454SJoe Stringer * of *key_size* * *count*. The *values* buffer must be of 567*0cb80454SJoe Stringer * *value_size* * *count*. 568*0cb80454SJoe Stringer * 569*0cb80454SJoe Stringer * The *elem_flags* argument may be specified as one of the 570*0cb80454SJoe Stringer * following: 571*0cb80454SJoe Stringer * 572*0cb80454SJoe Stringer * **BPF_F_LOCK** 573*0cb80454SJoe Stringer * Look up the value of a spin-locked map without 574*0cb80454SJoe Stringer * returning the lock. This must be specified if the 575*0cb80454SJoe Stringer * elements contain a spinlock. 576*0cb80454SJoe Stringer * 577*0cb80454SJoe Stringer * On success, *count* elements from the map are copied into the 578*0cb80454SJoe Stringer * user buffer, with the keys copied into *keys* and the values 579*0cb80454SJoe Stringer * copied into the corresponding indices in *values*. 580*0cb80454SJoe Stringer * 581*0cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 582*0cb80454SJoe Stringer * is set to the number of successfully processed elements. 583*0cb80454SJoe Stringer * 584f67c9cbfSJoe Stringer * Return 585f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 586f67c9cbfSJoe Stringer * is set appropriately. 587f67c9cbfSJoe Stringer * 588*0cb80454SJoe Stringer * May set *errno* to **ENOSPC** to indicate that *keys* or 589*0cb80454SJoe Stringer * *values* is too small to dump an entire bucket during 590*0cb80454SJoe Stringer * iteration of a hash-based map type. 591*0cb80454SJoe Stringer * 592f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_AND_DELETE_BATCH 593f67c9cbfSJoe Stringer * Description 594*0cb80454SJoe Stringer * Iterate and delete all elements in a map. 595*0cb80454SJoe Stringer * 596*0cb80454SJoe Stringer * This operation has the same behavior as 597*0cb80454SJoe Stringer * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 598*0cb80454SJoe Stringer * 599*0cb80454SJoe Stringer * * Every element that is successfully returned is also deleted 600*0cb80454SJoe Stringer * from the map. This is at least *count* elements. Note that 601*0cb80454SJoe Stringer * *count* is both an input and an output parameter. 602*0cb80454SJoe Stringer * * Upon returning with *errno* set to **EFAULT**, up to 603*0cb80454SJoe Stringer * *count* elements may be deleted without returning the keys 604*0cb80454SJoe Stringer * and values of the deleted elements. 605f67c9cbfSJoe Stringer * 606f67c9cbfSJoe Stringer * Return 607f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 608f67c9cbfSJoe Stringer * is set appropriately. 609f67c9cbfSJoe Stringer * 610f67c9cbfSJoe Stringer * BPF_MAP_UPDATE_BATCH 611f67c9cbfSJoe Stringer * Description 612*0cb80454SJoe Stringer * Update multiple elements in a map by *key*. 613*0cb80454SJoe Stringer * 614*0cb80454SJoe Stringer * The *keys* and *values* are input parameters which must point 615*0cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 616*0cb80454SJoe Stringer * and value size of the map *map_fd*. The *keys* buffer must be 617*0cb80454SJoe Stringer * of *key_size* * *count*. The *values* buffer must be of 618*0cb80454SJoe Stringer * *value_size* * *count*. 619*0cb80454SJoe Stringer * 620*0cb80454SJoe Stringer * Each element specified in *keys* is sequentially updated to the 621*0cb80454SJoe Stringer * value in the corresponding index in *values*. The *in_batch* 622*0cb80454SJoe Stringer * and *out_batch* parameters are ignored and should be zeroed. 623*0cb80454SJoe Stringer * 624*0cb80454SJoe Stringer * The *elem_flags* argument should be specified as one of the 625*0cb80454SJoe Stringer * following: 626*0cb80454SJoe Stringer * 627*0cb80454SJoe Stringer * **BPF_ANY** 628*0cb80454SJoe Stringer * Create new elements or update a existing elements. 629*0cb80454SJoe Stringer * **BPF_NOEXIST** 630*0cb80454SJoe Stringer * Create new elements only if they do not exist. 631*0cb80454SJoe Stringer * **BPF_EXIST** 632*0cb80454SJoe Stringer * Update existing elements. 633*0cb80454SJoe Stringer * **BPF_F_LOCK** 634*0cb80454SJoe Stringer * Update spin_lock-ed map elements. This must be 635*0cb80454SJoe Stringer * specified if the map value contains a spinlock. 636*0cb80454SJoe Stringer * 637*0cb80454SJoe Stringer * On success, *count* elements from the map are updated. 638*0cb80454SJoe Stringer * 639*0cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 640*0cb80454SJoe Stringer * is set to the number of successfully processed elements. 641f67c9cbfSJoe Stringer * 642f67c9cbfSJoe Stringer * Return 643f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 644f67c9cbfSJoe Stringer * is set appropriately. 645f67c9cbfSJoe Stringer * 646*0cb80454SJoe Stringer * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 647*0cb80454SJoe Stringer * **E2BIG**. **E2BIG** indicates that the number of elements in 648*0cb80454SJoe Stringer * the map reached the *max_entries* limit specified at map 649*0cb80454SJoe Stringer * creation time. 650*0cb80454SJoe Stringer * 651*0cb80454SJoe Stringer * May set *errno* to one of the following error codes under 652*0cb80454SJoe Stringer * specific circumstances: 653*0cb80454SJoe Stringer * 654*0cb80454SJoe Stringer * **EEXIST** 655*0cb80454SJoe Stringer * If *flags* specifies **BPF_NOEXIST** and the element 656*0cb80454SJoe Stringer * with *key* already exists in the map. 657*0cb80454SJoe Stringer * **ENOENT** 658*0cb80454SJoe Stringer * If *flags* specifies **BPF_EXIST** and the element with 659*0cb80454SJoe Stringer * *key* does not exist in the map. 660*0cb80454SJoe Stringer * 661f67c9cbfSJoe Stringer * BPF_MAP_DELETE_BATCH 662f67c9cbfSJoe Stringer * Description 663*0cb80454SJoe Stringer * Delete multiple elements in a map by *key*. 664*0cb80454SJoe Stringer * 665*0cb80454SJoe Stringer * The *keys* parameter is an input parameter which must point 666*0cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 667*0cb80454SJoe Stringer * size of the map *map_fd*, that is, *key_size* * *count*. 668*0cb80454SJoe Stringer * 669*0cb80454SJoe Stringer * Each element specified in *keys* is sequentially deleted. The 670*0cb80454SJoe Stringer * *in_batch*, *out_batch*, and *values* parameters are ignored 671*0cb80454SJoe Stringer * and should be zeroed. 672*0cb80454SJoe Stringer * 673*0cb80454SJoe Stringer * The *elem_flags* argument may be specified as one of the 674*0cb80454SJoe Stringer * following: 675*0cb80454SJoe Stringer * 676*0cb80454SJoe Stringer * **BPF_F_LOCK** 677*0cb80454SJoe Stringer * Look up the value of a spin-locked map without 678*0cb80454SJoe Stringer * returning the lock. This must be specified if the 679*0cb80454SJoe Stringer * elements contain a spinlock. 680*0cb80454SJoe Stringer * 681*0cb80454SJoe Stringer * On success, *count* elements from the map are updated. 682*0cb80454SJoe Stringer * 683*0cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 684*0cb80454SJoe Stringer * is set to the number of successfully processed elements. If 685*0cb80454SJoe Stringer * *errno* is **EFAULT**, up to *count* elements may be been 686*0cb80454SJoe Stringer * deleted. 687f67c9cbfSJoe Stringer * 688f67c9cbfSJoe Stringer * Return 689f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 690f67c9cbfSJoe Stringer * is set appropriately. 691f67c9cbfSJoe Stringer * 692f67c9cbfSJoe Stringer * BPF_LINK_CREATE 693f67c9cbfSJoe Stringer * Description 694f67c9cbfSJoe Stringer * Attach an eBPF program to a *target_fd* at the specified 695f67c9cbfSJoe Stringer * *attach_type* hook and return a file descriptor handle for 696f67c9cbfSJoe Stringer * managing the link. 697f67c9cbfSJoe Stringer * 698f67c9cbfSJoe Stringer * Return 699f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 700f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 701f67c9cbfSJoe Stringer * 702f67c9cbfSJoe Stringer * BPF_LINK_UPDATE 703f67c9cbfSJoe Stringer * Description 704f67c9cbfSJoe Stringer * Update the eBPF program in the specified *link_fd* to 705f67c9cbfSJoe Stringer * *new_prog_fd*. 706f67c9cbfSJoe Stringer * 707f67c9cbfSJoe Stringer * Return 708f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 709f67c9cbfSJoe Stringer * is set appropriately. 710f67c9cbfSJoe Stringer * 711f67c9cbfSJoe Stringer * BPF_LINK_GET_FD_BY_ID 712f67c9cbfSJoe Stringer * Description 713f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF Link corresponding to 714f67c9cbfSJoe Stringer * *link_id*. 715f67c9cbfSJoe Stringer * 716f67c9cbfSJoe Stringer * Return 717f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 718f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 719f67c9cbfSJoe Stringer * 720f67c9cbfSJoe Stringer * BPF_LINK_GET_NEXT_ID 721f67c9cbfSJoe Stringer * Description 722f67c9cbfSJoe Stringer * Fetch the next eBPF link currently loaded into the kernel. 723f67c9cbfSJoe Stringer * 724f67c9cbfSJoe Stringer * Looks for the eBPF link with an id greater than *start_id* 725f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF links 726f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 727f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 728f67c9cbfSJoe Stringer * 729f67c9cbfSJoe Stringer * Return 730f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 731f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 732f67c9cbfSJoe Stringer * 733f67c9cbfSJoe Stringer * BPF_ENABLE_STATS 734f67c9cbfSJoe Stringer * Description 735f67c9cbfSJoe Stringer * Enable eBPF runtime statistics gathering. 736f67c9cbfSJoe Stringer * 737f67c9cbfSJoe Stringer * Runtime statistics gathering for the eBPF runtime is disabled 738f67c9cbfSJoe Stringer * by default to minimize the corresponding performance overhead. 739f67c9cbfSJoe Stringer * This command enables statistics globally. 740f67c9cbfSJoe Stringer * 741f67c9cbfSJoe Stringer * Multiple programs may independently enable statistics. 742f67c9cbfSJoe Stringer * After gathering the desired statistics, eBPF runtime statistics 743f67c9cbfSJoe Stringer * may be disabled again by calling **close**\ (2) for the file 744f67c9cbfSJoe Stringer * descriptor returned by this function. Statistics will only be 745f67c9cbfSJoe Stringer * disabled system-wide when all outstanding file descriptors 746f67c9cbfSJoe Stringer * returned by prior calls for this subcommand are closed. 747f67c9cbfSJoe Stringer * 748f67c9cbfSJoe Stringer * Return 749f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 750f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 751f67c9cbfSJoe Stringer * 752f67c9cbfSJoe Stringer * BPF_ITER_CREATE 753f67c9cbfSJoe Stringer * Description 754f67c9cbfSJoe Stringer * Create an iterator on top of the specified *link_fd* (as 755f67c9cbfSJoe Stringer * previously created using **BPF_LINK_CREATE**) and return a 756f67c9cbfSJoe Stringer * file descriptor that can be used to trigger the iteration. 757f67c9cbfSJoe Stringer * 758f67c9cbfSJoe Stringer * If the resulting file descriptor is pinned to the filesystem 759f67c9cbfSJoe Stringer * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 760f67c9cbfSJoe Stringer * for that path will trigger the iterator to read kernel state 761f67c9cbfSJoe Stringer * using the eBPF program attached to *link_fd*. 762f67c9cbfSJoe Stringer * 763f67c9cbfSJoe Stringer * Return 764f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 765f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 766f67c9cbfSJoe Stringer * 767f67c9cbfSJoe Stringer * BPF_LINK_DETACH 768f67c9cbfSJoe Stringer * Description 769f67c9cbfSJoe Stringer * Forcefully detach the specified *link_fd* from its 770f67c9cbfSJoe Stringer * corresponding attachment point. 771f67c9cbfSJoe Stringer * 772f67c9cbfSJoe Stringer * Return 773f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 774f67c9cbfSJoe Stringer * is set appropriately. 775f67c9cbfSJoe Stringer * 776f67c9cbfSJoe Stringer * BPF_PROG_BIND_MAP 777f67c9cbfSJoe Stringer * Description 778f67c9cbfSJoe Stringer * Bind a map to the lifetime of an eBPF program. 779f67c9cbfSJoe Stringer * 780f67c9cbfSJoe Stringer * The map identified by *map_fd* is bound to the program 781f67c9cbfSJoe Stringer * identified by *prog_fd* and only released when *prog_fd* is 782f67c9cbfSJoe Stringer * released. This may be used in cases where metadata should be 783f67c9cbfSJoe Stringer * associated with a program which otherwise does not contain any 784f67c9cbfSJoe Stringer * references to the map (for example, embedded in the eBPF 785f67c9cbfSJoe Stringer * program instructions). 786f67c9cbfSJoe Stringer * 787f67c9cbfSJoe Stringer * Return 788f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 789f67c9cbfSJoe Stringer * is set appropriately. 790f67c9cbfSJoe Stringer * 7917799e4d9SJoe Stringer * NOTES 7927799e4d9SJoe Stringer * eBPF objects (maps and programs) can be shared between processes. 7938aacb3c8SJoe Stringer * 7948aacb3c8SJoe Stringer * * After **fork**\ (2), the child inherits file descriptors 7958aacb3c8SJoe Stringer * referring to the same eBPF objects. 7968aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be transferred over 7978aacb3c8SJoe Stringer * **unix**\ (7) domain sockets. 7988aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be duplicated in the 7998aacb3c8SJoe Stringer * usual way, using **dup**\ (2) and similar calls. 8008aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be pinned to the 8018aacb3c8SJoe Stringer * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 8028aacb3c8SJoe Stringer * 8038aacb3c8SJoe Stringer * An eBPF object is deallocated only after all file descriptors referring 8048aacb3c8SJoe Stringer * to the object have been closed and no references remain pinned to the 8058aacb3c8SJoe Stringer * filesystem or attached (for example, bound to a program or device). 8067799e4d9SJoe Stringer */ 80799c55f7dSAlexei Starovoitov enum bpf_cmd { 80899c55f7dSAlexei Starovoitov BPF_MAP_CREATE, 809db20fd2bSAlexei Starovoitov BPF_MAP_LOOKUP_ELEM, 810db20fd2bSAlexei Starovoitov BPF_MAP_UPDATE_ELEM, 811db20fd2bSAlexei Starovoitov BPF_MAP_DELETE_ELEM, 812db20fd2bSAlexei Starovoitov BPF_MAP_GET_NEXT_KEY, 81309756af4SAlexei Starovoitov BPF_PROG_LOAD, 814b2197755SDaniel Borkmann BPF_OBJ_PIN, 815b2197755SDaniel Borkmann BPF_OBJ_GET, 816f4324551SDaniel Mack BPF_PROG_ATTACH, 817f4324551SDaniel Mack BPF_PROG_DETACH, 8181cf1cae9SAlexei Starovoitov BPF_PROG_TEST_RUN, 81934ad5580SMartin KaFai Lau BPF_PROG_GET_NEXT_ID, 82034ad5580SMartin KaFai Lau BPF_MAP_GET_NEXT_ID, 821b16d9aa4SMartin KaFai Lau BPF_PROG_GET_FD_BY_ID, 822bd5f5f4eSMartin KaFai Lau BPF_MAP_GET_FD_BY_ID, 8231e270976SMartin KaFai Lau BPF_OBJ_GET_INFO_BY_FD, 824468e2f64SAlexei Starovoitov BPF_PROG_QUERY, 825c4f6699dSAlexei Starovoitov BPF_RAW_TRACEPOINT_OPEN, 826f56a653cSMartin KaFai Lau BPF_BTF_LOAD, 82778958fcaSMartin KaFai Lau BPF_BTF_GET_FD_BY_ID, 82841bdc4b4SYonghong Song BPF_TASK_FD_QUERY, 829bd513cd0SMauricio Vasquez B BPF_MAP_LOOKUP_AND_DELETE_ELEM, 83087df15deSDaniel Borkmann BPF_MAP_FREEZE, 8311b9ed84eSQuentin Monnet BPF_BTF_GET_NEXT_ID, 832cb4d03abSBrian Vazquez BPF_MAP_LOOKUP_BATCH, 83305799638SYonghong Song BPF_MAP_LOOKUP_AND_DELETE_BATCH, 834aa2e93b8SBrian Vazquez BPF_MAP_UPDATE_BATCH, 835aa2e93b8SBrian Vazquez BPF_MAP_DELETE_BATCH, 836af6eea57SAndrii Nakryiko BPF_LINK_CREATE, 8370c991ebcSAndrii Nakryiko BPF_LINK_UPDATE, 8382d602c8cSAndrii Nakryiko BPF_LINK_GET_FD_BY_ID, 8392d602c8cSAndrii Nakryiko BPF_LINK_GET_NEXT_ID, 840d46edd67SSong Liu BPF_ENABLE_STATS, 841ac51d99bSYonghong Song BPF_ITER_CREATE, 84273b11c2aSAndrii Nakryiko BPF_LINK_DETACH, 843ef15314aSYiFei Zhu BPF_PROG_BIND_MAP, 84499c55f7dSAlexei Starovoitov }; 84599c55f7dSAlexei Starovoitov 84699c55f7dSAlexei Starovoitov enum bpf_map_type { 84799c55f7dSAlexei Starovoitov BPF_MAP_TYPE_UNSPEC, 8480f8e4bd8SAlexei Starovoitov BPF_MAP_TYPE_HASH, 84928fbcfa0SAlexei Starovoitov BPF_MAP_TYPE_ARRAY, 85004fd61abSAlexei Starovoitov BPF_MAP_TYPE_PROG_ARRAY, 851ea317b26SKaixu Xia BPF_MAP_TYPE_PERF_EVENT_ARRAY, 852824bd0ceSAlexei Starovoitov BPF_MAP_TYPE_PERCPU_HASH, 853a10423b8SAlexei Starovoitov BPF_MAP_TYPE_PERCPU_ARRAY, 854d5a3b1f6SAlexei Starovoitov BPF_MAP_TYPE_STACK_TRACE, 8554ed8ec52SMartin KaFai Lau BPF_MAP_TYPE_CGROUP_ARRAY, 85629ba732aSMartin KaFai Lau BPF_MAP_TYPE_LRU_HASH, 8578f844938SMartin KaFai Lau BPF_MAP_TYPE_LRU_PERCPU_HASH, 858b95a5c4dSDaniel Mack BPF_MAP_TYPE_LPM_TRIE, 85956f668dfSMartin KaFai Lau BPF_MAP_TYPE_ARRAY_OF_MAPS, 860bcc6b1b7SMartin KaFai Lau BPF_MAP_TYPE_HASH_OF_MAPS, 861546ac1ffSJohn Fastabend BPF_MAP_TYPE_DEVMAP, 862174a79ffSJohn Fastabend BPF_MAP_TYPE_SOCKMAP, 8636710e112SJesper Dangaard Brouer BPF_MAP_TYPE_CPUMAP, 864fbfc504aSBjörn Töpel BPF_MAP_TYPE_XSKMAP, 86581110384SJohn Fastabend BPF_MAP_TYPE_SOCKHASH, 866de9cbbaaSRoman Gushchin BPF_MAP_TYPE_CGROUP_STORAGE, 8675dc4c4b7SMartin KaFai Lau BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 868b741f163SRoman Gushchin BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 869f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_QUEUE, 870f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_STACK, 8716ac99e8fSMartin KaFai Lau BPF_MAP_TYPE_SK_STORAGE, 8726f9d451aSToke Høiland-Jørgensen BPF_MAP_TYPE_DEVMAP_HASH, 87385d33df3SMartin KaFai Lau BPF_MAP_TYPE_STRUCT_OPS, 874457f4436SAndrii Nakryiko BPF_MAP_TYPE_RINGBUF, 8758ea63684SKP Singh BPF_MAP_TYPE_INODE_STORAGE, 8764cf1bc1fSKP Singh BPF_MAP_TYPE_TASK_STORAGE, 87799c55f7dSAlexei Starovoitov }; 87899c55f7dSAlexei Starovoitov 8796c4fc209SDaniel Borkmann /* Note that tracing related programs such as 8806c4fc209SDaniel Borkmann * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 8816c4fc209SDaniel Borkmann * are not subject to a stable API since kernel internal data 8826c4fc209SDaniel Borkmann * structures can change from release to release and may 8836c4fc209SDaniel Borkmann * therefore break existing tracing BPF programs. Tracing BPF 8846c4fc209SDaniel Borkmann * programs correspond to /a/ specific kernel which is to be 8856c4fc209SDaniel Borkmann * analyzed, and not /a/ specific kernel /and/ all future ones. 8866c4fc209SDaniel Borkmann */ 88709756af4SAlexei Starovoitov enum bpf_prog_type { 88809756af4SAlexei Starovoitov BPF_PROG_TYPE_UNSPEC, 889ddd872bcSAlexei Starovoitov BPF_PROG_TYPE_SOCKET_FILTER, 8902541517cSAlexei Starovoitov BPF_PROG_TYPE_KPROBE, 89196be4325SDaniel Borkmann BPF_PROG_TYPE_SCHED_CLS, 89294caee8cSDaniel Borkmann BPF_PROG_TYPE_SCHED_ACT, 89398b5c2c6SAlexei Starovoitov BPF_PROG_TYPE_TRACEPOINT, 8946a773a15SBrenden Blanco BPF_PROG_TYPE_XDP, 8950515e599SAlexei Starovoitov BPF_PROG_TYPE_PERF_EVENT, 8960e33661dSDaniel Mack BPF_PROG_TYPE_CGROUP_SKB, 89761023658SDavid Ahern BPF_PROG_TYPE_CGROUP_SOCK, 8983a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_IN, 8993a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_OUT, 9003a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_XMIT, 90140304b2aSLawrence Brakmo BPF_PROG_TYPE_SOCK_OPS, 902b005fd18SJohn Fastabend BPF_PROG_TYPE_SK_SKB, 903ebc614f6SRoman Gushchin BPF_PROG_TYPE_CGROUP_DEVICE, 9044f738adbSJohn Fastabend BPF_PROG_TYPE_SK_MSG, 905c4f6699dSAlexei Starovoitov BPF_PROG_TYPE_RAW_TRACEPOINT, 9064fbac77dSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 907004d4b27SMathieu Xhonneux BPF_PROG_TYPE_LWT_SEG6LOCAL, 908f4364dcfSSean Young BPF_PROG_TYPE_LIRC_MODE2, 9092dbb9b9eSMartin KaFai Lau BPF_PROG_TYPE_SK_REUSEPORT, 910d58e468bSPetar Penkov BPF_PROG_TYPE_FLOW_DISSECTOR, 9117b146cebSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SYSCTL, 9129df1c28bSMatt Mullins BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 9130d01da6aSStanislav Fomichev BPF_PROG_TYPE_CGROUP_SOCKOPT, 914f1b9509cSAlexei Starovoitov BPF_PROG_TYPE_TRACING, 91527ae7997SMartin KaFai Lau BPF_PROG_TYPE_STRUCT_OPS, 916be8704ffSAlexei Starovoitov BPF_PROG_TYPE_EXT, 917fc611f47SKP Singh BPF_PROG_TYPE_LSM, 918e9ddbb77SJakub Sitnicki BPF_PROG_TYPE_SK_LOOKUP, 91909756af4SAlexei Starovoitov }; 92009756af4SAlexei Starovoitov 9210e33661dSDaniel Mack enum bpf_attach_type { 9220e33661dSDaniel Mack BPF_CGROUP_INET_INGRESS, 9230e33661dSDaniel Mack BPF_CGROUP_INET_EGRESS, 92461023658SDavid Ahern BPF_CGROUP_INET_SOCK_CREATE, 92540304b2aSLawrence Brakmo BPF_CGROUP_SOCK_OPS, 926464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_PARSER, 927464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_VERDICT, 928ebc614f6SRoman Gushchin BPF_CGROUP_DEVICE, 9294f738adbSJohn Fastabend BPF_SK_MSG_VERDICT, 9304fbac77dSAndrey Ignatov BPF_CGROUP_INET4_BIND, 9314fbac77dSAndrey Ignatov BPF_CGROUP_INET6_BIND, 932d74bad4eSAndrey Ignatov BPF_CGROUP_INET4_CONNECT, 933d74bad4eSAndrey Ignatov BPF_CGROUP_INET6_CONNECT, 934aac3fc32SAndrey Ignatov BPF_CGROUP_INET4_POST_BIND, 935aac3fc32SAndrey Ignatov BPF_CGROUP_INET6_POST_BIND, 9361cedee13SAndrey Ignatov BPF_CGROUP_UDP4_SENDMSG, 9371cedee13SAndrey Ignatov BPF_CGROUP_UDP6_SENDMSG, 938f4364dcfSSean Young BPF_LIRC_MODE2, 939d58e468bSPetar Penkov BPF_FLOW_DISSECTOR, 9407b146cebSAndrey Ignatov BPF_CGROUP_SYSCTL, 941983695faSDaniel Borkmann BPF_CGROUP_UDP4_RECVMSG, 942983695faSDaniel Borkmann BPF_CGROUP_UDP6_RECVMSG, 9430d01da6aSStanislav Fomichev BPF_CGROUP_GETSOCKOPT, 9440d01da6aSStanislav Fomichev BPF_CGROUP_SETSOCKOPT, 945f1b9509cSAlexei Starovoitov BPF_TRACE_RAW_TP, 946fec56f58SAlexei Starovoitov BPF_TRACE_FENTRY, 947fec56f58SAlexei Starovoitov BPF_TRACE_FEXIT, 948ae240823SKP Singh BPF_MODIFY_RETURN, 949fc611f47SKP Singh BPF_LSM_MAC, 95015d83c4dSYonghong Song BPF_TRACE_ITER, 9511b66d253SDaniel Borkmann BPF_CGROUP_INET4_GETPEERNAME, 9521b66d253SDaniel Borkmann BPF_CGROUP_INET6_GETPEERNAME, 9531b66d253SDaniel Borkmann BPF_CGROUP_INET4_GETSOCKNAME, 9541b66d253SDaniel Borkmann BPF_CGROUP_INET6_GETSOCKNAME, 955fbee97feSDavid Ahern BPF_XDP_DEVMAP, 956f5836749SStanislav Fomichev BPF_CGROUP_INET_SOCK_RELEASE, 95792164774SLorenzo Bianconi BPF_XDP_CPUMAP, 958e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP, 959aa8d3a71SAndrii Nakryiko BPF_XDP, 9600e33661dSDaniel Mack __MAX_BPF_ATTACH_TYPE 9610e33661dSDaniel Mack }; 9620e33661dSDaniel Mack 9630e33661dSDaniel Mack #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 9640e33661dSDaniel Mack 965f2e10bffSAndrii Nakryiko enum bpf_link_type { 966f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_UNSPEC = 0, 967f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 968f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_TRACING = 2, 969f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_CGROUP = 3, 970de4e05caSYonghong Song BPF_LINK_TYPE_ITER = 4, 9717f045a49SJakub Sitnicki BPF_LINK_TYPE_NETNS = 5, 972aa8d3a71SAndrii Nakryiko BPF_LINK_TYPE_XDP = 6, 973f2e10bffSAndrii Nakryiko 974f2e10bffSAndrii Nakryiko MAX_BPF_LINK_TYPE, 975f2e10bffSAndrii Nakryiko }; 976f2e10bffSAndrii Nakryiko 977324bda9eSAlexei Starovoitov /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 978324bda9eSAlexei Starovoitov * 979324bda9eSAlexei Starovoitov * NONE(default): No further bpf programs allowed in the subtree. 980324bda9eSAlexei Starovoitov * 981324bda9eSAlexei Starovoitov * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 982324bda9eSAlexei Starovoitov * the program in this cgroup yields to sub-cgroup program. 983324bda9eSAlexei Starovoitov * 984324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 985324bda9eSAlexei Starovoitov * that cgroup program gets run in addition to the program in this cgroup. 986324bda9eSAlexei Starovoitov * 987324bda9eSAlexei Starovoitov * Only one program is allowed to be attached to a cgroup with 988324bda9eSAlexei Starovoitov * NONE or BPF_F_ALLOW_OVERRIDE flag. 989324bda9eSAlexei Starovoitov * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 990324bda9eSAlexei Starovoitov * release old program and attach the new one. Attach flags has to match. 991324bda9eSAlexei Starovoitov * 992324bda9eSAlexei Starovoitov * Multiple programs are allowed to be attached to a cgroup with 993324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 994324bda9eSAlexei Starovoitov * (those that were attached first, run first) 995324bda9eSAlexei Starovoitov * The programs of sub-cgroup are executed first, then programs of 996324bda9eSAlexei Starovoitov * this cgroup and then programs of parent cgroup. 997324bda9eSAlexei Starovoitov * When children program makes decision (like picking TCP CA or sock bind) 998324bda9eSAlexei Starovoitov * parent program has a chance to override it. 999324bda9eSAlexei Starovoitov * 10007dd68b32SAndrey Ignatov * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 10017dd68b32SAndrey Ignatov * programs for a cgroup. Though it's possible to replace an old program at 10027dd68b32SAndrey Ignatov * any position by also specifying BPF_F_REPLACE flag and position itself in 10037dd68b32SAndrey Ignatov * replace_bpf_fd attribute. Old program at this position will be released. 10047dd68b32SAndrey Ignatov * 1005324bda9eSAlexei Starovoitov * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1006324bda9eSAlexei Starovoitov * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1007324bda9eSAlexei Starovoitov * Ex1: 1008324bda9eSAlexei Starovoitov * cgrp1 (MULTI progs A, B) -> 1009324bda9eSAlexei Starovoitov * cgrp2 (OVERRIDE prog C) -> 1010324bda9eSAlexei Starovoitov * cgrp3 (MULTI prog D) -> 1011324bda9eSAlexei Starovoitov * cgrp4 (OVERRIDE prog E) -> 1012324bda9eSAlexei Starovoitov * cgrp5 (NONE prog F) 1013324bda9eSAlexei Starovoitov * the event in cgrp5 triggers execution of F,D,A,B in that order. 1014324bda9eSAlexei Starovoitov * if prog F is detached, the execution is E,D,A,B 1015324bda9eSAlexei Starovoitov * if prog F and D are detached, the execution is E,A,B 1016324bda9eSAlexei Starovoitov * if prog F, E and D are detached, the execution is C,A,B 1017324bda9eSAlexei Starovoitov * 1018324bda9eSAlexei Starovoitov * All eligible programs are executed regardless of return code from 1019324bda9eSAlexei Starovoitov * earlier programs. 10207f677633SAlexei Starovoitov */ 10217f677633SAlexei Starovoitov #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1022324bda9eSAlexei Starovoitov #define BPF_F_ALLOW_MULTI (1U << 1) 10237dd68b32SAndrey Ignatov #define BPF_F_REPLACE (1U << 2) 10247f677633SAlexei Starovoitov 1025e07b98d9SDavid S. Miller /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1026e07b98d9SDavid S. Miller * verifier will perform strict alignment checking as if the kernel 1027e07b98d9SDavid S. Miller * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1028e07b98d9SDavid S. Miller * and NET_IP_ALIGN defined to 2. 1029e07b98d9SDavid S. Miller */ 1030e07b98d9SDavid S. Miller #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1031e07b98d9SDavid S. Miller 1032e9ee9efcSDavid Miller /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 1033e9ee9efcSDavid Miller * verifier will allow any alignment whatsoever. On platforms 1034e9ee9efcSDavid Miller * with strict alignment requirements for loads ands stores (such 1035e9ee9efcSDavid Miller * as sparc and mips) the verifier validates that all loads and 1036e9ee9efcSDavid Miller * stores provably follow this requirement. This flag turns that 1037e9ee9efcSDavid Miller * checking and enforcement off. 1038e9ee9efcSDavid Miller * 1039e9ee9efcSDavid Miller * It is mostly used for testing when we want to validate the 1040e9ee9efcSDavid Miller * context and memory access aspects of the verifier, but because 1041e9ee9efcSDavid Miller * of an unaligned access the alignment check would trigger before 1042e9ee9efcSDavid Miller * the one we are interested in. 1043e9ee9efcSDavid Miller */ 1044e9ee9efcSDavid Miller #define BPF_F_ANY_ALIGNMENT (1U << 1) 1045e9ee9efcSDavid Miller 1046c240eff6SJiong Wang /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1047c240eff6SJiong Wang * Verifier does sub-register def/use analysis and identifies instructions whose 1048c240eff6SJiong Wang * def only matters for low 32-bit, high 32-bit is never referenced later 1049c240eff6SJiong Wang * through implicit zero extension. Therefore verifier notifies JIT back-ends 1050c240eff6SJiong Wang * that it is safe to ignore clearing high 32-bit for these instructions. This 1051c240eff6SJiong Wang * saves some back-ends a lot of code-gen. However such optimization is not 1052c240eff6SJiong Wang * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1053c240eff6SJiong Wang * hence hasn't used verifier's analysis result. But, we really want to have a 1054c240eff6SJiong Wang * way to be able to verify the correctness of the described optimization on 1055c240eff6SJiong Wang * x86_64 on which testsuites are frequently exercised. 1056c240eff6SJiong Wang * 1057c240eff6SJiong Wang * So, this flag is introduced. Once it is set, verifier will randomize high 1058c240eff6SJiong Wang * 32-bit for those instructions who has been identified as safe to ignore them. 1059c240eff6SJiong Wang * Then, if verifier is not doing correct analysis, such randomization will 1060c240eff6SJiong Wang * regress tests to expose bugs. 1061c240eff6SJiong Wang */ 1062c240eff6SJiong Wang #define BPF_F_TEST_RND_HI32 (1U << 2) 1063c240eff6SJiong Wang 106410d274e8SAlexei Starovoitov /* The verifier internal test flag. Behavior is undefined */ 106510d274e8SAlexei Starovoitov #define BPF_F_TEST_STATE_FREQ (1U << 3) 106610d274e8SAlexei Starovoitov 10671e6c62a8SAlexei Starovoitov /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 10681e6c62a8SAlexei Starovoitov * restrict map and helper usage for such programs. Sleepable BPF programs can 10691e6c62a8SAlexei Starovoitov * only be attached to hooks where kernel execution context allows sleeping. 10701e6c62a8SAlexei Starovoitov * Such programs are allowed to use helpers that may sleep like 10711e6c62a8SAlexei Starovoitov * bpf_copy_from_user(). 10721e6c62a8SAlexei Starovoitov */ 10731e6c62a8SAlexei Starovoitov #define BPF_F_SLEEPABLE (1U << 4) 10741e6c62a8SAlexei Starovoitov 1075d8eca5bbSDaniel Borkmann /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 10764976b718SHao Luo * the following extensions: 1077d8eca5bbSDaniel Borkmann * 10784976b718SHao Luo * insn[0].src_reg: BPF_PSEUDO_MAP_FD 10794976b718SHao Luo * insn[0].imm: map fd 10804976b718SHao Luo * insn[1].imm: 0 10814976b718SHao Luo * insn[0].off: 0 10824976b718SHao Luo * insn[1].off: 0 10834976b718SHao Luo * ldimm64 rewrite: address of map 10844976b718SHao Luo * verifier type: CONST_PTR_TO_MAP 1085d8eca5bbSDaniel Borkmann */ 1086f1a66f85SDaniel Borkmann #define BPF_PSEUDO_MAP_FD 1 10874976b718SHao Luo /* insn[0].src_reg: BPF_PSEUDO_MAP_VALUE 10884976b718SHao Luo * insn[0].imm: map fd 10894976b718SHao Luo * insn[1].imm: offset into value 10904976b718SHao Luo * insn[0].off: 0 10914976b718SHao Luo * insn[1].off: 0 10924976b718SHao Luo * ldimm64 rewrite: address of map[0]+offset 10934976b718SHao Luo * verifier type: PTR_TO_MAP_VALUE 10944976b718SHao Luo */ 1095d8eca5bbSDaniel Borkmann #define BPF_PSEUDO_MAP_VALUE 2 10964976b718SHao Luo /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 10974976b718SHao Luo * insn[0].imm: kernel btd id of VAR 10984976b718SHao Luo * insn[1].imm: 0 10994976b718SHao Luo * insn[0].off: 0 11004976b718SHao Luo * insn[1].off: 0 11014976b718SHao Luo * ldimm64 rewrite: address of the kernel variable 11024976b718SHao Luo * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 11034976b718SHao Luo * is struct/union. 11044976b718SHao Luo */ 11054976b718SHao Luo #define BPF_PSEUDO_BTF_ID 3 110669c087baSYonghong Song /* insn[0].src_reg: BPF_PSEUDO_FUNC 110769c087baSYonghong Song * insn[0].imm: insn offset to the func 110869c087baSYonghong Song * insn[1].imm: 0 110969c087baSYonghong Song * insn[0].off: 0 111069c087baSYonghong Song * insn[1].off: 0 111169c087baSYonghong Song * ldimm64 rewrite: address of the function 111269c087baSYonghong Song * verifier type: PTR_TO_FUNC. 111369c087baSYonghong Song */ 111469c087baSYonghong Song #define BPF_PSEUDO_FUNC 4 1115f1a66f85SDaniel Borkmann 1116cc8b0b92SAlexei Starovoitov /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1117cc8b0b92SAlexei Starovoitov * offset to another bpf function 1118cc8b0b92SAlexei Starovoitov */ 1119cc8b0b92SAlexei Starovoitov #define BPF_PSEUDO_CALL 1 1120cc8b0b92SAlexei Starovoitov 11213274f520SAlexei Starovoitov /* flags for BPF_MAP_UPDATE_ELEM command */ 11221aae4bddSAndrii Nakryiko enum { 11231aae4bddSAndrii Nakryiko BPF_ANY = 0, /* create new element or update existing */ 11241aae4bddSAndrii Nakryiko BPF_NOEXIST = 1, /* create new element if it didn't exist */ 11251aae4bddSAndrii Nakryiko BPF_EXIST = 2, /* update existing element */ 11261aae4bddSAndrii Nakryiko BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 11271aae4bddSAndrii Nakryiko }; 11283274f520SAlexei Starovoitov 112996eabe7aSMartin KaFai Lau /* flags for BPF_MAP_CREATE command */ 11301aae4bddSAndrii Nakryiko enum { 11311aae4bddSAndrii Nakryiko BPF_F_NO_PREALLOC = (1U << 0), 113229ba732aSMartin KaFai Lau /* Instead of having one common LRU list in the 11338f844938SMartin KaFai Lau * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 113429ba732aSMartin KaFai Lau * which can scale and perform better. 113529ba732aSMartin KaFai Lau * Note, the LRU nodes (including free nodes) cannot be moved 113629ba732aSMartin KaFai Lau * across different LRU lists. 113729ba732aSMartin KaFai Lau */ 11381aae4bddSAndrii Nakryiko BPF_F_NO_COMMON_LRU = (1U << 1), 113996eabe7aSMartin KaFai Lau /* Specify numa node during map creation */ 11401aae4bddSAndrii Nakryiko BPF_F_NUMA_NODE = (1U << 2), 1141cb4d2b3fSMartin KaFai Lau 1142591fe988SDaniel Borkmann /* Flags for accessing BPF object from syscall side. */ 11431aae4bddSAndrii Nakryiko BPF_F_RDONLY = (1U << 3), 11441aae4bddSAndrii Nakryiko BPF_F_WRONLY = (1U << 4), 11456e71b04aSChenbo Feng 1146615755a7SSong Liu /* Flag for stack_map, store build_id+offset instead of pointer */ 11471aae4bddSAndrii Nakryiko BPF_F_STACK_BUILD_ID = (1U << 5), 1148615755a7SSong Liu 114996b3b6c9SLorenz Bauer /* Zero-initialize hash function seed. This should only be used for testing. */ 11501aae4bddSAndrii Nakryiko BPF_F_ZERO_SEED = (1U << 6), 115196b3b6c9SLorenz Bauer 1152591fe988SDaniel Borkmann /* Flags for accessing BPF object from program side. */ 11531aae4bddSAndrii Nakryiko BPF_F_RDONLY_PROG = (1U << 7), 11541aae4bddSAndrii Nakryiko BPF_F_WRONLY_PROG = (1U << 8), 1155591fe988SDaniel Borkmann 11568f51dfc7SStanislav Fomichev /* Clone map from listener for newly accepted socket */ 11571aae4bddSAndrii Nakryiko BPF_F_CLONE = (1U << 9), 11588f51dfc7SStanislav Fomichev 1159fc970227SAndrii Nakryiko /* Enable memory-mapping BPF map */ 11601aae4bddSAndrii Nakryiko BPF_F_MMAPABLE = (1U << 10), 1161792cacccSSong Liu 1162792cacccSSong Liu /* Share perf_event among processes */ 1163792cacccSSong Liu BPF_F_PRESERVE_ELEMS = (1U << 11), 11644a8f87e6SDaniel Borkmann 11654a8f87e6SDaniel Borkmann /* Create a map that is suitable to be an inner map with dynamic max entries */ 11664a8f87e6SDaniel Borkmann BPF_F_INNER_MAP = (1U << 12), 11671aae4bddSAndrii Nakryiko }; 1168fc970227SAndrii Nakryiko 1169f5bfcd95SAndrey Ignatov /* Flags for BPF_PROG_QUERY. */ 1170f5bfcd95SAndrey Ignatov 1171f5bfcd95SAndrey Ignatov /* Query effective (directly attached + inherited from ancestor cgroups) 1172f5bfcd95SAndrey Ignatov * programs that will be executed for events within a cgroup. 1173f5bfcd95SAndrey Ignatov * attach_flags with this flag are returned only for directly attached programs. 1174f5bfcd95SAndrey Ignatov */ 11752f183360SLorenz Bauer #define BPF_F_QUERY_EFFECTIVE (1U << 0) 11762f183360SLorenz Bauer 11771b4d60ecSSong Liu /* Flags for BPF_PROG_TEST_RUN */ 11781b4d60ecSSong Liu 11791b4d60ecSSong Liu /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 11801b4d60ecSSong Liu #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 11811b4d60ecSSong Liu 1182d46edd67SSong Liu /* type for BPF_ENABLE_STATS */ 1183d46edd67SSong Liu enum bpf_stats_type { 1184d46edd67SSong Liu /* enabled run_time_ns and run_cnt */ 1185d46edd67SSong Liu BPF_STATS_RUN_TIME = 0, 1186d46edd67SSong Liu }; 1187d46edd67SSong Liu 1188615755a7SSong Liu enum bpf_stack_build_id_status { 1189615755a7SSong Liu /* user space need an empty entry to identify end of a trace */ 1190615755a7SSong Liu BPF_STACK_BUILD_ID_EMPTY = 0, 1191615755a7SSong Liu /* with valid build_id and offset */ 1192615755a7SSong Liu BPF_STACK_BUILD_ID_VALID = 1, 1193615755a7SSong Liu /* couldn't get build_id, fallback to ip */ 1194615755a7SSong Liu BPF_STACK_BUILD_ID_IP = 2, 1195615755a7SSong Liu }; 1196615755a7SSong Liu 1197615755a7SSong Liu #define BPF_BUILD_ID_SIZE 20 1198615755a7SSong Liu struct bpf_stack_build_id { 1199615755a7SSong Liu __s32 status; 1200615755a7SSong Liu unsigned char build_id[BPF_BUILD_ID_SIZE]; 1201615755a7SSong Liu union { 1202615755a7SSong Liu __u64 offset; 1203615755a7SSong Liu __u64 ip; 1204615755a7SSong Liu }; 1205615755a7SSong Liu }; 1206615755a7SSong Liu 12071aae4bddSAndrii Nakryiko #define BPF_OBJ_NAME_LEN 16U 12081aae4bddSAndrii Nakryiko 120999c55f7dSAlexei Starovoitov union bpf_attr { 121099c55f7dSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_CREATE command */ 121199c55f7dSAlexei Starovoitov __u32 map_type; /* one of enum bpf_map_type */ 121299c55f7dSAlexei Starovoitov __u32 key_size; /* size of key in bytes */ 121399c55f7dSAlexei Starovoitov __u32 value_size; /* size of value in bytes */ 121499c55f7dSAlexei Starovoitov __u32 max_entries; /* max number of entries in a map */ 121596eabe7aSMartin KaFai Lau __u32 map_flags; /* BPF_MAP_CREATE related 121696eabe7aSMartin KaFai Lau * flags defined above. 121796eabe7aSMartin KaFai Lau */ 121856f668dfSMartin KaFai Lau __u32 inner_map_fd; /* fd pointing to the inner map */ 121996eabe7aSMartin KaFai Lau __u32 numa_node; /* numa node (effective only if 122096eabe7aSMartin KaFai Lau * BPF_F_NUMA_NODE is set). 122196eabe7aSMartin KaFai Lau */ 1222067cae47SMartin KaFai Lau char map_name[BPF_OBJ_NAME_LEN]; 1223a3884572SJakub Kicinski __u32 map_ifindex; /* ifindex of netdev to create on */ 1224a26ca7c9SMartin KaFai Lau __u32 btf_fd; /* fd pointing to a BTF type data */ 12259b2cf328SMartin KaFai Lau __u32 btf_key_type_id; /* BTF type_id of the key */ 12269b2cf328SMartin KaFai Lau __u32 btf_value_type_id; /* BTF type_id of the value */ 122785d33df3SMartin KaFai Lau __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 122885d33df3SMartin KaFai Lau * struct stored as the 122985d33df3SMartin KaFai Lau * map value 123085d33df3SMartin KaFai Lau */ 123199c55f7dSAlexei Starovoitov }; 1232db20fd2bSAlexei Starovoitov 1233db20fd2bSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1234db20fd2bSAlexei Starovoitov __u32 map_fd; 1235db20fd2bSAlexei Starovoitov __aligned_u64 key; 1236db20fd2bSAlexei Starovoitov union { 1237db20fd2bSAlexei Starovoitov __aligned_u64 value; 1238db20fd2bSAlexei Starovoitov __aligned_u64 next_key; 1239db20fd2bSAlexei Starovoitov }; 12403274f520SAlexei Starovoitov __u64 flags; 1241db20fd2bSAlexei Starovoitov }; 124209756af4SAlexei Starovoitov 1243cb4d03abSBrian Vazquez struct { /* struct used by BPF_MAP_*_BATCH commands */ 1244cb4d03abSBrian Vazquez __aligned_u64 in_batch; /* start batch, 1245cb4d03abSBrian Vazquez * NULL to start from beginning 1246cb4d03abSBrian Vazquez */ 1247cb4d03abSBrian Vazquez __aligned_u64 out_batch; /* output: next start batch */ 1248cb4d03abSBrian Vazquez __aligned_u64 keys; 1249cb4d03abSBrian Vazquez __aligned_u64 values; 1250cb4d03abSBrian Vazquez __u32 count; /* input/output: 1251cb4d03abSBrian Vazquez * input: # of key/value 1252cb4d03abSBrian Vazquez * elements 1253cb4d03abSBrian Vazquez * output: # of filled elements 1254cb4d03abSBrian Vazquez */ 1255cb4d03abSBrian Vazquez __u32 map_fd; 1256cb4d03abSBrian Vazquez __u64 elem_flags; 1257cb4d03abSBrian Vazquez __u64 flags; 1258cb4d03abSBrian Vazquez } batch; 1259cb4d03abSBrian Vazquez 126009756af4SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_LOAD command */ 126109756af4SAlexei Starovoitov __u32 prog_type; /* one of enum bpf_prog_type */ 126209756af4SAlexei Starovoitov __u32 insn_cnt; 126309756af4SAlexei Starovoitov __aligned_u64 insns; 126409756af4SAlexei Starovoitov __aligned_u64 license; 1265cbd35700SAlexei Starovoitov __u32 log_level; /* verbosity level of verifier */ 1266cbd35700SAlexei Starovoitov __u32 log_size; /* size of user buffer */ 1267cbd35700SAlexei Starovoitov __aligned_u64 log_buf; /* user supplied buffer */ 12686c4fc209SDaniel Borkmann __u32 kern_version; /* not used */ 1269e07b98d9SDavid S. Miller __u32 prog_flags; 1270067cae47SMartin KaFai Lau char prog_name[BPF_OBJ_NAME_LEN]; 12711f6f4cb7SJakub Kicinski __u32 prog_ifindex; /* ifindex of netdev to prep for */ 12725e43f899SAndrey Ignatov /* For some prog types expected attach type must be known at 12735e43f899SAndrey Ignatov * load time to verify attach type specific parts of prog 12745e43f899SAndrey Ignatov * (context accesses, allowed helpers, etc). 12755e43f899SAndrey Ignatov */ 12765e43f899SAndrey Ignatov __u32 expected_attach_type; 1277838e9690SYonghong Song __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1278838e9690SYonghong Song __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1279838e9690SYonghong Song __aligned_u64 func_info; /* func info */ 1280838e9690SYonghong Song __u32 func_info_cnt; /* number of bpf_func_info records */ 1281c454a46bSMartin KaFai Lau __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1282c454a46bSMartin KaFai Lau __aligned_u64 line_info; /* line info */ 1283c454a46bSMartin KaFai Lau __u32 line_info_cnt; /* number of bpf_line_info records */ 1284ccfe29ebSAlexei Starovoitov __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1285290248a5SAndrii Nakryiko union { 1286290248a5SAndrii Nakryiko /* valid prog_fd to attach to bpf prog */ 1287290248a5SAndrii Nakryiko __u32 attach_prog_fd; 1288290248a5SAndrii Nakryiko /* or valid module BTF object fd or 0 to attach to vmlinux */ 1289290248a5SAndrii Nakryiko __u32 attach_btf_obj_fd; 1290290248a5SAndrii Nakryiko }; 129109756af4SAlexei Starovoitov }; 1292b2197755SDaniel Borkmann 1293b2197755SDaniel Borkmann struct { /* anonymous struct used by BPF_OBJ_* commands */ 1294b2197755SDaniel Borkmann __aligned_u64 pathname; 1295b2197755SDaniel Borkmann __u32 bpf_fd; 12966e71b04aSChenbo Feng __u32 file_flags; 1297b2197755SDaniel Borkmann }; 1298f4324551SDaniel Mack 1299f4324551SDaniel Mack struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1300f4324551SDaniel Mack __u32 target_fd; /* container object to attach to */ 1301f4324551SDaniel Mack __u32 attach_bpf_fd; /* eBPF program to attach */ 1302f4324551SDaniel Mack __u32 attach_type; 13037f677633SAlexei Starovoitov __u32 attach_flags; 13047dd68b32SAndrey Ignatov __u32 replace_bpf_fd; /* previously attached eBPF 13057dd68b32SAndrey Ignatov * program to replace if 13067dd68b32SAndrey Ignatov * BPF_F_REPLACE is used 13077dd68b32SAndrey Ignatov */ 1308f4324551SDaniel Mack }; 13091cf1cae9SAlexei Starovoitov 13101cf1cae9SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 13111cf1cae9SAlexei Starovoitov __u32 prog_fd; 13121cf1cae9SAlexei Starovoitov __u32 retval; 1313b5a36b1eSLorenz Bauer __u32 data_size_in; /* input: len of data_in */ 1314b5a36b1eSLorenz Bauer __u32 data_size_out; /* input/output: len of data_out 1315b5a36b1eSLorenz Bauer * returns ENOSPC if data_out 1316b5a36b1eSLorenz Bauer * is too small. 1317b5a36b1eSLorenz Bauer */ 13181cf1cae9SAlexei Starovoitov __aligned_u64 data_in; 13191cf1cae9SAlexei Starovoitov __aligned_u64 data_out; 13201cf1cae9SAlexei Starovoitov __u32 repeat; 13211cf1cae9SAlexei Starovoitov __u32 duration; 1322b0b9395dSStanislav Fomichev __u32 ctx_size_in; /* input: len of ctx_in */ 1323b0b9395dSStanislav Fomichev __u32 ctx_size_out; /* input/output: len of ctx_out 1324b0b9395dSStanislav Fomichev * returns ENOSPC if ctx_out 1325b0b9395dSStanislav Fomichev * is too small. 1326b0b9395dSStanislav Fomichev */ 1327b0b9395dSStanislav Fomichev __aligned_u64 ctx_in; 1328b0b9395dSStanislav Fomichev __aligned_u64 ctx_out; 13291b4d60ecSSong Liu __u32 flags; 13301b4d60ecSSong Liu __u32 cpu; 13311cf1cae9SAlexei Starovoitov } test; 133234ad5580SMartin KaFai Lau 1333b16d9aa4SMartin KaFai Lau struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1334b16d9aa4SMartin KaFai Lau union { 133534ad5580SMartin KaFai Lau __u32 start_id; 1336b16d9aa4SMartin KaFai Lau __u32 prog_id; 1337bd5f5f4eSMartin KaFai Lau __u32 map_id; 133878958fcaSMartin KaFai Lau __u32 btf_id; 1339a3b80e10SAndrii Nakryiko __u32 link_id; 1340b16d9aa4SMartin KaFai Lau }; 134134ad5580SMartin KaFai Lau __u32 next_id; 13426e71b04aSChenbo Feng __u32 open_flags; 134334ad5580SMartin KaFai Lau }; 13441e270976SMartin KaFai Lau 13451e270976SMartin KaFai Lau struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 13461e270976SMartin KaFai Lau __u32 bpf_fd; 13471e270976SMartin KaFai Lau __u32 info_len; 13481e270976SMartin KaFai Lau __aligned_u64 info; 13491e270976SMartin KaFai Lau } info; 1350468e2f64SAlexei Starovoitov 1351468e2f64SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1352468e2f64SAlexei Starovoitov __u32 target_fd; /* container object to query */ 1353468e2f64SAlexei Starovoitov __u32 attach_type; 1354468e2f64SAlexei Starovoitov __u32 query_flags; 1355468e2f64SAlexei Starovoitov __u32 attach_flags; 1356468e2f64SAlexei Starovoitov __aligned_u64 prog_ids; 1357468e2f64SAlexei Starovoitov __u32 prog_cnt; 1358468e2f64SAlexei Starovoitov } query; 1359c4f6699dSAlexei Starovoitov 1360af6eea57SAndrii Nakryiko struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1361c4f6699dSAlexei Starovoitov __u64 name; 1362c4f6699dSAlexei Starovoitov __u32 prog_fd; 1363c4f6699dSAlexei Starovoitov } raw_tracepoint; 1364f56a653cSMartin KaFai Lau 1365f56a653cSMartin KaFai Lau struct { /* anonymous struct for BPF_BTF_LOAD */ 1366f56a653cSMartin KaFai Lau __aligned_u64 btf; 1367f56a653cSMartin KaFai Lau __aligned_u64 btf_log_buf; 1368f56a653cSMartin KaFai Lau __u32 btf_size; 1369f56a653cSMartin KaFai Lau __u32 btf_log_size; 1370f56a653cSMartin KaFai Lau __u32 btf_log_level; 1371f56a653cSMartin KaFai Lau }; 137241bdc4b4SYonghong Song 137341bdc4b4SYonghong Song struct { 137441bdc4b4SYonghong Song __u32 pid; /* input: pid */ 137541bdc4b4SYonghong Song __u32 fd; /* input: fd */ 137641bdc4b4SYonghong Song __u32 flags; /* input: flags */ 137741bdc4b4SYonghong Song __u32 buf_len; /* input/output: buf len */ 137841bdc4b4SYonghong Song __aligned_u64 buf; /* input/output: 137941bdc4b4SYonghong Song * tp_name for tracepoint 138041bdc4b4SYonghong Song * symbol for kprobe 138141bdc4b4SYonghong Song * filename for uprobe 138241bdc4b4SYonghong Song */ 138341bdc4b4SYonghong Song __u32 prog_id; /* output: prod_id */ 138441bdc4b4SYonghong Song __u32 fd_type; /* output: BPF_FD_TYPE_* */ 138541bdc4b4SYonghong Song __u64 probe_offset; /* output: probe_offset */ 138641bdc4b4SYonghong Song __u64 probe_addr; /* output: probe_addr */ 138741bdc4b4SYonghong Song } task_fd_query; 1388af6eea57SAndrii Nakryiko 1389af6eea57SAndrii Nakryiko struct { /* struct used by BPF_LINK_CREATE command */ 1390af6eea57SAndrii Nakryiko __u32 prog_fd; /* eBPF program to attach */ 1391aa8d3a71SAndrii Nakryiko union { 1392af6eea57SAndrii Nakryiko __u32 target_fd; /* object to attach to */ 1393aa8d3a71SAndrii Nakryiko __u32 target_ifindex; /* target ifindex */ 1394aa8d3a71SAndrii Nakryiko }; 1395af6eea57SAndrii Nakryiko __u32 attach_type; /* attach type */ 1396af6eea57SAndrii Nakryiko __u32 flags; /* extra flags */ 13974a1e7c0cSToke Høiland-Jørgensen union { 13984a1e7c0cSToke Høiland-Jørgensen __u32 target_btf_id; /* btf_id of target to attach to */ 13994a1e7c0cSToke Høiland-Jørgensen struct { 14005e7b3020SYonghong Song __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 14015e7b3020SYonghong Song __u32 iter_info_len; /* iter_info length */ 14024a1e7c0cSToke Høiland-Jørgensen }; 14034a1e7c0cSToke Høiland-Jørgensen }; 1404af6eea57SAndrii Nakryiko } link_create; 14050c991ebcSAndrii Nakryiko 14060c991ebcSAndrii Nakryiko struct { /* struct used by BPF_LINK_UPDATE command */ 14070c991ebcSAndrii Nakryiko __u32 link_fd; /* link fd */ 14080c991ebcSAndrii Nakryiko /* new program fd to update link with */ 14090c991ebcSAndrii Nakryiko __u32 new_prog_fd; 14100c991ebcSAndrii Nakryiko __u32 flags; /* extra flags */ 14110c991ebcSAndrii Nakryiko /* expected link's program fd; is specified only if 14120c991ebcSAndrii Nakryiko * BPF_F_REPLACE flag is set in flags */ 14130c991ebcSAndrii Nakryiko __u32 old_prog_fd; 14140c991ebcSAndrii Nakryiko } link_update; 14150c991ebcSAndrii Nakryiko 141673b11c2aSAndrii Nakryiko struct { 141773b11c2aSAndrii Nakryiko __u32 link_fd; 141873b11c2aSAndrii Nakryiko } link_detach; 141973b11c2aSAndrii Nakryiko 1420d46edd67SSong Liu struct { /* struct used by BPF_ENABLE_STATS command */ 1421d46edd67SSong Liu __u32 type; 1422d46edd67SSong Liu } enable_stats; 1423d46edd67SSong Liu 1424ac51d99bSYonghong Song struct { /* struct used by BPF_ITER_CREATE command */ 1425ac51d99bSYonghong Song __u32 link_fd; 1426ac51d99bSYonghong Song __u32 flags; 1427ac51d99bSYonghong Song } iter_create; 1428ac51d99bSYonghong Song 1429ef15314aSYiFei Zhu struct { /* struct used by BPF_PROG_BIND_MAP command */ 1430ef15314aSYiFei Zhu __u32 prog_fd; 1431ef15314aSYiFei Zhu __u32 map_fd; 1432ef15314aSYiFei Zhu __u32 flags; /* extra flags */ 1433ef15314aSYiFei Zhu } prog_bind_map; 1434ef15314aSYiFei Zhu 143599c55f7dSAlexei Starovoitov } __attribute__((aligned(8))); 143699c55f7dSAlexei Starovoitov 143756a092c8SQuentin Monnet /* The description below is an attempt at providing documentation to eBPF 143856a092c8SQuentin Monnet * developers about the multiple available eBPF helper functions. It can be 143956a092c8SQuentin Monnet * parsed and used to produce a manual page. The workflow is the following, 144056a092c8SQuentin Monnet * and requires the rst2man utility: 1441ebb676daSThomas Graf * 144256a092c8SQuentin Monnet * $ ./scripts/bpf_helpers_doc.py \ 144356a092c8SQuentin Monnet * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 144456a092c8SQuentin Monnet * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 144556a092c8SQuentin Monnet * $ man /tmp/bpf-helpers.7 1446ebb676daSThomas Graf * 144756a092c8SQuentin Monnet * Note that in order to produce this external documentation, some RST 144856a092c8SQuentin Monnet * formatting is used in the descriptions to get "bold" and "italics" in 144956a092c8SQuentin Monnet * manual pages. Also note that the few trailing white spaces are 145056a092c8SQuentin Monnet * intentional, removing them would break paragraphs for rst2man. 1451ebb676daSThomas Graf * 145256a092c8SQuentin Monnet * Start of BPF helper function descriptions: 1453ad4a5223SQuentin Monnet * 1454ad4a5223SQuentin Monnet * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1455ad4a5223SQuentin Monnet * Description 1456ad4a5223SQuentin Monnet * Perform a lookup in *map* for an entry associated to *key*. 1457ad4a5223SQuentin Monnet * Return 1458ad4a5223SQuentin Monnet * Map value associated to *key*, or **NULL** if no entry was 1459ad4a5223SQuentin Monnet * found. 1460ad4a5223SQuentin Monnet * 1461bdb7b79bSAndrii Nakryiko * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1462ad4a5223SQuentin Monnet * Description 1463ad4a5223SQuentin Monnet * Add or update the value of the entry associated to *key* in 1464ad4a5223SQuentin Monnet * *map* with *value*. *flags* is one of: 1465ad4a5223SQuentin Monnet * 1466ad4a5223SQuentin Monnet * **BPF_NOEXIST** 1467ad4a5223SQuentin Monnet * The entry for *key* must not exist in the map. 1468ad4a5223SQuentin Monnet * **BPF_EXIST** 1469ad4a5223SQuentin Monnet * The entry for *key* must already exist in the map. 1470ad4a5223SQuentin Monnet * **BPF_ANY** 1471ad4a5223SQuentin Monnet * No condition on the existence of the entry for *key*. 1472ad4a5223SQuentin Monnet * 1473ad4a5223SQuentin Monnet * Flag value **BPF_NOEXIST** cannot be used for maps of types 1474ad4a5223SQuentin Monnet * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1475ad4a5223SQuentin Monnet * elements always exist), the helper would return an error. 1476ad4a5223SQuentin Monnet * Return 1477ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1478ad4a5223SQuentin Monnet * 1479bdb7b79bSAndrii Nakryiko * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1480ad4a5223SQuentin Monnet * Description 1481ad4a5223SQuentin Monnet * Delete entry with *key* from *map*. 1482ad4a5223SQuentin Monnet * Return 1483ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1484ad4a5223SQuentin Monnet * 1485bdb7b79bSAndrii Nakryiko * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1486ad4a5223SQuentin Monnet * Description 1487ad4a5223SQuentin Monnet * For tracing programs, safely attempt to read *size* bytes from 14886ae08ae3SDaniel Borkmann * kernel space address *unsafe_ptr* and store the data in *dst*. 14896ae08ae3SDaniel Borkmann * 1490ab8d7809SQuentin Monnet * Generally, use **bpf_probe_read_user**\ () or 1491ab8d7809SQuentin Monnet * **bpf_probe_read_kernel**\ () instead. 1492ad4a5223SQuentin Monnet * Return 1493ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1494ad4a5223SQuentin Monnet * 1495ad4a5223SQuentin Monnet * u64 bpf_ktime_get_ns(void) 1496ad4a5223SQuentin Monnet * Description 1497ad4a5223SQuentin Monnet * Return the time elapsed since system boot, in nanoseconds. 149871d19214SMaciej Żenczykowski * Does not include time the system was suspended. 1499ab8d7809SQuentin Monnet * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1500ad4a5223SQuentin Monnet * Return 1501ad4a5223SQuentin Monnet * Current *ktime*. 1502ad4a5223SQuentin Monnet * 1503bdb7b79bSAndrii Nakryiko * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1504ad4a5223SQuentin Monnet * Description 1505ad4a5223SQuentin Monnet * This helper is a "printk()-like" facility for debugging. It 1506ad4a5223SQuentin Monnet * prints a message defined by format *fmt* (of size *fmt_size*) 1507ad4a5223SQuentin Monnet * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 1508ad4a5223SQuentin Monnet * available. It can take up to three additional **u64** 1509ad4a5223SQuentin Monnet * arguments (as an eBPF helpers, the total number of arguments is 1510ad4a5223SQuentin Monnet * limited to five). 1511ad4a5223SQuentin Monnet * 1512ad4a5223SQuentin Monnet * Each time the helper is called, it appends a line to the trace. 151355c33dfbSPeter Wu * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 151455c33dfbSPeter Wu * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 1515ad4a5223SQuentin Monnet * The format of the trace is customizable, and the exact output 1516ad4a5223SQuentin Monnet * one will get depends on the options set in 1517ad4a5223SQuentin Monnet * *\/sys/kernel/debug/tracing/trace_options* (see also the 1518ad4a5223SQuentin Monnet * *README* file under the same directory). However, it usually 1519ad4a5223SQuentin Monnet * defaults to something like: 1520ad4a5223SQuentin Monnet * 1521ad4a5223SQuentin Monnet * :: 1522ad4a5223SQuentin Monnet * 1523ad4a5223SQuentin Monnet * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1524ad4a5223SQuentin Monnet * 1525ad4a5223SQuentin Monnet * In the above: 1526ad4a5223SQuentin Monnet * 1527ad4a5223SQuentin Monnet * * ``telnet`` is the name of the current task. 1528ad4a5223SQuentin Monnet * * ``470`` is the PID of the current task. 1529ad4a5223SQuentin Monnet * * ``001`` is the CPU number on which the task is 1530ad4a5223SQuentin Monnet * running. 1531ad4a5223SQuentin Monnet * * In ``.N..``, each character refers to a set of 1532ad4a5223SQuentin Monnet * options (whether irqs are enabled, scheduling 1533ad4a5223SQuentin Monnet * options, whether hard/softirqs are running, level of 1534ad4a5223SQuentin Monnet * preempt_disabled respectively). **N** means that 1535ad4a5223SQuentin Monnet * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1536ad4a5223SQuentin Monnet * are set. 1537ad4a5223SQuentin Monnet * * ``419421.045894`` is a timestamp. 1538ad4a5223SQuentin Monnet * * ``0x00000001`` is a fake value used by BPF for the 1539ad4a5223SQuentin Monnet * instruction pointer register. 1540ad4a5223SQuentin Monnet * * ``<formatted msg>`` is the message formatted with 1541ad4a5223SQuentin Monnet * *fmt*. 1542ad4a5223SQuentin Monnet * 1543ad4a5223SQuentin Monnet * The conversion specifiers supported by *fmt* are similar, but 1544ad4a5223SQuentin Monnet * more limited than for printk(). They are **%d**, **%i**, 1545ad4a5223SQuentin Monnet * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1546ad4a5223SQuentin Monnet * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1547ad4a5223SQuentin Monnet * of field, padding with zeroes, etc.) is available, and the 1548ad4a5223SQuentin Monnet * helper will return **-EINVAL** (but print nothing) if it 1549ad4a5223SQuentin Monnet * encounters an unknown specifier. 1550ad4a5223SQuentin Monnet * 1551ad4a5223SQuentin Monnet * Also, note that **bpf_trace_printk**\ () is slow, and should 1552ad4a5223SQuentin Monnet * only be used for debugging purposes. For this reason, a notice 1553b16fc097STobias Klauser * block (spanning several lines) is printed to kernel logs and 1554ad4a5223SQuentin Monnet * states that the helper should not be used "for production use" 1555ad4a5223SQuentin Monnet * the first time this helper is used (or more precisely, when 1556ad4a5223SQuentin Monnet * **trace_printk**\ () buffers are allocated). For passing values 1557ad4a5223SQuentin Monnet * to user space, perf events should be preferred. 1558ad4a5223SQuentin Monnet * Return 1559ad4a5223SQuentin Monnet * The number of bytes written to the buffer, or a negative error 1560ad4a5223SQuentin Monnet * in case of failure. 1561ad4a5223SQuentin Monnet * 15621fdd08beSQuentin Monnet * u32 bpf_get_prandom_u32(void) 15631fdd08beSQuentin Monnet * Description 15641fdd08beSQuentin Monnet * Get a pseudo-random number. 15651fdd08beSQuentin Monnet * 15661fdd08beSQuentin Monnet * From a security point of view, this helper uses its own 15671fdd08beSQuentin Monnet * pseudo-random internal state, and cannot be used to infer the 15681fdd08beSQuentin Monnet * seed of other random functions in the kernel. However, it is 15691fdd08beSQuentin Monnet * essential to note that the generator used by the helper is not 15701fdd08beSQuentin Monnet * cryptographically secure. 15711fdd08beSQuentin Monnet * Return 15721fdd08beSQuentin Monnet * A random 32-bit unsigned value. 15731fdd08beSQuentin Monnet * 15741fdd08beSQuentin Monnet * u32 bpf_get_smp_processor_id(void) 15751fdd08beSQuentin Monnet * Description 15761fdd08beSQuentin Monnet * Get the SMP (symmetric multiprocessing) processor id. Note that 15771fdd08beSQuentin Monnet * all programs run with preemption disabled, which means that the 15781fdd08beSQuentin Monnet * SMP processor id is stable during all the execution of the 15791fdd08beSQuentin Monnet * program. 15801fdd08beSQuentin Monnet * Return 15811fdd08beSQuentin Monnet * The SMP id of the processor running the program. 15821fdd08beSQuentin Monnet * 1583bdb7b79bSAndrii Nakryiko * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1584ad4a5223SQuentin Monnet * Description 1585ad4a5223SQuentin Monnet * Store *len* bytes from address *from* into the packet 1586ad4a5223SQuentin Monnet * associated to *skb*, at *offset*. *flags* are a combination of 1587ad4a5223SQuentin Monnet * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1588ad4a5223SQuentin Monnet * checksum for the packet after storing the bytes) and 1589ad4a5223SQuentin Monnet * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1590ad4a5223SQuentin Monnet * **->swhash** and *skb*\ **->l4hash** to 0). 1591ad4a5223SQuentin Monnet * 159232e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1593ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1594ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1595ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1596ad4a5223SQuentin Monnet * direct packet access. 1597ad4a5223SQuentin Monnet * Return 1598ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1599ad4a5223SQuentin Monnet * 1600bdb7b79bSAndrii Nakryiko * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1601ad4a5223SQuentin Monnet * Description 1602ad4a5223SQuentin Monnet * Recompute the layer 3 (e.g. IP) checksum for the packet 1603ad4a5223SQuentin Monnet * associated to *skb*. Computation is incremental, so the helper 1604ad4a5223SQuentin Monnet * must know the former value of the header field that was 1605ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 1606ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored in *size*. 1607ad4a5223SQuentin Monnet * Alternatively, it is possible to store the difference between 1608ad4a5223SQuentin Monnet * the previous and the new values of the header field in *to*, by 1609ad4a5223SQuentin Monnet * setting *from* and *size* to 0. For both methods, *offset* 1610ad4a5223SQuentin Monnet * indicates the location of the IP checksum within the packet. 1611ad4a5223SQuentin Monnet * 1612ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 1613ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 1614ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 1615ad4a5223SQuentin Monnet * checksum to update. 1616ad4a5223SQuentin Monnet * 161732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1618ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1619ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1620ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1621ad4a5223SQuentin Monnet * direct packet access. 1622ad4a5223SQuentin Monnet * Return 1623ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1624ad4a5223SQuentin Monnet * 1625bdb7b79bSAndrii Nakryiko * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1626ad4a5223SQuentin Monnet * Description 1627ad4a5223SQuentin Monnet * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1628ad4a5223SQuentin Monnet * packet associated to *skb*. Computation is incremental, so the 1629ad4a5223SQuentin Monnet * helper must know the former value of the header field that was 1630ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 1631ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored on the lowest 1632ad4a5223SQuentin Monnet * four bits of *flags*. Alternatively, it is possible to store 1633ad4a5223SQuentin Monnet * the difference between the previous and the new values of the 1634ad4a5223SQuentin Monnet * header field in *to*, by setting *from* and the four lowest 1635ad4a5223SQuentin Monnet * bits of *flags* to 0. For both methods, *offset* indicates the 1636ad4a5223SQuentin Monnet * location of the IP checksum within the packet. In addition to 1637ad4a5223SQuentin Monnet * the size of the field, *flags* can be added (bitwise OR) actual 1638ad4a5223SQuentin Monnet * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1639ad4a5223SQuentin Monnet * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1640ad4a5223SQuentin Monnet * for updates resulting in a null checksum the value is set to 1641ad4a5223SQuentin Monnet * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1642ad4a5223SQuentin Monnet * the checksum is to be computed against a pseudo-header. 1643ad4a5223SQuentin Monnet * 1644ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 1645ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 1646ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 1647ad4a5223SQuentin Monnet * checksum to update. 1648ad4a5223SQuentin Monnet * 164932e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1650ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1651ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1652ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1653ad4a5223SQuentin Monnet * direct packet access. 1654ad4a5223SQuentin Monnet * Return 1655ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1656ad4a5223SQuentin Monnet * 1657bdb7b79bSAndrii Nakryiko * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 1658ad4a5223SQuentin Monnet * Description 1659ad4a5223SQuentin Monnet * This special helper is used to trigger a "tail call", or in 1660ad4a5223SQuentin Monnet * other words, to jump into another eBPF program. The same stack 1661ad4a5223SQuentin Monnet * frame is used (but values on stack and in registers for the 1662ad4a5223SQuentin Monnet * caller are not accessible to the callee). This mechanism allows 1663ad4a5223SQuentin Monnet * for program chaining, either for raising the maximum number of 1664ad4a5223SQuentin Monnet * available eBPF instructions, or to execute given programs in 1665ad4a5223SQuentin Monnet * conditional blocks. For security reasons, there is an upper 1666ad4a5223SQuentin Monnet * limit to the number of successive tail calls that can be 1667ad4a5223SQuentin Monnet * performed. 1668ad4a5223SQuentin Monnet * 1669ad4a5223SQuentin Monnet * Upon call of this helper, the program attempts to jump into a 1670ad4a5223SQuentin Monnet * program referenced at index *index* in *prog_array_map*, a 1671ad4a5223SQuentin Monnet * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 1672ad4a5223SQuentin Monnet * *ctx*, a pointer to the context. 1673ad4a5223SQuentin Monnet * 1674ad4a5223SQuentin Monnet * If the call succeeds, the kernel immediately runs the first 1675ad4a5223SQuentin Monnet * instruction of the new program. This is not a function call, 1676ad4a5223SQuentin Monnet * and it never returns to the previous program. If the call 1677ad4a5223SQuentin Monnet * fails, then the helper has no effect, and the caller continues 1678ad4a5223SQuentin Monnet * to run its subsequent instructions. A call can fail if the 1679ad4a5223SQuentin Monnet * destination program for the jump does not exist (i.e. *index* 1680ad4a5223SQuentin Monnet * is superior to the number of entries in *prog_array_map*), or 1681ad4a5223SQuentin Monnet * if the maximum number of tail calls has been reached for this 1682ad4a5223SQuentin Monnet * chain of programs. This limit is defined in the kernel by the 1683ad4a5223SQuentin Monnet * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 1684ad4a5223SQuentin Monnet * which is currently set to 32. 1685ad4a5223SQuentin Monnet * Return 1686ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1687ad4a5223SQuentin Monnet * 1688bdb7b79bSAndrii Nakryiko * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 1689ad4a5223SQuentin Monnet * Description 1690ad4a5223SQuentin Monnet * Clone and redirect the packet associated to *skb* to another 1691ad4a5223SQuentin Monnet * net device of index *ifindex*. Both ingress and egress 1692ad4a5223SQuentin Monnet * interfaces can be used for redirection. The **BPF_F_INGRESS** 1693ad4a5223SQuentin Monnet * value in *flags* is used to make the distinction (ingress path 1694ad4a5223SQuentin Monnet * is selected if the flag is present, egress path otherwise). 1695ad4a5223SQuentin Monnet * This is the only flag supported for now. 1696ad4a5223SQuentin Monnet * 1697ad4a5223SQuentin Monnet * In comparison with **bpf_redirect**\ () helper, 1698ad4a5223SQuentin Monnet * **bpf_clone_redirect**\ () has the associated cost of 1699ad4a5223SQuentin Monnet * duplicating the packet buffer, but this can be executed out of 1700ad4a5223SQuentin Monnet * the eBPF program. Conversely, **bpf_redirect**\ () is more 1701ad4a5223SQuentin Monnet * efficient, but it is handled through an action code where the 1702ad4a5223SQuentin Monnet * redirection happens only after the eBPF program has returned. 1703ad4a5223SQuentin Monnet * 170432e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1705ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1706ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1707ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1708ad4a5223SQuentin Monnet * direct packet access. 1709ad4a5223SQuentin Monnet * Return 1710ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1711c456dec4SQuentin Monnet * 1712c456dec4SQuentin Monnet * u64 bpf_get_current_pid_tgid(void) 1713c456dec4SQuentin Monnet * Return 1714c456dec4SQuentin Monnet * A 64-bit integer containing the current tgid and pid, and 1715c456dec4SQuentin Monnet * created as such: 1716c456dec4SQuentin Monnet * *current_task*\ **->tgid << 32 \|** 1717c456dec4SQuentin Monnet * *current_task*\ **->pid**. 1718c456dec4SQuentin Monnet * 1719c456dec4SQuentin Monnet * u64 bpf_get_current_uid_gid(void) 1720c456dec4SQuentin Monnet * Return 1721c456dec4SQuentin Monnet * A 64-bit integer containing the current GID and UID, and 1722c456dec4SQuentin Monnet * created as such: *current_gid* **<< 32 \|** *current_uid*. 1723c456dec4SQuentin Monnet * 1724bdb7b79bSAndrii Nakryiko * long bpf_get_current_comm(void *buf, u32 size_of_buf) 1725c456dec4SQuentin Monnet * Description 1726c456dec4SQuentin Monnet * Copy the **comm** attribute of the current task into *buf* of 1727c456dec4SQuentin Monnet * *size_of_buf*. The **comm** attribute contains the name of 1728c456dec4SQuentin Monnet * the executable (excluding the path) for the current task. The 1729c456dec4SQuentin Monnet * *size_of_buf* must be strictly positive. On success, the 1730c456dec4SQuentin Monnet * helper makes sure that the *buf* is NUL-terminated. On failure, 1731c456dec4SQuentin Monnet * it is filled with zeroes. 1732c456dec4SQuentin Monnet * Return 1733c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1734c456dec4SQuentin Monnet * 17351fdd08beSQuentin Monnet * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 17361fdd08beSQuentin Monnet * Description 17371fdd08beSQuentin Monnet * Retrieve the classid for the current task, i.e. for the net_cls 17381fdd08beSQuentin Monnet * cgroup to which *skb* belongs. 17391fdd08beSQuentin Monnet * 17401fdd08beSQuentin Monnet * This helper can be used on TC egress path, but not on ingress. 17411fdd08beSQuentin Monnet * 17421fdd08beSQuentin Monnet * The net_cls cgroup provides an interface to tag network packets 17431fdd08beSQuentin Monnet * based on a user-provided identifier for all traffic coming from 17441fdd08beSQuentin Monnet * the tasks belonging to the related cgroup. See also the related 17451fdd08beSQuentin Monnet * kernel documentation, available from the Linux sources in file 1746da82c92fSMauro Carvalho Chehab * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 17471fdd08beSQuentin Monnet * 17481fdd08beSQuentin Monnet * The Linux kernel has two versions for cgroups: there are 17491fdd08beSQuentin Monnet * cgroups v1 and cgroups v2. Both are available to users, who can 17501fdd08beSQuentin Monnet * use a mixture of them, but note that the net_cls cgroup is for 17511fdd08beSQuentin Monnet * cgroup v1 only. This makes it incompatible with BPF programs 17521fdd08beSQuentin Monnet * run on cgroups, which is a cgroup-v2-only feature (a socket can 17531fdd08beSQuentin Monnet * only hold data for one version of cgroups at a time). 17541fdd08beSQuentin Monnet * 17551fdd08beSQuentin Monnet * This helper is only available is the kernel was compiled with 17561fdd08beSQuentin Monnet * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 17571fdd08beSQuentin Monnet * "**y**" or to "**m**". 17581fdd08beSQuentin Monnet * Return 17591fdd08beSQuentin Monnet * The classid, or 0 for the default unconfigured classid. 17601fdd08beSQuentin Monnet * 1761bdb7b79bSAndrii Nakryiko * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1762c456dec4SQuentin Monnet * Description 1763c456dec4SQuentin Monnet * Push a *vlan_tci* (VLAN tag control information) of protocol 1764c456dec4SQuentin Monnet * *vlan_proto* to the packet associated to *skb*, then update 1765c456dec4SQuentin Monnet * the checksum. Note that if *vlan_proto* is different from 1766c456dec4SQuentin Monnet * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1767c456dec4SQuentin Monnet * be **ETH_P_8021Q**. 1768c456dec4SQuentin Monnet * 176932e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1770c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1771c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 1772c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 1773c456dec4SQuentin Monnet * direct packet access. 1774c456dec4SQuentin Monnet * Return 1775c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1776c456dec4SQuentin Monnet * 1777bdb7b79bSAndrii Nakryiko * long bpf_skb_vlan_pop(struct sk_buff *skb) 1778c456dec4SQuentin Monnet * Description 1779c456dec4SQuentin Monnet * Pop a VLAN header from the packet associated to *skb*. 1780c456dec4SQuentin Monnet * 178132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1782c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1783c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 1784c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 1785c456dec4SQuentin Monnet * direct packet access. 1786c456dec4SQuentin Monnet * Return 1787c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1788c456dec4SQuentin Monnet * 1789bdb7b79bSAndrii Nakryiko * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1790c456dec4SQuentin Monnet * Description 1791c456dec4SQuentin Monnet * Get tunnel metadata. This helper takes a pointer *key* to an 1792c456dec4SQuentin Monnet * empty **struct bpf_tunnel_key** of **size**, that will be 1793c456dec4SQuentin Monnet * filled with tunnel metadata for the packet associated to *skb*. 1794c456dec4SQuentin Monnet * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1795c456dec4SQuentin Monnet * indicates that the tunnel is based on IPv6 protocol instead of 1796c456dec4SQuentin Monnet * IPv4. 1797c456dec4SQuentin Monnet * 1798c456dec4SQuentin Monnet * The **struct bpf_tunnel_key** is an object that generalizes the 1799c456dec4SQuentin Monnet * principal parameters used by various tunneling protocols into a 1800c456dec4SQuentin Monnet * single struct. This way, it can be used to easily make a 1801c456dec4SQuentin Monnet * decision based on the contents of the encapsulation header, 1802c456dec4SQuentin Monnet * "summarized" in this struct. In particular, it holds the IP 1803c456dec4SQuentin Monnet * address of the remote end (IPv4 or IPv6, depending on the case) 1804c456dec4SQuentin Monnet * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1805c456dec4SQuentin Monnet * this struct exposes the *key*\ **->tunnel_id**, which is 1806c456dec4SQuentin Monnet * generally mapped to a VNI (Virtual Network Identifier), making 1807c456dec4SQuentin Monnet * it programmable together with the **bpf_skb_set_tunnel_key**\ 1808c456dec4SQuentin Monnet * () helper. 1809c456dec4SQuentin Monnet * 1810c456dec4SQuentin Monnet * Let's imagine that the following code is part of a program 1811c456dec4SQuentin Monnet * attached to the TC ingress interface, on one end of a GRE 1812c456dec4SQuentin Monnet * tunnel, and is supposed to filter out all messages coming from 1813c456dec4SQuentin Monnet * remote ends with IPv4 address other than 10.0.0.1: 1814c456dec4SQuentin Monnet * 1815c456dec4SQuentin Monnet * :: 1816c456dec4SQuentin Monnet * 1817c456dec4SQuentin Monnet * int ret; 1818c456dec4SQuentin Monnet * struct bpf_tunnel_key key = {}; 1819c456dec4SQuentin Monnet * 1820c456dec4SQuentin Monnet * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1821c456dec4SQuentin Monnet * if (ret < 0) 1822c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 1823c456dec4SQuentin Monnet * 1824c456dec4SQuentin Monnet * if (key.remote_ipv4 != 0x0a000001) 1825c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 1826c456dec4SQuentin Monnet * 1827c456dec4SQuentin Monnet * return TC_ACT_OK; // accept packet 1828c456dec4SQuentin Monnet * 1829c456dec4SQuentin Monnet * This interface can also be used with all encapsulation devices 1830c456dec4SQuentin Monnet * that can operate in "collect metadata" mode: instead of having 1831c456dec4SQuentin Monnet * one network device per specific configuration, the "collect 1832c456dec4SQuentin Monnet * metadata" mode only requires a single device where the 1833c456dec4SQuentin Monnet * configuration can be extracted from this helper. 1834c456dec4SQuentin Monnet * 1835c456dec4SQuentin Monnet * This can be used together with various tunnels such as VXLan, 1836c456dec4SQuentin Monnet * Geneve, GRE or IP in IP (IPIP). 1837c456dec4SQuentin Monnet * Return 1838c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1839c456dec4SQuentin Monnet * 1840bdb7b79bSAndrii Nakryiko * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1841c456dec4SQuentin Monnet * Description 1842c456dec4SQuentin Monnet * Populate tunnel metadata for packet associated to *skb.* The 1843c456dec4SQuentin Monnet * tunnel metadata is set to the contents of *key*, of *size*. The 1844c456dec4SQuentin Monnet * *flags* can be set to a combination of the following values: 1845c456dec4SQuentin Monnet * 1846c456dec4SQuentin Monnet * **BPF_F_TUNINFO_IPV6** 1847c456dec4SQuentin Monnet * Indicate that the tunnel is based on IPv6 protocol 1848c456dec4SQuentin Monnet * instead of IPv4. 1849c456dec4SQuentin Monnet * **BPF_F_ZERO_CSUM_TX** 1850c456dec4SQuentin Monnet * For IPv4 packets, add a flag to tunnel metadata 1851c456dec4SQuentin Monnet * indicating that checksum computation should be skipped 1852c456dec4SQuentin Monnet * and checksum set to zeroes. 1853c456dec4SQuentin Monnet * **BPF_F_DONT_FRAGMENT** 1854c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that the 1855c456dec4SQuentin Monnet * packet should not be fragmented. 1856c456dec4SQuentin Monnet * **BPF_F_SEQ_NUMBER** 1857c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that a 1858c456dec4SQuentin Monnet * sequence number should be added to tunnel header before 1859c456dec4SQuentin Monnet * sending the packet. This flag was added for GRE 1860c456dec4SQuentin Monnet * encapsulation, but might be used with other protocols 1861c456dec4SQuentin Monnet * as well in the future. 1862c456dec4SQuentin Monnet * 1863c456dec4SQuentin Monnet * Here is a typical usage on the transmit path: 1864c456dec4SQuentin Monnet * 1865c456dec4SQuentin Monnet * :: 1866c456dec4SQuentin Monnet * 1867c456dec4SQuentin Monnet * struct bpf_tunnel_key key; 1868c456dec4SQuentin Monnet * populate key ... 1869c456dec4SQuentin Monnet * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1870c456dec4SQuentin Monnet * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1871c456dec4SQuentin Monnet * 1872c456dec4SQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_key**\ () 1873c456dec4SQuentin Monnet * helper for additional information. 1874c456dec4SQuentin Monnet * Return 1875c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1876c456dec4SQuentin Monnet * 1877c6b5fb86SQuentin Monnet * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1878c6b5fb86SQuentin Monnet * Description 1879c6b5fb86SQuentin Monnet * Read the value of a perf event counter. This helper relies on a 1880c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1881c6b5fb86SQuentin Monnet * the perf event counter is selected when *map* is updated with 1882c6b5fb86SQuentin Monnet * perf event file descriptors. The *map* is an array whose size 1883c6b5fb86SQuentin Monnet * is the number of available CPUs, and each cell contains a value 1884c6b5fb86SQuentin Monnet * relative to one CPU. The value to retrieve is indicated by 1885c6b5fb86SQuentin Monnet * *flags*, that contains the index of the CPU to look up, masked 1886c6b5fb86SQuentin Monnet * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1887c6b5fb86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 1888c6b5fb86SQuentin Monnet * current CPU should be retrieved. 1889c6b5fb86SQuentin Monnet * 1890c6b5fb86SQuentin Monnet * Note that before Linux 4.13, only hardware perf event can be 1891c6b5fb86SQuentin Monnet * retrieved. 1892c6b5fb86SQuentin Monnet * 1893c6b5fb86SQuentin Monnet * Also, be aware that the newer helper 1894c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () is recommended over 18953bd5a09bSQuentin Monnet * **bpf_perf_event_read**\ () in general. The latter has some ABI 1896c6b5fb86SQuentin Monnet * quirks where error and counter value are used as a return code 1897c6b5fb86SQuentin Monnet * (which is wrong to do since ranges may overlap). This issue is 18983bd5a09bSQuentin Monnet * fixed with **bpf_perf_event_read_value**\ (), which at the same 18993bd5a09bSQuentin Monnet * time provides more features over the **bpf_perf_event_read**\ 19003bd5a09bSQuentin Monnet * () interface. Please refer to the description of 1901c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () for details. 1902c6b5fb86SQuentin Monnet * Return 1903c6b5fb86SQuentin Monnet * The value of the perf event counter read from the map, or a 1904c6b5fb86SQuentin Monnet * negative error code in case of failure. 1905c6b5fb86SQuentin Monnet * 1906bdb7b79bSAndrii Nakryiko * long bpf_redirect(u32 ifindex, u64 flags) 1907c456dec4SQuentin Monnet * Description 1908c456dec4SQuentin Monnet * Redirect the packet to another net device of index *ifindex*. 1909c456dec4SQuentin Monnet * This helper is somewhat similar to **bpf_clone_redirect**\ 1910c456dec4SQuentin Monnet * (), except that the packet is not cloned, which provides 1911c456dec4SQuentin Monnet * increased performance. 1912c456dec4SQuentin Monnet * 1913c456dec4SQuentin Monnet * Except for XDP, both ingress and egress interfaces can be used 1914c456dec4SQuentin Monnet * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1915c456dec4SQuentin Monnet * to make the distinction (ingress path is selected if the flag 1916c456dec4SQuentin Monnet * is present, egress path otherwise). Currently, XDP only 1917c456dec4SQuentin Monnet * supports redirection to the egress interface, and accepts no 1918c456dec4SQuentin Monnet * flag at all. 1919c456dec4SQuentin Monnet * 1920f25975f4SToke Høiland-Jørgensen * The same effect can also be attained with the more generic 1921f25975f4SToke Høiland-Jørgensen * **bpf_redirect_map**\ (), which uses a BPF map to store the 1922f25975f4SToke Høiland-Jørgensen * redirect target instead of providing it directly to the helper. 1923c456dec4SQuentin Monnet * Return 1924c456dec4SQuentin Monnet * For XDP, the helper returns **XDP_REDIRECT** on success or 1925c456dec4SQuentin Monnet * **XDP_ABORTED** on error. For other program types, the values 1926c456dec4SQuentin Monnet * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1927c456dec4SQuentin Monnet * error. 1928c456dec4SQuentin Monnet * 19291fdd08beSQuentin Monnet * u32 bpf_get_route_realm(struct sk_buff *skb) 19301fdd08beSQuentin Monnet * Description 19311fdd08beSQuentin Monnet * Retrieve the realm or the route, that is to say the 19321fdd08beSQuentin Monnet * **tclassid** field of the destination for the *skb*. The 1933b16fc097STobias Klauser * identifier retrieved is a user-provided tag, similar to the 19341fdd08beSQuentin Monnet * one used with the net_cls cgroup (see description for 19351fdd08beSQuentin Monnet * **bpf_get_cgroup_classid**\ () helper), but here this tag is 19361fdd08beSQuentin Monnet * held by a route (a destination entry), not by a task. 19371fdd08beSQuentin Monnet * 19381fdd08beSQuentin Monnet * Retrieving this identifier works with the clsact TC egress hook 19391fdd08beSQuentin Monnet * (see also **tc-bpf(8)**), or alternatively on conventional 19401fdd08beSQuentin Monnet * classful egress qdiscs, but not on TC ingress path. In case of 19411fdd08beSQuentin Monnet * clsact TC egress hook, this has the advantage that, internally, 19421fdd08beSQuentin Monnet * the destination entry has not been dropped yet in the transmit 19431fdd08beSQuentin Monnet * path. Therefore, the destination entry does not need to be 19441fdd08beSQuentin Monnet * artificially held via **netif_keep_dst**\ () for a classful 19451fdd08beSQuentin Monnet * qdisc until the *skb* is freed. 19461fdd08beSQuentin Monnet * 19471fdd08beSQuentin Monnet * This helper is available only if the kernel was compiled with 19481fdd08beSQuentin Monnet * **CONFIG_IP_ROUTE_CLASSID** configuration option. 19491fdd08beSQuentin Monnet * Return 19501fdd08beSQuentin Monnet * The realm of the route for the packet associated to *skb*, or 0 19511fdd08beSQuentin Monnet * if none was found. 19521fdd08beSQuentin Monnet * 1953bdb7b79bSAndrii Nakryiko * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1954c456dec4SQuentin Monnet * Description 1955c456dec4SQuentin Monnet * Write raw *data* blob into a special BPF perf event held by 1956c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1957c456dec4SQuentin Monnet * event must have the following attributes: **PERF_SAMPLE_RAW** 1958c456dec4SQuentin Monnet * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1959c456dec4SQuentin Monnet * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1960c456dec4SQuentin Monnet * 1961c456dec4SQuentin Monnet * The *flags* are used to indicate the index in *map* for which 1962c456dec4SQuentin Monnet * the value must be put, masked with **BPF_F_INDEX_MASK**. 1963c456dec4SQuentin Monnet * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1964c456dec4SQuentin Monnet * to indicate that the index of the current CPU core should be 1965c456dec4SQuentin Monnet * used. 1966c456dec4SQuentin Monnet * 1967c456dec4SQuentin Monnet * The value to write, of *size*, is passed through eBPF stack and 1968c456dec4SQuentin Monnet * pointed by *data*. 1969c456dec4SQuentin Monnet * 1970c456dec4SQuentin Monnet * The context of the program *ctx* needs also be passed to the 1971c456dec4SQuentin Monnet * helper. 1972c456dec4SQuentin Monnet * 1973c456dec4SQuentin Monnet * On user space, a program willing to read the values needs to 1974c456dec4SQuentin Monnet * call **perf_event_open**\ () on the perf event (either for 1975c456dec4SQuentin Monnet * one or for all CPUs) and to store the file descriptor into the 1976c456dec4SQuentin Monnet * *map*. This must be done before the eBPF program can send data 1977c456dec4SQuentin Monnet * into it. An example is available in file 1978c456dec4SQuentin Monnet * *samples/bpf/trace_output_user.c* in the Linux kernel source 1979c456dec4SQuentin Monnet * tree (the eBPF program counterpart is in 1980c456dec4SQuentin Monnet * *samples/bpf/trace_output_kern.c*). 1981c456dec4SQuentin Monnet * 1982c456dec4SQuentin Monnet * **bpf_perf_event_output**\ () achieves better performance 1983c456dec4SQuentin Monnet * than **bpf_trace_printk**\ () for sharing data with user 1984c456dec4SQuentin Monnet * space, and is much better suitable for streaming data from eBPF 1985c456dec4SQuentin Monnet * programs. 1986c456dec4SQuentin Monnet * 1987c456dec4SQuentin Monnet * Note that this helper is not restricted to tracing use cases 1988c456dec4SQuentin Monnet * and can be used with programs attached to TC or XDP as well, 1989c456dec4SQuentin Monnet * where it allows for passing data to user space listeners. Data 1990c456dec4SQuentin Monnet * can be: 1991c456dec4SQuentin Monnet * 1992c456dec4SQuentin Monnet * * Only custom structs, 1993c456dec4SQuentin Monnet * * Only the packet payload, or 1994c456dec4SQuentin Monnet * * A combination of both. 1995c456dec4SQuentin Monnet * Return 1996c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1997c456dec4SQuentin Monnet * 1998bdb7b79bSAndrii Nakryiko * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 19991fdd08beSQuentin Monnet * Description 20001fdd08beSQuentin Monnet * This helper was provided as an easy way to load data from a 20011fdd08beSQuentin Monnet * packet. It can be used to load *len* bytes from *offset* from 20021fdd08beSQuentin Monnet * the packet associated to *skb*, into the buffer pointed by 20031fdd08beSQuentin Monnet * *to*. 20041fdd08beSQuentin Monnet * 20051fdd08beSQuentin Monnet * Since Linux 4.7, usage of this helper has mostly been replaced 20061fdd08beSQuentin Monnet * by "direct packet access", enabling packet data to be 20071fdd08beSQuentin Monnet * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 20081fdd08beSQuentin Monnet * pointing respectively to the first byte of packet data and to 20091fdd08beSQuentin Monnet * the byte after the last byte of packet data. However, it 20101fdd08beSQuentin Monnet * remains useful if one wishes to read large quantities of data 20111fdd08beSQuentin Monnet * at once from a packet into the eBPF stack. 20121fdd08beSQuentin Monnet * Return 20131fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 20141fdd08beSQuentin Monnet * 2015bdb7b79bSAndrii Nakryiko * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2016c456dec4SQuentin Monnet * Description 2017c456dec4SQuentin Monnet * Walk a user or a kernel stack and return its id. To achieve 2018c456dec4SQuentin Monnet * this, the helper needs *ctx*, which is a pointer to the context 2019c456dec4SQuentin Monnet * on which the tracing program is executed, and a pointer to a 2020c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2021c456dec4SQuentin Monnet * 2022c456dec4SQuentin Monnet * The last argument, *flags*, holds the number of stack frames to 2023c456dec4SQuentin Monnet * skip (from 0 to 255), masked with 2024c456dec4SQuentin Monnet * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2025c456dec4SQuentin Monnet * a combination of the following flags: 2026c456dec4SQuentin Monnet * 2027c456dec4SQuentin Monnet * **BPF_F_USER_STACK** 2028c456dec4SQuentin Monnet * Collect a user space stack instead of a kernel stack. 2029c456dec4SQuentin Monnet * **BPF_F_FAST_STACK_CMP** 2030c456dec4SQuentin Monnet * Compare stacks by hash only. 2031c456dec4SQuentin Monnet * **BPF_F_REUSE_STACKID** 2032c456dec4SQuentin Monnet * If two different stacks hash into the same *stackid*, 2033c456dec4SQuentin Monnet * discard the old one. 2034c456dec4SQuentin Monnet * 2035c456dec4SQuentin Monnet * The stack id retrieved is a 32 bit long integer handle which 2036c456dec4SQuentin Monnet * can be further combined with other data (including other stack 2037c456dec4SQuentin Monnet * ids) and used as a key into maps. This can be useful for 2038c456dec4SQuentin Monnet * generating a variety of graphs (such as flame graphs or off-cpu 2039c456dec4SQuentin Monnet * graphs). 2040c456dec4SQuentin Monnet * 2041c456dec4SQuentin Monnet * For walking a stack, this helper is an improvement over 2042c456dec4SQuentin Monnet * **bpf_probe_read**\ (), which can be used with unrolled loops 2043c456dec4SQuentin Monnet * but is not efficient and consumes a lot of eBPF instructions. 2044c456dec4SQuentin Monnet * Instead, **bpf_get_stackid**\ () can collect up to 2045c456dec4SQuentin Monnet * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2046c456dec4SQuentin Monnet * this limit can be controlled with the **sysctl** program, and 2047c456dec4SQuentin Monnet * that it should be manually increased in order to profile long 2048c456dec4SQuentin Monnet * user stacks (such as stacks for Java programs). To do so, use: 2049c456dec4SQuentin Monnet * 2050c456dec4SQuentin Monnet * :: 2051c456dec4SQuentin Monnet * 2052c456dec4SQuentin Monnet * # sysctl kernel.perf_event_max_stack=<new value> 2053c456dec4SQuentin Monnet * Return 2054c456dec4SQuentin Monnet * The positive or null stack id on success, or a negative error 2055c456dec4SQuentin Monnet * in case of failure. 2056c456dec4SQuentin Monnet * 20571fdd08beSQuentin Monnet * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 20581fdd08beSQuentin Monnet * Description 20591fdd08beSQuentin Monnet * Compute a checksum difference, from the raw buffer pointed by 20601fdd08beSQuentin Monnet * *from*, of length *from_size* (that must be a multiple of 4), 20611fdd08beSQuentin Monnet * towards the raw buffer pointed by *to*, of size *to_size* 20621fdd08beSQuentin Monnet * (same remark). An optional *seed* can be added to the value 20631fdd08beSQuentin Monnet * (this can be cascaded, the seed may come from a previous call 20641fdd08beSQuentin Monnet * to the helper). 20651fdd08beSQuentin Monnet * 20661fdd08beSQuentin Monnet * This is flexible enough to be used in several ways: 20671fdd08beSQuentin Monnet * 20681fdd08beSQuentin Monnet * * With *from_size* == 0, *to_size* > 0 and *seed* set to 20691fdd08beSQuentin Monnet * checksum, it can be used when pushing new data. 20701fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* == 0 and *seed* set to 20711fdd08beSQuentin Monnet * checksum, it can be used when removing data from a packet. 20721fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 20731fdd08beSQuentin Monnet * can be used to compute a diff. Note that *from_size* and 20741fdd08beSQuentin Monnet * *to_size* do not need to be equal. 20751fdd08beSQuentin Monnet * 20761fdd08beSQuentin Monnet * This helper can be used in combination with 20771fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 20781fdd08beSQuentin Monnet * which one can feed in the difference computed with 20791fdd08beSQuentin Monnet * **bpf_csum_diff**\ (). 20801fdd08beSQuentin Monnet * Return 20811fdd08beSQuentin Monnet * The checksum result, or a negative error code in case of 20821fdd08beSQuentin Monnet * failure. 20831fdd08beSQuentin Monnet * 2084bdb7b79bSAndrii Nakryiko * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 20851fdd08beSQuentin Monnet * Description 20861fdd08beSQuentin Monnet * Retrieve tunnel options metadata for the packet associated to 20871fdd08beSQuentin Monnet * *skb*, and store the raw tunnel option data to the buffer *opt* 20881fdd08beSQuentin Monnet * of *size*. 20891fdd08beSQuentin Monnet * 20901fdd08beSQuentin Monnet * This helper can be used with encapsulation devices that can 20911fdd08beSQuentin Monnet * operate in "collect metadata" mode (please refer to the related 20921fdd08beSQuentin Monnet * note in the description of **bpf_skb_get_tunnel_key**\ () for 20931fdd08beSQuentin Monnet * more details). A particular example where this can be used is 20941fdd08beSQuentin Monnet * in combination with the Geneve encapsulation protocol, where it 20951fdd08beSQuentin Monnet * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 20961fdd08beSQuentin Monnet * and retrieving arbitrary TLVs (Type-Length-Value headers) from 20971fdd08beSQuentin Monnet * the eBPF program. This allows for full customization of these 20981fdd08beSQuentin Monnet * headers. 20991fdd08beSQuentin Monnet * Return 21001fdd08beSQuentin Monnet * The size of the option data retrieved. 21011fdd08beSQuentin Monnet * 2102bdb7b79bSAndrii Nakryiko * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 21031fdd08beSQuentin Monnet * Description 21041fdd08beSQuentin Monnet * Set tunnel options metadata for the packet associated to *skb* 21051fdd08beSQuentin Monnet * to the option data contained in the raw buffer *opt* of *size*. 21061fdd08beSQuentin Monnet * 21071fdd08beSQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_opt**\ () 21081fdd08beSQuentin Monnet * helper for additional information. 21091fdd08beSQuentin Monnet * Return 21101fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 21111fdd08beSQuentin Monnet * 2112bdb7b79bSAndrii Nakryiko * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 21131fdd08beSQuentin Monnet * Description 21141fdd08beSQuentin Monnet * Change the protocol of the *skb* to *proto*. Currently 21151fdd08beSQuentin Monnet * supported are transition from IPv4 to IPv6, and from IPv6 to 21161fdd08beSQuentin Monnet * IPv4. The helper takes care of the groundwork for the 21171fdd08beSQuentin Monnet * transition, including resizing the socket buffer. The eBPF 21181fdd08beSQuentin Monnet * program is expected to fill the new headers, if any, via 21191fdd08beSQuentin Monnet * **skb_store_bytes**\ () and to recompute the checksums with 21201fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 21211fdd08beSQuentin Monnet * (). The main case for this helper is to perform NAT64 21221fdd08beSQuentin Monnet * operations out of an eBPF program. 21231fdd08beSQuentin Monnet * 21241fdd08beSQuentin Monnet * Internally, the GSO type is marked as dodgy so that headers are 21251fdd08beSQuentin Monnet * checked and segments are recalculated by the GSO/GRO engine. 21261fdd08beSQuentin Monnet * The size for GSO target is adapted as well. 21271fdd08beSQuentin Monnet * 21281fdd08beSQuentin Monnet * All values for *flags* are reserved for future usage, and must 21291fdd08beSQuentin Monnet * be left at zero. 21301fdd08beSQuentin Monnet * 213132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 21321fdd08beSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 21331fdd08beSQuentin Monnet * previously done by the verifier are invalidated and must be 21341fdd08beSQuentin Monnet * performed again, if the helper is used in combination with 21351fdd08beSQuentin Monnet * direct packet access. 21361fdd08beSQuentin Monnet * Return 21371fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 21381fdd08beSQuentin Monnet * 2139bdb7b79bSAndrii Nakryiko * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 21401fdd08beSQuentin Monnet * Description 21411fdd08beSQuentin Monnet * Change the packet type for the packet associated to *skb*. This 21421fdd08beSQuentin Monnet * comes down to setting *skb*\ **->pkt_type** to *type*, except 21431fdd08beSQuentin Monnet * the eBPF program does not have a write access to *skb*\ 21441fdd08beSQuentin Monnet * **->pkt_type** beside this helper. Using a helper here allows 21451fdd08beSQuentin Monnet * for graceful handling of errors. 21461fdd08beSQuentin Monnet * 21471fdd08beSQuentin Monnet * The major use case is to change incoming *skb*s to 21481fdd08beSQuentin Monnet * **PACKET_HOST** in a programmatic way instead of having to 21491fdd08beSQuentin Monnet * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 21501fdd08beSQuentin Monnet * example. 21511fdd08beSQuentin Monnet * 21521fdd08beSQuentin Monnet * Note that *type* only allows certain values. At this time, they 21531fdd08beSQuentin Monnet * are: 21541fdd08beSQuentin Monnet * 21551fdd08beSQuentin Monnet * **PACKET_HOST** 21561fdd08beSQuentin Monnet * Packet is for us. 21571fdd08beSQuentin Monnet * **PACKET_BROADCAST** 21581fdd08beSQuentin Monnet * Send packet to all. 21591fdd08beSQuentin Monnet * **PACKET_MULTICAST** 21601fdd08beSQuentin Monnet * Send packet to group. 21611fdd08beSQuentin Monnet * **PACKET_OTHERHOST** 21621fdd08beSQuentin Monnet * Send packet to someone else. 21631fdd08beSQuentin Monnet * Return 21641fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 21651fdd08beSQuentin Monnet * 2166bdb7b79bSAndrii Nakryiko * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2167c6b5fb86SQuentin Monnet * Description 2168c6b5fb86SQuentin Monnet * Check whether *skb* is a descendant of the cgroup2 held by 2169c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2170c6b5fb86SQuentin Monnet * Return 2171c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 2172c6b5fb86SQuentin Monnet * 2173c6b5fb86SQuentin Monnet * * 0, if the *skb* failed the cgroup2 descendant test. 2174c6b5fb86SQuentin Monnet * * 1, if the *skb* succeeded the cgroup2 descendant test. 2175c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 2176c6b5fb86SQuentin Monnet * 2177fa15601aSQuentin Monnet * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2178fa15601aSQuentin Monnet * Description 2179fa15601aSQuentin Monnet * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2180fa15601aSQuentin Monnet * not set, in particular if the hash was cleared due to mangling, 2181fa15601aSQuentin Monnet * recompute this hash. Later accesses to the hash can be done 2182fa15601aSQuentin Monnet * directly with *skb*\ **->hash**. 2183fa15601aSQuentin Monnet * 2184fa15601aSQuentin Monnet * Calling **bpf_set_hash_invalid**\ (), changing a packet 2185fa15601aSQuentin Monnet * prototype with **bpf_skb_change_proto**\ (), or calling 2186fa15601aSQuentin Monnet * **bpf_skb_store_bytes**\ () with the 2187fa15601aSQuentin Monnet * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2188fa15601aSQuentin Monnet * the hash and to trigger a new computation for the next call to 2189fa15601aSQuentin Monnet * **bpf_get_hash_recalc**\ (). 2190fa15601aSQuentin Monnet * Return 2191fa15601aSQuentin Monnet * The 32-bit hash. 2192fa15601aSQuentin Monnet * 2193c456dec4SQuentin Monnet * u64 bpf_get_current_task(void) 2194c456dec4SQuentin Monnet * Return 2195c456dec4SQuentin Monnet * A pointer to the current task struct. 2196fa15601aSQuentin Monnet * 2197bdb7b79bSAndrii Nakryiko * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2198c6b5fb86SQuentin Monnet * Description 2199c6b5fb86SQuentin Monnet * Attempt in a safe way to write *len* bytes from the buffer 2200c6b5fb86SQuentin Monnet * *src* to *dst* in memory. It only works for threads that are in 2201c6b5fb86SQuentin Monnet * user context, and *dst* must be a valid user space address. 2202c6b5fb86SQuentin Monnet * 2203c6b5fb86SQuentin Monnet * This helper should not be used to implement any kind of 2204c6b5fb86SQuentin Monnet * security mechanism because of TOC-TOU attacks, but rather to 2205c6b5fb86SQuentin Monnet * debug, divert, and manipulate execution of semi-cooperative 2206c6b5fb86SQuentin Monnet * processes. 2207c6b5fb86SQuentin Monnet * 2208c6b5fb86SQuentin Monnet * Keep in mind that this feature is meant for experiments, and it 2209c6b5fb86SQuentin Monnet * has a risk of crashing the system and running programs. 2210c6b5fb86SQuentin Monnet * Therefore, when an eBPF program using this helper is attached, 2211c6b5fb86SQuentin Monnet * a warning including PID and process name is printed to kernel 2212c6b5fb86SQuentin Monnet * logs. 2213c6b5fb86SQuentin Monnet * Return 2214c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2215c6b5fb86SQuentin Monnet * 2216bdb7b79bSAndrii Nakryiko * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2217c6b5fb86SQuentin Monnet * Description 2218c6b5fb86SQuentin Monnet * Check whether the probe is being run is the context of a given 2219c6b5fb86SQuentin Monnet * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2220c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2221c6b5fb86SQuentin Monnet * Return 2222c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 2223c6b5fb86SQuentin Monnet * 22241aef5b43SSong Liu * * 0, if current task belongs to the cgroup2. 22251aef5b43SSong Liu * * 1, if current task does not belong to the cgroup2. 2226c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 2227c6b5fb86SQuentin Monnet * 2228bdb7b79bSAndrii Nakryiko * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2229fa15601aSQuentin Monnet * Description 2230fa15601aSQuentin Monnet * Resize (trim or grow) the packet associated to *skb* to the 2231fa15601aSQuentin Monnet * new *len*. The *flags* are reserved for future usage, and must 2232fa15601aSQuentin Monnet * be left at zero. 2233fa15601aSQuentin Monnet * 2234fa15601aSQuentin Monnet * The basic idea is that the helper performs the needed work to 2235fa15601aSQuentin Monnet * change the size of the packet, then the eBPF program rewrites 2236fa15601aSQuentin Monnet * the rest via helpers like **bpf_skb_store_bytes**\ (), 2237fa15601aSQuentin Monnet * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2238fa15601aSQuentin Monnet * and others. This helper is a slow path utility intended for 2239fa15601aSQuentin Monnet * replies with control messages. And because it is targeted for 2240fa15601aSQuentin Monnet * slow path, the helper itself can afford to be slow: it 2241fa15601aSQuentin Monnet * implicitly linearizes, unclones and drops offloads from the 2242fa15601aSQuentin Monnet * *skb*. 2243fa15601aSQuentin Monnet * 224432e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2245fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2246fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2247fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2248fa15601aSQuentin Monnet * direct packet access. 2249fa15601aSQuentin Monnet * Return 2250fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2251fa15601aSQuentin Monnet * 2252bdb7b79bSAndrii Nakryiko * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2253fa15601aSQuentin Monnet * Description 2254fa15601aSQuentin Monnet * Pull in non-linear data in case the *skb* is non-linear and not 2255fa15601aSQuentin Monnet * all of *len* are part of the linear section. Make *len* bytes 2256fa15601aSQuentin Monnet * from *skb* readable and writable. If a zero value is passed for 2257fa15601aSQuentin Monnet * *len*, then the whole length of the *skb* is pulled. 2258fa15601aSQuentin Monnet * 2259fa15601aSQuentin Monnet * This helper is only needed for reading and writing with direct 2260fa15601aSQuentin Monnet * packet access. 2261fa15601aSQuentin Monnet * 2262fa15601aSQuentin Monnet * For direct packet access, testing that offsets to access 2263fa15601aSQuentin Monnet * are within packet boundaries (test on *skb*\ **->data_end**) is 2264fa15601aSQuentin Monnet * susceptible to fail if offsets are invalid, or if the requested 2265fa15601aSQuentin Monnet * data is in non-linear parts of the *skb*. On failure the 2266fa15601aSQuentin Monnet * program can just bail out, or in the case of a non-linear 2267fa15601aSQuentin Monnet * buffer, use a helper to make the data available. The 2268fa15601aSQuentin Monnet * **bpf_skb_load_bytes**\ () helper is a first solution to access 2269fa15601aSQuentin Monnet * the data. Another one consists in using **bpf_skb_pull_data** 2270fa15601aSQuentin Monnet * to pull in once the non-linear parts, then retesting and 2271fa15601aSQuentin Monnet * eventually access the data. 2272fa15601aSQuentin Monnet * 2273fa15601aSQuentin Monnet * At the same time, this also makes sure the *skb* is uncloned, 2274fa15601aSQuentin Monnet * which is a necessary condition for direct write. As this needs 2275fa15601aSQuentin Monnet * to be an invariant for the write part only, the verifier 2276fa15601aSQuentin Monnet * detects writes and adds a prologue that is calling 2277fa15601aSQuentin Monnet * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2278fa15601aSQuentin Monnet * the very beginning in case it is indeed cloned. 2279fa15601aSQuentin Monnet * 228032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2281fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2282fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2283fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2284fa15601aSQuentin Monnet * direct packet access. 2285fa15601aSQuentin Monnet * Return 2286fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2287fa15601aSQuentin Monnet * 2288fa15601aSQuentin Monnet * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2289fa15601aSQuentin Monnet * Description 2290fa15601aSQuentin Monnet * Add the checksum *csum* into *skb*\ **->csum** in case the 2291fa15601aSQuentin Monnet * driver has supplied a checksum for the entire packet into that 2292fa15601aSQuentin Monnet * field. Return an error otherwise. This helper is intended to be 2293fa15601aSQuentin Monnet * used in combination with **bpf_csum_diff**\ (), in particular 2294fa15601aSQuentin Monnet * when the checksum needs to be updated after data has been 2295fa15601aSQuentin Monnet * written into the packet through direct packet access. 2296fa15601aSQuentin Monnet * Return 2297fa15601aSQuentin Monnet * The checksum on success, or a negative error code in case of 2298fa15601aSQuentin Monnet * failure. 2299fa15601aSQuentin Monnet * 2300fa15601aSQuentin Monnet * void bpf_set_hash_invalid(struct sk_buff *skb) 2301fa15601aSQuentin Monnet * Description 2302fa15601aSQuentin Monnet * Invalidate the current *skb*\ **->hash**. It can be used after 2303fa15601aSQuentin Monnet * mangling on headers through direct packet access, in order to 2304fa15601aSQuentin Monnet * indicate that the hash is outdated and to trigger a 2305fa15601aSQuentin Monnet * recalculation the next time the kernel tries to access this 2306fa15601aSQuentin Monnet * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2307fa15601aSQuentin Monnet * 2308bdb7b79bSAndrii Nakryiko * long bpf_get_numa_node_id(void) 2309fa15601aSQuentin Monnet * Description 2310fa15601aSQuentin Monnet * Return the id of the current NUMA node. The primary use case 2311fa15601aSQuentin Monnet * for this helper is the selection of sockets for the local NUMA 2312fa15601aSQuentin Monnet * node, when the program is attached to sockets using the 2313fa15601aSQuentin Monnet * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2314fa15601aSQuentin Monnet * but the helper is also available to other eBPF program types, 2315fa15601aSQuentin Monnet * similarly to **bpf_get_smp_processor_id**\ (). 2316fa15601aSQuentin Monnet * Return 2317fa15601aSQuentin Monnet * The id of current NUMA node. 2318fa15601aSQuentin Monnet * 2319bdb7b79bSAndrii Nakryiko * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2320c6b5fb86SQuentin Monnet * Description 2321c6b5fb86SQuentin Monnet * Grows headroom of packet associated to *skb* and adjusts the 2322c6b5fb86SQuentin Monnet * offset of the MAC header accordingly, adding *len* bytes of 2323c6b5fb86SQuentin Monnet * space. It automatically extends and reallocates memory as 2324c6b5fb86SQuentin Monnet * required. 2325c6b5fb86SQuentin Monnet * 2326c6b5fb86SQuentin Monnet * This helper can be used on a layer 3 *skb* to push a MAC header 2327c6b5fb86SQuentin Monnet * for redirection into a layer 2 device. 2328c6b5fb86SQuentin Monnet * 2329c6b5fb86SQuentin Monnet * All values for *flags* are reserved for future usage, and must 2330c6b5fb86SQuentin Monnet * be left at zero. 2331c6b5fb86SQuentin Monnet * 233232e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2333c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2334c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 2335c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 2336c6b5fb86SQuentin Monnet * direct packet access. 2337c6b5fb86SQuentin Monnet * Return 2338c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2339c6b5fb86SQuentin Monnet * 2340bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2341c6b5fb86SQuentin Monnet * Description 2342c6b5fb86SQuentin Monnet * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2343c6b5fb86SQuentin Monnet * it is possible to use a negative value for *delta*. This helper 2344c6b5fb86SQuentin Monnet * can be used to prepare the packet for pushing or popping 2345c6b5fb86SQuentin Monnet * headers. 2346c6b5fb86SQuentin Monnet * 234732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2348c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2349c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 2350c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 2351c6b5fb86SQuentin Monnet * direct packet access. 2352c6b5fb86SQuentin Monnet * Return 2353c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2354c6b5fb86SQuentin Monnet * 2355bdb7b79bSAndrii Nakryiko * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2356c6b5fb86SQuentin Monnet * Description 23576ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe kernel address 2358ab8d7809SQuentin Monnet * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 23596ae08ae3SDaniel Borkmann * more details. 2360c6b5fb86SQuentin Monnet * 2361ab8d7809SQuentin Monnet * Generally, use **bpf_probe_read_user_str**\ () or 2362ab8d7809SQuentin Monnet * **bpf_probe_read_kernel_str**\ () instead. 2363c6b5fb86SQuentin Monnet * Return 2364c6b5fb86SQuentin Monnet * On success, the strictly positive length of the string, 2365c6b5fb86SQuentin Monnet * including the trailing NUL character. On error, a negative 2366c6b5fb86SQuentin Monnet * value. 2367c6b5fb86SQuentin Monnet * 2368c6b5fb86SQuentin Monnet * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2369c6b5fb86SQuentin Monnet * Description 2370c6b5fb86SQuentin Monnet * If the **struct sk_buff** pointed by *skb* has a known socket, 2371c6b5fb86SQuentin Monnet * retrieve the cookie (generated by the kernel) of this socket. 2372c6b5fb86SQuentin Monnet * If no cookie has been set yet, generate a new cookie. Once 2373c6b5fb86SQuentin Monnet * generated, the socket cookie remains stable for the life of the 2374c6b5fb86SQuentin Monnet * socket. This helper can be useful for monitoring per socket 2375cd48bddaSDaniel Borkmann * networking traffic statistics as it provides a global socket 2376cd48bddaSDaniel Borkmann * identifier that can be assumed unique. 2377c6b5fb86SQuentin Monnet * Return 237807881ccbSFlorent Revest * A 8-byte long unique number on success, or 0 if the socket 237907881ccbSFlorent Revest * field is missing inside *skb*. 2380c6b5fb86SQuentin Monnet * 2381d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2382d692f113SAndrey Ignatov * Description 2383d692f113SAndrey Ignatov * Equivalent to bpf_get_socket_cookie() helper that accepts 238462369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_addr** context. 2385d692f113SAndrey Ignatov * Return 238607881ccbSFlorent Revest * A 8-byte long unique number. 2387d692f113SAndrey Ignatov * 2388d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2389d692f113SAndrey Ignatov * Description 2390ab8d7809SQuentin Monnet * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 239162369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_ops** context. 2392d692f113SAndrey Ignatov * Return 239307881ccbSFlorent Revest * A 8-byte long unique number. 2394d692f113SAndrey Ignatov * 2395c5dbb89fSFlorent Revest * u64 bpf_get_socket_cookie(struct sock *sk) 2396c5dbb89fSFlorent Revest * Description 2397c5dbb89fSFlorent Revest * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2398c5dbb89fSFlorent Revest * *sk*, but gets socket from a BTF **struct sock**. This helper 2399c5dbb89fSFlorent Revest * also works for sleepable programs. 2400c5dbb89fSFlorent Revest * Return 2401c5dbb89fSFlorent Revest * A 8-byte long unique number or 0 if *sk* is NULL. 2402c5dbb89fSFlorent Revest * 2403c6b5fb86SQuentin Monnet * u32 bpf_get_socket_uid(struct sk_buff *skb) 2404c6b5fb86SQuentin Monnet * Return 2405c6b5fb86SQuentin Monnet * The owner UID of the socket associated to *skb*. If the socket 2406c6b5fb86SQuentin Monnet * is **NULL**, or if it is not a full socket (i.e. if it is a 2407c6b5fb86SQuentin Monnet * time-wait or a request socket instead), **overflowuid** value 2408c6b5fb86SQuentin Monnet * is returned (note that **overflowuid** might also be the actual 2409c6b5fb86SQuentin Monnet * UID value for the socket). 2410c6b5fb86SQuentin Monnet * 2411bdb7b79bSAndrii Nakryiko * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2412fa15601aSQuentin Monnet * Description 2413fa15601aSQuentin Monnet * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2414fa15601aSQuentin Monnet * to value *hash*. 2415fa15601aSQuentin Monnet * Return 2416fa15601aSQuentin Monnet * 0 2417fa15601aSQuentin Monnet * 2418bdb7b79bSAndrii Nakryiko * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 24197aa79a86SQuentin Monnet * Description 24207aa79a86SQuentin Monnet * Emulate a call to **setsockopt()** on the socket associated to 24217aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 24227aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 24237aa79a86SQuentin Monnet * must be specified, see **setsockopt(2)** for more information. 24247aa79a86SQuentin Monnet * The option value of length *optlen* is pointed by *optval*. 24257aa79a86SQuentin Monnet * 2426beecf11bSStanislav Fomichev * *bpf_socket* should be one of the following: 2427ab8d7809SQuentin Monnet * 2428beecf11bSStanislav Fomichev * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2429beecf11bSStanislav Fomichev * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2430beecf11bSStanislav Fomichev * and **BPF_CGROUP_INET6_CONNECT**. 2431beecf11bSStanislav Fomichev * 24327aa79a86SQuentin Monnet * This helper actually implements a subset of **setsockopt()**. 24337aa79a86SQuentin Monnet * It supports the following *level*\ s: 24347aa79a86SQuentin Monnet * 24357aa79a86SQuentin Monnet * * **SOL_SOCKET**, which supports the following *optname*\ s: 24367aa79a86SQuentin Monnet * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2437f9bcf968SDmitry Yakunin * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2438f9bcf968SDmitry Yakunin * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 24397aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports the following *optname*\ s: 24407aa79a86SQuentin Monnet * **TCP_CONGESTION**, **TCP_BPF_IW**, 2441f9bcf968SDmitry Yakunin * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2442f9bcf968SDmitry Yakunin * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2443eca43ee6SNikita V. Shirokov * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**. 24447aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 24457aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 24467aa79a86SQuentin Monnet * Return 24477aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 24487aa79a86SQuentin Monnet * 2449bdb7b79bSAndrii Nakryiko * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2450fa15601aSQuentin Monnet * Description 2451fa15601aSQuentin Monnet * Grow or shrink the room for data in the packet associated to 2452fa15601aSQuentin Monnet * *skb* by *len_diff*, and according to the selected *mode*. 2453fa15601aSQuentin Monnet * 2454836e66c2SDaniel Borkmann * By default, the helper will reset any offloaded checksum 2455836e66c2SDaniel Borkmann * indicator of the skb to CHECKSUM_NONE. This can be avoided 2456836e66c2SDaniel Borkmann * by the following flag: 2457836e66c2SDaniel Borkmann * 2458836e66c2SDaniel Borkmann * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2459836e66c2SDaniel Borkmann * checksum data of the skb to CHECKSUM_NONE. 2460836e66c2SDaniel Borkmann * 246114aa3192SWillem de Bruijn * There are two supported modes at this time: 246214aa3192SWillem de Bruijn * 246314aa3192SWillem de Bruijn * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 246414aa3192SWillem de Bruijn * (room space is added or removed below the layer 2 header). 2465fa15601aSQuentin Monnet * 2466fa15601aSQuentin Monnet * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2467fa15601aSQuentin Monnet * (room space is added or removed below the layer 3 header). 2468fa15601aSQuentin Monnet * 2469868d5235SWillem de Bruijn * The following flags are supported at this time: 24702278f6ccSWillem de Bruijn * 24712278f6ccSWillem de Bruijn * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 24722278f6ccSWillem de Bruijn * Adjusting mss in this way is not allowed for datagrams. 2473fa15601aSQuentin Monnet * 247480867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 247580867c5eSQuentin Monnet * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2476868d5235SWillem de Bruijn * Any new space is reserved to hold a tunnel header. 2477868d5235SWillem de Bruijn * Configure skb offsets and other fields accordingly. 2478868d5235SWillem de Bruijn * 247980867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 248080867c5eSQuentin Monnet * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2481868d5235SWillem de Bruijn * Use with ENCAP_L3 flags to further specify the tunnel type. 2482868d5235SWillem de Bruijn * 248380867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 248458dfc900SAlan Maguire * Use with ENCAP_L3/L4 flags to further specify the tunnel 248580867c5eSQuentin Monnet * type; *len* is the length of the inner MAC header. 248658dfc900SAlan Maguire * 248732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2488fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2489fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2490fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2491fa15601aSQuentin Monnet * direct packet access. 2492fa15601aSQuentin Monnet * Return 2493fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2494fa15601aSQuentin Monnet * 2495bdb7b79bSAndrii Nakryiko * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 2496ab127040SQuentin Monnet * Description 2497ab127040SQuentin Monnet * Redirect the packet to the endpoint referenced by *map* at 2498ab127040SQuentin Monnet * index *key*. Depending on its type, this *map* can contain 2499ab127040SQuentin Monnet * references to net devices (for forwarding packets through other 2500ab127040SQuentin Monnet * ports), or to CPUs (for redirecting XDP frames to another CPU; 2501ab127040SQuentin Monnet * but this is only implemented for native XDP (with driver 2502ab127040SQuentin Monnet * support) as of this writing). 2503ab127040SQuentin Monnet * 250443e74c02SToke Høiland-Jørgensen * The lower two bits of *flags* are used as the return code if 250543e74c02SToke Høiland-Jørgensen * the map lookup fails. This is so that the return value can be 2506ab8d7809SQuentin Monnet * one of the XDP program return codes up to **XDP_TX**, as chosen 2507ab8d7809SQuentin Monnet * by the caller. Any higher bits in the *flags* argument must be 250843e74c02SToke Høiland-Jørgensen * unset. 2509ab127040SQuentin Monnet * 2510ab8d7809SQuentin Monnet * See also **bpf_redirect**\ (), which only supports redirecting 2511ab8d7809SQuentin Monnet * to an ifindex, but doesn't require a map to do so. 2512ab127040SQuentin Monnet * Return 2513f25975f4SToke Høiland-Jørgensen * **XDP_REDIRECT** on success, or the value of the two lower bits 2514a33d3147SJakub Wilk * of the *flags* argument on error. 2515ab127040SQuentin Monnet * 2516bdb7b79bSAndrii Nakryiko * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2517ab127040SQuentin Monnet * Description 2518ab127040SQuentin Monnet * Redirect the packet to the socket referenced by *map* (of type 2519ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2520ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 2521ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 2522ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 2523ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 2524ab127040SQuentin Monnet * Return 2525ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 2526ab127040SQuentin Monnet * 2527bdb7b79bSAndrii Nakryiko * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2528ab127040SQuentin Monnet * Description 2529ab127040SQuentin Monnet * Add an entry to, or update a *map* referencing sockets. The 2530ab127040SQuentin Monnet * *skops* is used as a new value for the entry associated to 2531ab127040SQuentin Monnet * *key*. *flags* is one of: 2532ab127040SQuentin Monnet * 2533ab127040SQuentin Monnet * **BPF_NOEXIST** 2534ab127040SQuentin Monnet * The entry for *key* must not exist in the map. 2535ab127040SQuentin Monnet * **BPF_EXIST** 2536ab127040SQuentin Monnet * The entry for *key* must already exist in the map. 2537ab127040SQuentin Monnet * **BPF_ANY** 2538ab127040SQuentin Monnet * No condition on the existence of the entry for *key*. 2539ab127040SQuentin Monnet * 2540ab127040SQuentin Monnet * If the *map* has eBPF programs (parser and verdict), those will 2541ab127040SQuentin Monnet * be inherited by the socket being added. If the socket is 2542ab127040SQuentin Monnet * already attached to eBPF programs, this results in an error. 2543ab127040SQuentin Monnet * Return 2544ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 2545ab127040SQuentin Monnet * 2546bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2547fa15601aSQuentin Monnet * Description 2548fa15601aSQuentin Monnet * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2549fa15601aSQuentin Monnet * *delta* (which can be positive or negative). Note that this 2550fa15601aSQuentin Monnet * operation modifies the address stored in *xdp_md*\ **->data**, 2551fa15601aSQuentin Monnet * so the latter must be loaded only after the helper has been 2552fa15601aSQuentin Monnet * called. 2553fa15601aSQuentin Monnet * 2554fa15601aSQuentin Monnet * The use of *xdp_md*\ **->data_meta** is optional and programs 2555fa15601aSQuentin Monnet * are not required to use it. The rationale is that when the 2556fa15601aSQuentin Monnet * packet is processed with XDP (e.g. as DoS filter), it is 2557fa15601aSQuentin Monnet * possible to push further meta data along with it before passing 2558fa15601aSQuentin Monnet * to the stack, and to give the guarantee that an ingress eBPF 2559fa15601aSQuentin Monnet * program attached as a TC classifier on the same device can pick 2560fa15601aSQuentin Monnet * this up for further post-processing. Since TC works with socket 2561fa15601aSQuentin Monnet * buffers, it remains possible to set from XDP the **mark** or 2562fa15601aSQuentin Monnet * **priority** pointers, or other pointers for the socket buffer. 2563fa15601aSQuentin Monnet * Having this scratch space generic and programmable allows for 2564fa15601aSQuentin Monnet * more flexibility as the user is free to store whatever meta 2565fa15601aSQuentin Monnet * data they need. 2566fa15601aSQuentin Monnet * 256732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2568fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2569fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2570fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2571fa15601aSQuentin Monnet * direct packet access. 2572fa15601aSQuentin Monnet * Return 2573fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 25747aa79a86SQuentin Monnet * 2575bdb7b79bSAndrii Nakryiko * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 25767aa79a86SQuentin Monnet * Description 25777aa79a86SQuentin Monnet * Read the value of a perf event counter, and store it into *buf* 25787aa79a86SQuentin Monnet * of size *buf_size*. This helper relies on a *map* of type 25797aa79a86SQuentin Monnet * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 25807aa79a86SQuentin Monnet * counter is selected when *map* is updated with perf event file 25817aa79a86SQuentin Monnet * descriptors. The *map* is an array whose size is the number of 25827aa79a86SQuentin Monnet * available CPUs, and each cell contains a value relative to one 25837aa79a86SQuentin Monnet * CPU. The value to retrieve is indicated by *flags*, that 25847aa79a86SQuentin Monnet * contains the index of the CPU to look up, masked with 25857aa79a86SQuentin Monnet * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 25867aa79a86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 25877aa79a86SQuentin Monnet * current CPU should be retrieved. 25887aa79a86SQuentin Monnet * 25897aa79a86SQuentin Monnet * This helper behaves in a way close to 25907aa79a86SQuentin Monnet * **bpf_perf_event_read**\ () helper, save that instead of 25917aa79a86SQuentin Monnet * just returning the value observed, it fills the *buf* 25927aa79a86SQuentin Monnet * structure. This allows for additional data to be retrieved: in 25937aa79a86SQuentin Monnet * particular, the enabled and running times (in *buf*\ 25947aa79a86SQuentin Monnet * **->enabled** and *buf*\ **->running**, respectively) are 25957aa79a86SQuentin Monnet * copied. In general, **bpf_perf_event_read_value**\ () is 25967aa79a86SQuentin Monnet * recommended over **bpf_perf_event_read**\ (), which has some 25977aa79a86SQuentin Monnet * ABI issues and provides fewer functionalities. 25987aa79a86SQuentin Monnet * 25997aa79a86SQuentin Monnet * These values are interesting, because hardware PMU (Performance 26007aa79a86SQuentin Monnet * Monitoring Unit) counters are limited resources. When there are 26017aa79a86SQuentin Monnet * more PMU based perf events opened than available counters, 26027aa79a86SQuentin Monnet * kernel will multiplex these events so each event gets certain 26037aa79a86SQuentin Monnet * percentage (but not all) of the PMU time. In case that 26047aa79a86SQuentin Monnet * multiplexing happens, the number of samples or counter value 26057aa79a86SQuentin Monnet * will not reflect the case compared to when no multiplexing 26067aa79a86SQuentin Monnet * occurs. This makes comparison between different runs difficult. 26077aa79a86SQuentin Monnet * Typically, the counter value should be normalized before 26087aa79a86SQuentin Monnet * comparing to other experiments. The usual normalization is done 26097aa79a86SQuentin Monnet * as follows. 26107aa79a86SQuentin Monnet * 26117aa79a86SQuentin Monnet * :: 26127aa79a86SQuentin Monnet * 26137aa79a86SQuentin Monnet * normalized_counter = counter * t_enabled / t_running 26147aa79a86SQuentin Monnet * 26157aa79a86SQuentin Monnet * Where t_enabled is the time enabled for event and t_running is 26167aa79a86SQuentin Monnet * the time running for event since last normalization. The 26177aa79a86SQuentin Monnet * enabled and running times are accumulated since the perf event 26187aa79a86SQuentin Monnet * open. To achieve scaling factor between two invocations of an 2619ab8d7809SQuentin Monnet * eBPF program, users can use CPU id as the key (which is 26207aa79a86SQuentin Monnet * typical for perf array usage model) to remember the previous 26217aa79a86SQuentin Monnet * value and do the calculation inside the eBPF program. 26227aa79a86SQuentin Monnet * Return 26237aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 26247aa79a86SQuentin Monnet * 2625bdb7b79bSAndrii Nakryiko * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 26267aa79a86SQuentin Monnet * Description 26277aa79a86SQuentin Monnet * For en eBPF program attached to a perf event, retrieve the 26287aa79a86SQuentin Monnet * value of the event counter associated to *ctx* and store it in 26297aa79a86SQuentin Monnet * the structure pointed by *buf* and of size *buf_size*. Enabled 26307aa79a86SQuentin Monnet * and running times are also stored in the structure (see 26317aa79a86SQuentin Monnet * description of helper **bpf_perf_event_read_value**\ () for 26327aa79a86SQuentin Monnet * more details). 26337aa79a86SQuentin Monnet * Return 26347aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 26357aa79a86SQuentin Monnet * 2636bdb7b79bSAndrii Nakryiko * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 26377aa79a86SQuentin Monnet * Description 26387aa79a86SQuentin Monnet * Emulate a call to **getsockopt()** on the socket associated to 26397aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 26407aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 26417aa79a86SQuentin Monnet * must be specified, see **getsockopt(2)** for more information. 26427aa79a86SQuentin Monnet * The retrieved value is stored in the structure pointed by 26437aa79a86SQuentin Monnet * *opval* and of length *optlen*. 26447aa79a86SQuentin Monnet * 2645beecf11bSStanislav Fomichev * *bpf_socket* should be one of the following: 2646ab8d7809SQuentin Monnet * 2647beecf11bSStanislav Fomichev * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2648beecf11bSStanislav Fomichev * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2649beecf11bSStanislav Fomichev * and **BPF_CGROUP_INET6_CONNECT**. 2650beecf11bSStanislav Fomichev * 26517aa79a86SQuentin Monnet * This helper actually implements a subset of **getsockopt()**. 26527aa79a86SQuentin Monnet * It supports the following *level*\ s: 26537aa79a86SQuentin Monnet * 26547aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports *optname* 26557aa79a86SQuentin Monnet * **TCP_CONGESTION**. 26567aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 26577aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 26587aa79a86SQuentin Monnet * Return 26597aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 26607aa79a86SQuentin Monnet * 2661bdb7b79bSAndrii Nakryiko * long bpf_override_return(struct pt_regs *regs, u64 rc) 26627aa79a86SQuentin Monnet * Description 26637aa79a86SQuentin Monnet * Used for error injection, this helper uses kprobes to override 26647aa79a86SQuentin Monnet * the return value of the probed function, and to set it to *rc*. 26657aa79a86SQuentin Monnet * The first argument is the context *regs* on which the kprobe 26667aa79a86SQuentin Monnet * works. 26677aa79a86SQuentin Monnet * 2668ab8d7809SQuentin Monnet * This helper works by setting the PC (program counter) 26697aa79a86SQuentin Monnet * to an override function which is run in place of the original 26707aa79a86SQuentin Monnet * probed function. This means the probed function is not run at 26717aa79a86SQuentin Monnet * all. The replacement function just returns with the required 26727aa79a86SQuentin Monnet * value. 26737aa79a86SQuentin Monnet * 26747aa79a86SQuentin Monnet * This helper has security implications, and thus is subject to 26757aa79a86SQuentin Monnet * restrictions. It is only available if the kernel was compiled 26767aa79a86SQuentin Monnet * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 26777aa79a86SQuentin Monnet * option, and in this case it only works on functions tagged with 26787aa79a86SQuentin Monnet * **ALLOW_ERROR_INJECTION** in the kernel code. 26797aa79a86SQuentin Monnet * 26807aa79a86SQuentin Monnet * Also, the helper is only available for the architectures having 26817aa79a86SQuentin Monnet * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 26827aa79a86SQuentin Monnet * x86 architecture is the only one to support this feature. 26837aa79a86SQuentin Monnet * Return 26847aa79a86SQuentin Monnet * 0 26857aa79a86SQuentin Monnet * 2686bdb7b79bSAndrii Nakryiko * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 26877aa79a86SQuentin Monnet * Description 26887aa79a86SQuentin Monnet * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 26897aa79a86SQuentin Monnet * for the full TCP socket associated to *bpf_sock_ops* to 26907aa79a86SQuentin Monnet * *argval*. 26917aa79a86SQuentin Monnet * 26927aa79a86SQuentin Monnet * The primary use of this field is to determine if there should 26937aa79a86SQuentin Monnet * be calls to eBPF programs of type 26947aa79a86SQuentin Monnet * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 26957aa79a86SQuentin Monnet * code. A program of the same type can change its value, per 26967aa79a86SQuentin Monnet * connection and as necessary, when the connection is 26977aa79a86SQuentin Monnet * established. This field is directly accessible for reading, but 26987aa79a86SQuentin Monnet * this helper must be used for updates in order to return an 26997aa79a86SQuentin Monnet * error if an eBPF program tries to set a callback that is not 27007aa79a86SQuentin Monnet * supported in the current kernel. 27017aa79a86SQuentin Monnet * 2702725721a6SViet Hoang Tran * *argval* is a flag array which can combine these flags: 27037aa79a86SQuentin Monnet * 27047aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 27057aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 27067aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 270723729ff2SStanislav Fomichev * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 27087aa79a86SQuentin Monnet * 2709725721a6SViet Hoang Tran * Therefore, this function can be used to clear a callback flag by 2710725721a6SViet Hoang Tran * setting the appropriate bit to zero. e.g. to disable the RTO 2711725721a6SViet Hoang Tran * callback: 2712725721a6SViet Hoang Tran * 2713725721a6SViet Hoang Tran * **bpf_sock_ops_cb_flags_set(bpf_sock,** 2714725721a6SViet Hoang Tran * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 2715725721a6SViet Hoang Tran * 27167aa79a86SQuentin Monnet * Here are some examples of where one could call such eBPF 27177aa79a86SQuentin Monnet * program: 27187aa79a86SQuentin Monnet * 27197aa79a86SQuentin Monnet * * When RTO fires. 27207aa79a86SQuentin Monnet * * When a packet is retransmitted. 27217aa79a86SQuentin Monnet * * When the connection terminates. 27227aa79a86SQuentin Monnet * * When a packet is sent. 27237aa79a86SQuentin Monnet * * When a packet is received. 27247aa79a86SQuentin Monnet * Return 27257aa79a86SQuentin Monnet * Code **-EINVAL** if the socket is not a full TCP socket; 27267aa79a86SQuentin Monnet * otherwise, a positive number containing the bits that could not 27277aa79a86SQuentin Monnet * be set is returned (which comes down to 0 if all bits were set 27287aa79a86SQuentin Monnet * as required). 27297aa79a86SQuentin Monnet * 2730bdb7b79bSAndrii Nakryiko * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 2731ab127040SQuentin Monnet * Description 2732ab127040SQuentin Monnet * This helper is used in programs implementing policies at the 2733ab127040SQuentin Monnet * socket level. If the message *msg* is allowed to pass (i.e. if 2734ab127040SQuentin Monnet * the verdict eBPF program returns **SK_PASS**), redirect it to 2735ab127040SQuentin Monnet * the socket referenced by *map* (of type 2736ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2737ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 2738ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 2739ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 2740ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 2741ab127040SQuentin Monnet * Return 2742ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 2743ab127040SQuentin Monnet * 2744bdb7b79bSAndrii Nakryiko * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 2745ab127040SQuentin Monnet * Description 2746ab127040SQuentin Monnet * For socket policies, apply the verdict of the eBPF program to 2747ab127040SQuentin Monnet * the next *bytes* (number of bytes) of message *msg*. 2748ab127040SQuentin Monnet * 2749ab127040SQuentin Monnet * For example, this helper can be used in the following cases: 2750ab127040SQuentin Monnet * 2751ab127040SQuentin Monnet * * A single **sendmsg**\ () or **sendfile**\ () system call 2752ab127040SQuentin Monnet * contains multiple logical messages that the eBPF program is 2753ab127040SQuentin Monnet * supposed to read and for which it should apply a verdict. 2754ab127040SQuentin Monnet * * An eBPF program only cares to read the first *bytes* of a 2755ab127040SQuentin Monnet * *msg*. If the message has a large payload, then setting up 2756ab127040SQuentin Monnet * and calling the eBPF program repeatedly for all bytes, even 2757ab127040SQuentin Monnet * though the verdict is already known, would create unnecessary 2758ab127040SQuentin Monnet * overhead. 2759ab127040SQuentin Monnet * 2760ab127040SQuentin Monnet * When called from within an eBPF program, the helper sets a 2761ab127040SQuentin Monnet * counter internal to the BPF infrastructure, that is used to 2762ab127040SQuentin Monnet * apply the last verdict to the next *bytes*. If *bytes* is 2763ab127040SQuentin Monnet * smaller than the current data being processed from a 2764ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () system call, the first 2765ab127040SQuentin Monnet * *bytes* will be sent and the eBPF program will be re-run with 2766ab127040SQuentin Monnet * the pointer for start of data pointing to byte number *bytes* 2767ab127040SQuentin Monnet * **+ 1**. If *bytes* is larger than the current data being 2768ab127040SQuentin Monnet * processed, then the eBPF verdict will be applied to multiple 2769ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2770ab127040SQuentin Monnet * consumed. 2771ab127040SQuentin Monnet * 2772ab127040SQuentin Monnet * Note that if a socket closes with the internal counter holding 2773ab127040SQuentin Monnet * a non-zero value, this is not a problem because data is not 2774ab127040SQuentin Monnet * being buffered for *bytes* and is sent as it is received. 2775ab127040SQuentin Monnet * Return 2776ab127040SQuentin Monnet * 0 2777ab127040SQuentin Monnet * 2778bdb7b79bSAndrii Nakryiko * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 2779ab127040SQuentin Monnet * Description 2780ab127040SQuentin Monnet * For socket policies, prevent the execution of the verdict eBPF 2781ab127040SQuentin Monnet * program for message *msg* until *bytes* (byte number) have been 2782ab127040SQuentin Monnet * accumulated. 2783ab127040SQuentin Monnet * 2784ab127040SQuentin Monnet * This can be used when one needs a specific number of bytes 2785ab127040SQuentin Monnet * before a verdict can be assigned, even if the data spans 2786ab127040SQuentin Monnet * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 2787ab127040SQuentin Monnet * case would be a user calling **sendmsg**\ () repeatedly with 2788ab127040SQuentin Monnet * 1-byte long message segments. Obviously, this is bad for 2789ab127040SQuentin Monnet * performance, but it is still valid. If the eBPF program needs 2790ab127040SQuentin Monnet * *bytes* bytes to validate a header, this helper can be used to 2791ab127040SQuentin Monnet * prevent the eBPF program to be called again until *bytes* have 2792ab127040SQuentin Monnet * been accumulated. 2793ab127040SQuentin Monnet * Return 2794ab127040SQuentin Monnet * 0 2795ab127040SQuentin Monnet * 2796bdb7b79bSAndrii Nakryiko * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2797ab127040SQuentin Monnet * Description 2798ab127040SQuentin Monnet * For socket policies, pull in non-linear data from user space 2799ab127040SQuentin Monnet * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2800ab127040SQuentin Monnet * **->data_end** to *start* and *end* bytes offsets into *msg*, 2801ab127040SQuentin Monnet * respectively. 2802ab127040SQuentin Monnet * 2803ab127040SQuentin Monnet * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2804ab127040SQuentin Monnet * *msg* it can only parse data that the (**data**, **data_end**) 2805ab127040SQuentin Monnet * pointers have already consumed. For **sendmsg**\ () hooks this 2806ab127040SQuentin Monnet * is likely the first scatterlist element. But for calls relying 2807ab127040SQuentin Monnet * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2808ab127040SQuentin Monnet * be the range (**0**, **0**) because the data is shared with 2809ab127040SQuentin Monnet * user space and by default the objective is to avoid allowing 2810ab127040SQuentin Monnet * user space to modify data while (or after) eBPF verdict is 2811ab127040SQuentin Monnet * being decided. This helper can be used to pull in data and to 2812ab127040SQuentin Monnet * set the start and end pointer to given values. Data will be 2813ab127040SQuentin Monnet * copied if necessary (i.e. if data was not linear and if start 2814ab127040SQuentin Monnet * and end pointers do not point to the same chunk). 2815ab127040SQuentin Monnet * 281632e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2817ab127040SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2818ab127040SQuentin Monnet * previously done by the verifier are invalidated and must be 2819ab127040SQuentin Monnet * performed again, if the helper is used in combination with 2820ab127040SQuentin Monnet * direct packet access. 2821ab127040SQuentin Monnet * 2822ab127040SQuentin Monnet * All values for *flags* are reserved for future usage, and must 2823ab127040SQuentin Monnet * be left at zero. 2824ab127040SQuentin Monnet * Return 2825ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 2826ab127040SQuentin Monnet * 2827bdb7b79bSAndrii Nakryiko * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 28287aa79a86SQuentin Monnet * Description 28297aa79a86SQuentin Monnet * Bind the socket associated to *ctx* to the address pointed by 28307aa79a86SQuentin Monnet * *addr*, of length *addr_len*. This allows for making outgoing 28317aa79a86SQuentin Monnet * connection from the desired IP address, which can be useful for 28327aa79a86SQuentin Monnet * example when all processes inside a cgroup should use one 28337aa79a86SQuentin Monnet * single IP address on a host that has multiple IP configured. 28347aa79a86SQuentin Monnet * 28357aa79a86SQuentin Monnet * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 28367aa79a86SQuentin Monnet * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 28378086fbafSStanislav Fomichev * **AF_INET6**). It's advised to pass zero port (**sin_port** 28388086fbafSStanislav Fomichev * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 28398086fbafSStanislav Fomichev * behavior and lets the kernel efficiently pick up an unused 28408086fbafSStanislav Fomichev * port as long as 4-tuple is unique. Passing non-zero port might 28418086fbafSStanislav Fomichev * lead to degraded performance. 28427aa79a86SQuentin Monnet * Return 28437aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 28442d020dd7SQuentin Monnet * 2845bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 28462d020dd7SQuentin Monnet * Description 28472d020dd7SQuentin Monnet * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2848c8741e2bSJesper Dangaard Brouer * possible to both shrink and grow the packet tail. 2849c8741e2bSJesper Dangaard Brouer * Shrink done via *delta* being a negative integer. 28502d020dd7SQuentin Monnet * 285132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 28522d020dd7SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 28532d020dd7SQuentin Monnet * previously done by the verifier are invalidated and must be 28542d020dd7SQuentin Monnet * performed again, if the helper is used in combination with 28552d020dd7SQuentin Monnet * direct packet access. 28562d020dd7SQuentin Monnet * Return 28572d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 28582d020dd7SQuentin Monnet * 2859bdb7b79bSAndrii Nakryiko * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 28602d020dd7SQuentin Monnet * Description 28612d020dd7SQuentin Monnet * Retrieve the XFRM state (IP transform framework, see also 28622d020dd7SQuentin Monnet * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 28632d020dd7SQuentin Monnet * 28642d020dd7SQuentin Monnet * The retrieved value is stored in the **struct bpf_xfrm_state** 28652d020dd7SQuentin Monnet * pointed by *xfrm_state* and of length *size*. 28662d020dd7SQuentin Monnet * 28672d020dd7SQuentin Monnet * All values for *flags* are reserved for future usage, and must 28682d020dd7SQuentin Monnet * be left at zero. 28692d020dd7SQuentin Monnet * 28702d020dd7SQuentin Monnet * This helper is available only if the kernel was compiled with 28712d020dd7SQuentin Monnet * **CONFIG_XFRM** configuration option. 28722d020dd7SQuentin Monnet * Return 28732d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 2874c195651eSYonghong Song * 2875bdb7b79bSAndrii Nakryiko * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2876c195651eSYonghong Song * Description 2877c195651eSYonghong Song * Return a user or a kernel stack in bpf program provided buffer. 2878c195651eSYonghong Song * To achieve this, the helper needs *ctx*, which is a pointer 2879c195651eSYonghong Song * to the context on which the tracing program is executed. 2880c195651eSYonghong Song * To store the stacktrace, the bpf program provides *buf* with 2881c195651eSYonghong Song * a nonnegative *size*. 2882c195651eSYonghong Song * 2883c195651eSYonghong Song * The last argument, *flags*, holds the number of stack frames to 2884c195651eSYonghong Song * skip (from 0 to 255), masked with 2885c195651eSYonghong Song * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2886c195651eSYonghong Song * the following flags: 2887c195651eSYonghong Song * 2888c195651eSYonghong Song * **BPF_F_USER_STACK** 2889c195651eSYonghong Song * Collect a user space stack instead of a kernel stack. 2890c195651eSYonghong Song * **BPF_F_USER_BUILD_ID** 2891c195651eSYonghong Song * Collect buildid+offset instead of ips for user stack, 2892c195651eSYonghong Song * only valid if **BPF_F_USER_STACK** is also specified. 2893c195651eSYonghong Song * 2894c195651eSYonghong Song * **bpf_get_stack**\ () can collect up to 2895c195651eSYonghong Song * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 2896c195651eSYonghong Song * to sufficient large buffer size. Note that 2897c195651eSYonghong Song * this limit can be controlled with the **sysctl** program, and 2898c195651eSYonghong Song * that it should be manually increased in order to profile long 2899c195651eSYonghong Song * user stacks (such as stacks for Java programs). To do so, use: 2900c195651eSYonghong Song * 2901c195651eSYonghong Song * :: 2902c195651eSYonghong Song * 2903c195651eSYonghong Song * # sysctl kernel.perf_event_max_stack=<new value> 2904c195651eSYonghong Song * Return 29057a279e93SQuentin Monnet * A non-negative value equal to or less than *size* on success, 29067a279e93SQuentin Monnet * or a negative error in case of failure. 29074e1ec56cSDaniel Borkmann * 2908bdb7b79bSAndrii Nakryiko * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 29094e1ec56cSDaniel Borkmann * Description 29104e1ec56cSDaniel Borkmann * This helper is similar to **bpf_skb_load_bytes**\ () in that 29114e1ec56cSDaniel Borkmann * it provides an easy way to load *len* bytes from *offset* 29124e1ec56cSDaniel Borkmann * from the packet associated to *skb*, into the buffer pointed 29134e1ec56cSDaniel Borkmann * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 29144e1ec56cSDaniel Borkmann * a fifth argument *start_header* exists in order to select a 29154e1ec56cSDaniel Borkmann * base offset to start from. *start_header* can be one of: 29164e1ec56cSDaniel Borkmann * 29174e1ec56cSDaniel Borkmann * **BPF_HDR_START_MAC** 29184e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s mac header. 29194e1ec56cSDaniel Borkmann * **BPF_HDR_START_NET** 29204e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s network header. 29214e1ec56cSDaniel Borkmann * 29224e1ec56cSDaniel Borkmann * In general, "direct packet access" is the preferred method to 29234e1ec56cSDaniel Borkmann * access packet data, however, this helper is in particular useful 29244e1ec56cSDaniel Borkmann * in socket filters where *skb*\ **->data** does not always point 29254e1ec56cSDaniel Borkmann * to the start of the mac header and where "direct packet access" 29264e1ec56cSDaniel Borkmann * is not available. 29274e1ec56cSDaniel Borkmann * Return 29284e1ec56cSDaniel Borkmann * 0 on success, or a negative error in case of failure. 29294e1ec56cSDaniel Borkmann * 2930bdb7b79bSAndrii Nakryiko * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 293187f5fc7eSDavid Ahern * Description 293287f5fc7eSDavid Ahern * Do FIB lookup in kernel tables using parameters in *params*. 293387f5fc7eSDavid Ahern * If lookup is successful and result shows packet is to be 293487f5fc7eSDavid Ahern * forwarded, the neighbor tables are searched for the nexthop. 293587f5fc7eSDavid Ahern * If successful (ie., FIB lookup shows forwarding and nexthop 2936fa898d76SDavid Ahern * is resolved), the nexthop address is returned in ipv4_dst 2937fa898d76SDavid Ahern * or ipv6_dst based on family, smac is set to mac address of 2938fa898d76SDavid Ahern * egress device, dmac is set to nexthop mac address, rt_metric 29394c79579bSDavid Ahern * is set to metric from route (IPv4/IPv6 only), and ifindex 29404c79579bSDavid Ahern * is set to the device index of the nexthop from the FIB lookup. 294187f5fc7eSDavid Ahern * 294287f5fc7eSDavid Ahern * *plen* argument is the size of the passed in struct. 29437a279e93SQuentin Monnet * *flags* argument can be a combination of one or more of the 29447a279e93SQuentin Monnet * following values: 294587f5fc7eSDavid Ahern * 29467a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_DIRECT** 29477a279e93SQuentin Monnet * Do a direct table lookup vs full lookup using FIB 29487a279e93SQuentin Monnet * rules. 29497a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_OUTPUT** 29507a279e93SQuentin Monnet * Perform lookup from an egress perspective (default is 29517a279e93SQuentin Monnet * ingress). 295287f5fc7eSDavid Ahern * 295387f5fc7eSDavid Ahern * *ctx* is either **struct xdp_md** for XDP programs or 295487f5fc7eSDavid Ahern * **struct sk_buff** tc cls_act programs. 295587f5fc7eSDavid Ahern * Return 29564c79579bSDavid Ahern * * < 0 if any input argument is invalid 29574c79579bSDavid Ahern * * 0 on success (packet is forwarded, nexthop neighbor exists) 29584c79579bSDavid Ahern * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 29592bae79d2SQuentin Monnet * packet is not forwarded or needs assist from full stack 296081110384SJohn Fastabend * 2961e1850ea9SJesper Dangaard Brouer * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 2962e1850ea9SJesper Dangaard Brouer * was exceeded and output params->mtu_result contains the MTU. 2963e1850ea9SJesper Dangaard Brouer * 2964bdb7b79bSAndrii Nakryiko * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 296581110384SJohn Fastabend * Description 296681110384SJohn Fastabend * Add an entry to, or update a sockhash *map* referencing sockets. 296781110384SJohn Fastabend * The *skops* is used as a new value for the entry associated to 296881110384SJohn Fastabend * *key*. *flags* is one of: 296981110384SJohn Fastabend * 297081110384SJohn Fastabend * **BPF_NOEXIST** 297181110384SJohn Fastabend * The entry for *key* must not exist in the map. 297281110384SJohn Fastabend * **BPF_EXIST** 297381110384SJohn Fastabend * The entry for *key* must already exist in the map. 297481110384SJohn Fastabend * **BPF_ANY** 297581110384SJohn Fastabend * No condition on the existence of the entry for *key*. 297681110384SJohn Fastabend * 297781110384SJohn Fastabend * If the *map* has eBPF programs (parser and verdict), those will 297881110384SJohn Fastabend * be inherited by the socket being added. If the socket is 297981110384SJohn Fastabend * already attached to eBPF programs, this results in an error. 298081110384SJohn Fastabend * Return 298181110384SJohn Fastabend * 0 on success, or a negative error in case of failure. 298281110384SJohn Fastabend * 2983bdb7b79bSAndrii Nakryiko * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 298481110384SJohn Fastabend * Description 298581110384SJohn Fastabend * This helper is used in programs implementing policies at the 298681110384SJohn Fastabend * socket level. If the message *msg* is allowed to pass (i.e. if 298781110384SJohn Fastabend * the verdict eBPF program returns **SK_PASS**), redirect it to 298881110384SJohn Fastabend * the socket referenced by *map* (of type 298981110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 299081110384SJohn Fastabend * egress interfaces can be used for redirection. The 299181110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 299281110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 299381110384SJohn Fastabend * egress path otherwise). This is the only flag supported for now. 299481110384SJohn Fastabend * Return 299581110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 299681110384SJohn Fastabend * 2997bdb7b79bSAndrii Nakryiko * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 299881110384SJohn Fastabend * Description 299981110384SJohn Fastabend * This helper is used in programs implementing policies at the 300081110384SJohn Fastabend * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 300149f3d12bSJakub Wilk * if the verdict eBPF program returns **SK_PASS**), redirect it 300281110384SJohn Fastabend * to the socket referenced by *map* (of type 300381110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 300481110384SJohn Fastabend * egress interfaces can be used for redirection. The 300581110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 300681110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 300781110384SJohn Fastabend * egress otherwise). This is the only flag supported for now. 300881110384SJohn Fastabend * Return 300981110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 3010fe94cc29SMathieu Xhonneux * 3011bdb7b79bSAndrii Nakryiko * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3012fe94cc29SMathieu Xhonneux * Description 3013fe94cc29SMathieu Xhonneux * Encapsulate the packet associated to *skb* within a Layer 3 3014fe94cc29SMathieu Xhonneux * protocol header. This header is provided in the buffer at 3015fe94cc29SMathieu Xhonneux * address *hdr*, with *len* its size in bytes. *type* indicates 3016fe94cc29SMathieu Xhonneux * the protocol of the header and can be one of: 3017fe94cc29SMathieu Xhonneux * 3018fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6** 3019fe94cc29SMathieu Xhonneux * IPv6 encapsulation with Segment Routing Header 3020fe94cc29SMathieu Xhonneux * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3021fe94cc29SMathieu Xhonneux * the IPv6 header is computed by the kernel. 3022fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6_INLINE** 3023fe94cc29SMathieu Xhonneux * Only works if *skb* contains an IPv6 packet. Insert a 3024fe94cc29SMathieu Xhonneux * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3025fe94cc29SMathieu Xhonneux * the IPv6 header. 30263e0bd37cSPeter Oskolkov * **BPF_LWT_ENCAP_IP** 30273e0bd37cSPeter Oskolkov * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 30283e0bd37cSPeter Oskolkov * must be IPv4 or IPv6, followed by zero or more 302980867c5eSQuentin Monnet * additional headers, up to **LWT_BPF_MAX_HEADROOM** 303080867c5eSQuentin Monnet * total bytes in all prepended headers. Please note that 303180867c5eSQuentin Monnet * if **skb_is_gso**\ (*skb*) is true, no more than two 303280867c5eSQuentin Monnet * headers can be prepended, and the inner header, if 303380867c5eSQuentin Monnet * present, should be either GRE or UDP/GUE. 30343e0bd37cSPeter Oskolkov * 303580867c5eSQuentin Monnet * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 303680867c5eSQuentin Monnet * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 303780867c5eSQuentin Monnet * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 303880867c5eSQuentin Monnet * **BPF_PROG_TYPE_LWT_XMIT**. 3039fe94cc29SMathieu Xhonneux * 304032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3041fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3042fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3043fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3044fe94cc29SMathieu Xhonneux * direct packet access. 3045fe94cc29SMathieu Xhonneux * Return 3046fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3047fe94cc29SMathieu Xhonneux * 3048bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3049fe94cc29SMathieu Xhonneux * Description 3050fe94cc29SMathieu Xhonneux * Store *len* bytes from address *from* into the packet 3051fe94cc29SMathieu Xhonneux * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3052fe94cc29SMathieu Xhonneux * inside the outermost IPv6 Segment Routing Header can be 3053fe94cc29SMathieu Xhonneux * modified through this helper. 3054fe94cc29SMathieu Xhonneux * 305532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3056fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3057fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3058fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3059fe94cc29SMathieu Xhonneux * direct packet access. 3060fe94cc29SMathieu Xhonneux * Return 3061fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3062fe94cc29SMathieu Xhonneux * 3063bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3064fe94cc29SMathieu Xhonneux * Description 3065fe94cc29SMathieu Xhonneux * Adjust the size allocated to TLVs in the outermost IPv6 3066fe94cc29SMathieu Xhonneux * Segment Routing Header contained in the packet associated to 3067fe94cc29SMathieu Xhonneux * *skb*, at position *offset* by *delta* bytes. Only offsets 3068fe94cc29SMathieu Xhonneux * after the segments are accepted. *delta* can be as well 3069fe94cc29SMathieu Xhonneux * positive (growing) as negative (shrinking). 3070fe94cc29SMathieu Xhonneux * 307132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3072fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3073fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3074fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3075fe94cc29SMathieu Xhonneux * direct packet access. 3076fe94cc29SMathieu Xhonneux * Return 3077fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3078fe94cc29SMathieu Xhonneux * 3079bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3080fe94cc29SMathieu Xhonneux * Description 3081fe94cc29SMathieu Xhonneux * Apply an IPv6 Segment Routing action of type *action* to the 3082fe94cc29SMathieu Xhonneux * packet associated to *skb*. Each action takes a parameter 3083fe94cc29SMathieu Xhonneux * contained at address *param*, and of length *param_len* bytes. 3084fe94cc29SMathieu Xhonneux * *action* can be one of: 3085fe94cc29SMathieu Xhonneux * 3086fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_X** 3087fe94cc29SMathieu Xhonneux * End.X action: Endpoint with Layer-3 cross-connect. 3088fe94cc29SMathieu Xhonneux * Type of *param*: **struct in6_addr**. 3089fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_T** 3090fe94cc29SMathieu Xhonneux * End.T action: Endpoint with specific IPv6 table lookup. 3091fe94cc29SMathieu Xhonneux * Type of *param*: **int**. 3092fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6** 3093fe94cc29SMathieu Xhonneux * End.B6 action: Endpoint bound to an SRv6 policy. 309480867c5eSQuentin Monnet * Type of *param*: **struct ipv6_sr_hdr**. 3095fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3096fe94cc29SMathieu Xhonneux * End.B6.Encap action: Endpoint bound to an SRv6 3097fe94cc29SMathieu Xhonneux * encapsulation policy. 309880867c5eSQuentin Monnet * Type of *param*: **struct ipv6_sr_hdr**. 3099fe94cc29SMathieu Xhonneux * 310032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3101fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3102fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3103fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3104fe94cc29SMathieu Xhonneux * direct packet access. 3105fe94cc29SMathieu Xhonneux * Return 3106fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3107f4364dcfSSean Young * 3108bdb7b79bSAndrii Nakryiko * long bpf_rc_repeat(void *ctx) 310962369db2SQuentin Monnet * Description 311062369db2SQuentin Monnet * This helper is used in programs implementing IR decoding, to 311162369db2SQuentin Monnet * report a successfully decoded repeat key message. This delays 311262369db2SQuentin Monnet * the generation of a key up event for previously generated 311362369db2SQuentin Monnet * key down event. 311462369db2SQuentin Monnet * 311562369db2SQuentin Monnet * Some IR protocols like NEC have a special IR message for 311662369db2SQuentin Monnet * repeating last button, for when a button is held down. 311762369db2SQuentin Monnet * 311862369db2SQuentin Monnet * The *ctx* should point to the lirc sample as passed into 311962369db2SQuentin Monnet * the program. 312062369db2SQuentin Monnet * 312162369db2SQuentin Monnet * This helper is only available is the kernel was compiled with 312262369db2SQuentin Monnet * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 312362369db2SQuentin Monnet * "**y**". 312462369db2SQuentin Monnet * Return 312562369db2SQuentin Monnet * 0 312662369db2SQuentin Monnet * 3127bdb7b79bSAndrii Nakryiko * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3128f4364dcfSSean Young * Description 3129f4364dcfSSean Young * This helper is used in programs implementing IR decoding, to 3130f4364dcfSSean Young * report a successfully decoded key press with *scancode*, 3131f4364dcfSSean Young * *toggle* value in the given *protocol*. The scancode will be 3132f4364dcfSSean Young * translated to a keycode using the rc keymap, and reported as 3133f4364dcfSSean Young * an input key down event. After a period a key up event is 3134f4364dcfSSean Young * generated. This period can be extended by calling either 313590b1023fSQuentin Monnet * **bpf_rc_keydown**\ () again with the same values, or calling 313690b1023fSQuentin Monnet * **bpf_rc_repeat**\ (). 3137f4364dcfSSean Young * 3138f4364dcfSSean Young * Some protocols include a toggle bit, in case the button was 3139f4364dcfSSean Young * released and pressed again between consecutive scancodes. 3140f4364dcfSSean Young * 3141f4364dcfSSean Young * The *ctx* should point to the lirc sample as passed into 3142f4364dcfSSean Young * the program. 3143f4364dcfSSean Young * 3144f4364dcfSSean Young * The *protocol* is the decoded protocol number (see 3145f4364dcfSSean Young * **enum rc_proto** for some predefined values). 3146f4364dcfSSean Young * 3147f4364dcfSSean Young * This helper is only available is the kernel was compiled with 3148f4364dcfSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3149f4364dcfSSean Young * "**y**". 3150f4364dcfSSean Young * Return 3151f4364dcfSSean Young * 0 3152f4364dcfSSean Young * 315362369db2SQuentin Monnet * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3154cb20b08eSDaniel Borkmann * Description 3155cb20b08eSDaniel Borkmann * Return the cgroup v2 id of the socket associated with the *skb*. 3156cb20b08eSDaniel Borkmann * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3157cb20b08eSDaniel Borkmann * helper for cgroup v1 by providing a tag resp. identifier that 3158cb20b08eSDaniel Borkmann * can be matched on or used for map lookups e.g. to implement 3159cb20b08eSDaniel Borkmann * policy. The cgroup v2 id of a given path in the hierarchy is 3160cb20b08eSDaniel Borkmann * exposed in user space through the f_handle API in order to get 3161cb20b08eSDaniel Borkmann * to the same 64-bit id. 3162cb20b08eSDaniel Borkmann * 3163cb20b08eSDaniel Borkmann * This helper can be used on TC egress path, but not on ingress, 3164cb20b08eSDaniel Borkmann * and is available only if the kernel was compiled with the 3165cb20b08eSDaniel Borkmann * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3166cb20b08eSDaniel Borkmann * Return 3167cb20b08eSDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 3168bf6fa2c8SYonghong Song * 3169bf6fa2c8SYonghong Song * u64 bpf_get_current_cgroup_id(void) 3170bf6fa2c8SYonghong Song * Return 3171bf6fa2c8SYonghong Song * A 64-bit integer containing the current cgroup id based 3172bf6fa2c8SYonghong Song * on the cgroup within which the current task is running. 3173cd339431SRoman Gushchin * 317462369db2SQuentin Monnet * void *bpf_get_local_storage(void *map, u64 flags) 3175cd339431SRoman Gushchin * Description 3176cd339431SRoman Gushchin * Get the pointer to the local storage area. 3177cd339431SRoman Gushchin * The type and the size of the local storage is defined 3178cd339431SRoman Gushchin * by the *map* argument. 3179cd339431SRoman Gushchin * The *flags* meaning is specific for each map type, 3180cd339431SRoman Gushchin * and has to be 0 for cgroup local storage. 3181cd339431SRoman Gushchin * 318290b1023fSQuentin Monnet * Depending on the BPF program type, a local storage area 318390b1023fSQuentin Monnet * can be shared between multiple instances of the BPF program, 3184cd339431SRoman Gushchin * running simultaneously. 3185cd339431SRoman Gushchin * 3186cd339431SRoman Gushchin * A user should care about the synchronization by himself. 318791c960b0SBrendan Jackman * For example, by using the **BPF_ATOMIC** instructions to alter 3188cd339431SRoman Gushchin * the shared data. 3189cd339431SRoman Gushchin * Return 319090b1023fSQuentin Monnet * A pointer to the local storage area. 31912dbb9b9eSMartin KaFai Lau * 3192bdb7b79bSAndrii Nakryiko * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 31932dbb9b9eSMartin KaFai Lau * Description 319490b1023fSQuentin Monnet * Select a **SO_REUSEPORT** socket from a 319590b1023fSQuentin Monnet * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 319690b1023fSQuentin Monnet * It checks the selected socket is matching the incoming 319790b1023fSQuentin Monnet * request in the socket buffer. 31982dbb9b9eSMartin KaFai Lau * Return 31992dbb9b9eSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 32006acc9b43SJoe Stringer * 320162369db2SQuentin Monnet * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 320262369db2SQuentin Monnet * Description 320362369db2SQuentin Monnet * Return id of cgroup v2 that is ancestor of cgroup associated 320462369db2SQuentin Monnet * with the *skb* at the *ancestor_level*. The root cgroup is at 320562369db2SQuentin Monnet * *ancestor_level* zero and each step down the hierarchy 320662369db2SQuentin Monnet * increments the level. If *ancestor_level* == level of cgroup 320762369db2SQuentin Monnet * associated with *skb*, then return value will be same as that 320862369db2SQuentin Monnet * of **bpf_skb_cgroup_id**\ (). 320962369db2SQuentin Monnet * 321062369db2SQuentin Monnet * The helper is useful to implement policies based on cgroups 321162369db2SQuentin Monnet * that are upper in hierarchy than immediate cgroup associated 321262369db2SQuentin Monnet * with *skb*. 321362369db2SQuentin Monnet * 321462369db2SQuentin Monnet * The format of returned id and helper limitations are same as in 321562369db2SQuentin Monnet * **bpf_skb_cgroup_id**\ (). 321662369db2SQuentin Monnet * Return 321762369db2SQuentin Monnet * The id is returned or 0 in case the id could not be retrieved. 321862369db2SQuentin Monnet * 3219f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 32206acc9b43SJoe Stringer * Description 32216acc9b43SJoe Stringer * Look for TCP socket matching *tuple*, optionally in a child 32226acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 322390b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 32246acc9b43SJoe Stringer * 32256acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 32266acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 32276acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 32286acc9b43SJoe Stringer * 32296acc9b43SJoe Stringer * *tuple_size* must be one of: 32306acc9b43SJoe Stringer * 32316acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 32326acc9b43SJoe Stringer * Look for an IPv4 socket. 32336acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 32346acc9b43SJoe Stringer * Look for an IPv6 socket. 32356acc9b43SJoe Stringer * 3236f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 3237bfdfa517SRandy Dunlap * socket lookup table in the netns associated with the *ctx* 3238f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 3239f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 3240f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 3241f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 3242f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 3243f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 32446acc9b43SJoe Stringer * 32456acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 32466acc9b43SJoe Stringer * be left at zero. 32476acc9b43SJoe Stringer * 32486acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 32496acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 32506acc9b43SJoe Stringer * Return 32510bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 32520bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 325380867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 325480867c5eSQuentin Monnet * tuple. 32556acc9b43SJoe Stringer * 3256f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 32576acc9b43SJoe Stringer * Description 32586acc9b43SJoe Stringer * Look for UDP socket matching *tuple*, optionally in a child 32596acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 326090b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 32616acc9b43SJoe Stringer * 32626acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 32636acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 32646acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 32656acc9b43SJoe Stringer * 32666acc9b43SJoe Stringer * *tuple_size* must be one of: 32676acc9b43SJoe Stringer * 32686acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 32696acc9b43SJoe Stringer * Look for an IPv4 socket. 32706acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 32716acc9b43SJoe Stringer * Look for an IPv6 socket. 32726acc9b43SJoe Stringer * 3273f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 3274bfdfa517SRandy Dunlap * socket lookup table in the netns associated with the *ctx* 3275f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 3276f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 3277f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 3278f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 3279f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 3280f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 32816acc9b43SJoe Stringer * 32826acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 32836acc9b43SJoe Stringer * be left at zero. 32846acc9b43SJoe Stringer * 32856acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 32866acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 32876acc9b43SJoe Stringer * Return 32880bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 32890bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 329080867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 329180867c5eSQuentin Monnet * tuple. 32926acc9b43SJoe Stringer * 3293a5fa25adSMartin KaFai Lau * long bpf_sk_release(void *sock) 32946acc9b43SJoe Stringer * Description 329590b1023fSQuentin Monnet * Release the reference held by *sock*. *sock* must be a 329690b1023fSQuentin Monnet * non-**NULL** pointer that was returned from 329790b1023fSQuentin Monnet * **bpf_sk_lookup_xxx**\ (). 329890b1023fSQuentin Monnet * Return 329990b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 330090b1023fSQuentin Monnet * 3301bdb7b79bSAndrii Nakryiko * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 330262369db2SQuentin Monnet * Description 330362369db2SQuentin Monnet * Push an element *value* in *map*. *flags* is one of: 330462369db2SQuentin Monnet * 330562369db2SQuentin Monnet * **BPF_EXIST** 330662369db2SQuentin Monnet * If the queue/stack is full, the oldest element is 330762369db2SQuentin Monnet * removed to make room for this. 330862369db2SQuentin Monnet * Return 330962369db2SQuentin Monnet * 0 on success, or a negative error in case of failure. 331062369db2SQuentin Monnet * 3311bdb7b79bSAndrii Nakryiko * long bpf_map_pop_elem(struct bpf_map *map, void *value) 331290b1023fSQuentin Monnet * Description 331390b1023fSQuentin Monnet * Pop an element from *map*. 331490b1023fSQuentin Monnet * Return 331590b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 331690b1023fSQuentin Monnet * 3317bdb7b79bSAndrii Nakryiko * long bpf_map_peek_elem(struct bpf_map *map, void *value) 331890b1023fSQuentin Monnet * Description 331990b1023fSQuentin Monnet * Get an element from *map* without removing it. 33206acc9b43SJoe Stringer * Return 33216acc9b43SJoe Stringer * 0 on success, or a negative error in case of failure. 33226fff607eSJohn Fastabend * 3323bdb7b79bSAndrii Nakryiko * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 33246fff607eSJohn Fastabend * Description 332590b1023fSQuentin Monnet * For socket policies, insert *len* bytes into *msg* at offset 33266fff607eSJohn Fastabend * *start*. 33276fff607eSJohn Fastabend * 33286fff607eSJohn Fastabend * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 332990b1023fSQuentin Monnet * *msg* it may want to insert metadata or options into the *msg*. 33306fff607eSJohn Fastabend * This can later be read and used by any of the lower layer BPF 33316fff607eSJohn Fastabend * hooks. 33326fff607eSJohn Fastabend * 33336fff607eSJohn Fastabend * This helper may fail if under memory pressure (a malloc 33346fff607eSJohn Fastabend * fails) in these cases BPF programs will get an appropriate 33356fff607eSJohn Fastabend * error and BPF programs will need to handle them. 33366fff607eSJohn Fastabend * Return 33376fff607eSJohn Fastabend * 0 on success, or a negative error in case of failure. 33387246d8edSJohn Fastabend * 3339bdb7b79bSAndrii Nakryiko * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 33407246d8edSJohn Fastabend * Description 33415f0e5412SAndrii Nakryiko * Will remove *len* bytes from a *msg* starting at byte *start*. 33427246d8edSJohn Fastabend * This may result in **ENOMEM** errors under certain situations if 33437246d8edSJohn Fastabend * an allocation and copy are required due to a full ring buffer. 33447246d8edSJohn Fastabend * However, the helper will try to avoid doing the allocation 33457246d8edSJohn Fastabend * if possible. Other errors can occur if input parameters are 334690b1023fSQuentin Monnet * invalid either due to *start* byte not being valid part of *msg* 33477246d8edSJohn Fastabend * payload and/or *pop* value being to large. 33487246d8edSJohn Fastabend * Return 334990b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 335001d3240aSSean Young * 3351bdb7b79bSAndrii Nakryiko * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 335201d3240aSSean Young * Description 335301d3240aSSean Young * This helper is used in programs implementing IR decoding, to 335401d3240aSSean Young * report a successfully decoded pointer movement. 335501d3240aSSean Young * 335601d3240aSSean Young * The *ctx* should point to the lirc sample as passed into 335701d3240aSSean Young * the program. 335801d3240aSSean Young * 335901d3240aSSean Young * This helper is only available is the kernel was compiled with 336001d3240aSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 336101d3240aSSean Young * "**y**". 336201d3240aSSean Young * Return 336301d3240aSSean Young * 0 336446f8bc92SMartin KaFai Lau * 3365bdb7b79bSAndrii Nakryiko * long bpf_spin_lock(struct bpf_spin_lock *lock) 33660eb09785SQuentin Monnet * Description 33670eb09785SQuentin Monnet * Acquire a spinlock represented by the pointer *lock*, which is 33680eb09785SQuentin Monnet * stored as part of a value of a map. Taking the lock allows to 33690eb09785SQuentin Monnet * safely update the rest of the fields in that value. The 33700eb09785SQuentin Monnet * spinlock can (and must) later be released with a call to 33710eb09785SQuentin Monnet * **bpf_spin_unlock**\ (\ *lock*\ ). 33720eb09785SQuentin Monnet * 33730eb09785SQuentin Monnet * Spinlocks in BPF programs come with a number of restrictions 33740eb09785SQuentin Monnet * and constraints: 33750eb09785SQuentin Monnet * 33760eb09785SQuentin Monnet * * **bpf_spin_lock** objects are only allowed inside maps of 33770eb09785SQuentin Monnet * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 33780eb09785SQuentin Monnet * list could be extended in the future). 33790eb09785SQuentin Monnet * * BTF description of the map is mandatory. 33800eb09785SQuentin Monnet * * The BPF program can take ONE lock at a time, since taking two 33810eb09785SQuentin Monnet * or more could cause dead locks. 33820eb09785SQuentin Monnet * * Only one **struct bpf_spin_lock** is allowed per map element. 33830eb09785SQuentin Monnet * * When the lock is taken, calls (either BPF to BPF or helpers) 33840eb09785SQuentin Monnet * are not allowed. 33850eb09785SQuentin Monnet * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 33860eb09785SQuentin Monnet * allowed inside a spinlock-ed region. 33870eb09785SQuentin Monnet * * The BPF program MUST call **bpf_spin_unlock**\ () to release 33880eb09785SQuentin Monnet * the lock, on all execution paths, before it returns. 33890eb09785SQuentin Monnet * * The BPF program can access **struct bpf_spin_lock** only via 33900eb09785SQuentin Monnet * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 33910eb09785SQuentin Monnet * helpers. Loading or storing data into the **struct 33920eb09785SQuentin Monnet * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 33930eb09785SQuentin Monnet * * To use the **bpf_spin_lock**\ () helper, the BTF description 33940eb09785SQuentin Monnet * of the map value must be a struct and have **struct 33950eb09785SQuentin Monnet * bpf_spin_lock** *anyname*\ **;** field at the top level. 33960eb09785SQuentin Monnet * Nested lock inside another struct is not allowed. 33970eb09785SQuentin Monnet * * The **struct bpf_spin_lock** *lock* field in a map value must 33980eb09785SQuentin Monnet * be aligned on a multiple of 4 bytes in that value. 33990eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 34000eb09785SQuentin Monnet * the **bpf_spin_lock** field to user space. 34010eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 34020eb09785SQuentin Monnet * a BPF program, do not update the **bpf_spin_lock** field. 34030eb09785SQuentin Monnet * * **bpf_spin_lock** cannot be on the stack or inside a 34040eb09785SQuentin Monnet * networking packet (it can only be inside of a map values). 34050eb09785SQuentin Monnet * * **bpf_spin_lock** is available to root only. 34060eb09785SQuentin Monnet * * Tracing programs and socket filter programs cannot use 34070eb09785SQuentin Monnet * **bpf_spin_lock**\ () due to insufficient preemption checks 34080eb09785SQuentin Monnet * (but this may change in the future). 34090eb09785SQuentin Monnet * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 34100eb09785SQuentin Monnet * Return 34110eb09785SQuentin Monnet * 0 34120eb09785SQuentin Monnet * 3413bdb7b79bSAndrii Nakryiko * long bpf_spin_unlock(struct bpf_spin_lock *lock) 34140eb09785SQuentin Monnet * Description 34150eb09785SQuentin Monnet * Release the *lock* previously locked by a call to 34160eb09785SQuentin Monnet * **bpf_spin_lock**\ (\ *lock*\ ). 34170eb09785SQuentin Monnet * Return 34180eb09785SQuentin Monnet * 0 34190eb09785SQuentin Monnet * 342046f8bc92SMartin KaFai Lau * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 342146f8bc92SMartin KaFai Lau * Description 342246f8bc92SMartin KaFai Lau * This helper gets a **struct bpf_sock** pointer such 342362369db2SQuentin Monnet * that all the fields in this **bpf_sock** can be accessed. 342446f8bc92SMartin KaFai Lau * Return 342562369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 342646f8bc92SMartin KaFai Lau * case of failure. 3427655a51e5SMartin KaFai Lau * 3428655a51e5SMartin KaFai Lau * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3429655a51e5SMartin KaFai Lau * Description 3430655a51e5SMartin KaFai Lau * This helper gets a **struct bpf_tcp_sock** pointer from a 3431655a51e5SMartin KaFai Lau * **struct bpf_sock** pointer. 3432655a51e5SMartin KaFai Lau * Return 343362369db2SQuentin Monnet * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3434655a51e5SMartin KaFai Lau * case of failure. 3435f7c917baSbrakmo * 3436bdb7b79bSAndrii Nakryiko * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3437f7c917baSbrakmo * Description 343862369db2SQuentin Monnet * Set ECN (Explicit Congestion Notification) field of IP header 343962369db2SQuentin Monnet * to **CE** (Congestion Encountered) if current value is **ECT** 344062369db2SQuentin Monnet * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 344162369db2SQuentin Monnet * and IPv4. 3442f7c917baSbrakmo * Return 344362369db2SQuentin Monnet * 1 if the **CE** flag is set (either by the current helper call 344462369db2SQuentin Monnet * or because it was already present), 0 if it is not set. 3445dbafd7ddSMartin KaFai Lau * 3446dbafd7ddSMartin KaFai Lau * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3447dbafd7ddSMartin KaFai Lau * Description 344862369db2SQuentin Monnet * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 344962369db2SQuentin Monnet * **bpf_sk_release**\ () is unnecessary and not allowed. 3450dbafd7ddSMartin KaFai Lau * Return 345162369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 3452dbafd7ddSMartin KaFai Lau * case of failure. 3453edbf8c01SLorenz Bauer * 3454edbf8c01SLorenz Bauer * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3455edbf8c01SLorenz Bauer * Description 3456edbf8c01SLorenz Bauer * Look for TCP socket matching *tuple*, optionally in a child 3457edbf8c01SLorenz Bauer * network namespace *netns*. The return value must be checked, 3458edbf8c01SLorenz Bauer * and if non-**NULL**, released via **bpf_sk_release**\ (). 3459edbf8c01SLorenz Bauer * 346080867c5eSQuentin Monnet * This function is identical to **bpf_sk_lookup_tcp**\ (), except 346180867c5eSQuentin Monnet * that it also returns timewait or request sockets. Use 346280867c5eSQuentin Monnet * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 346380867c5eSQuentin Monnet * full structure. 3464edbf8c01SLorenz Bauer * 3465edbf8c01SLorenz Bauer * This helper is available only if the kernel was compiled with 3466edbf8c01SLorenz Bauer * **CONFIG_NET** configuration option. 3467edbf8c01SLorenz Bauer * Return 3468edbf8c01SLorenz Bauer * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3469edbf8c01SLorenz Bauer * For sockets with reuseport option, the **struct bpf_sock** 347080867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 347180867c5eSQuentin Monnet * tuple. 347239904084SLorenz Bauer * 3473c0df236eSMartin KaFai Lau * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 347439904084SLorenz Bauer * Description 347580867c5eSQuentin Monnet * Check whether *iph* and *th* contain a valid SYN cookie ACK for 347680867c5eSQuentin Monnet * the listening socket in *sk*. 347739904084SLorenz Bauer * 347880867c5eSQuentin Monnet * *iph* points to the start of the IPv4 or IPv6 header, while 347980867c5eSQuentin Monnet * *iph_len* contains **sizeof**\ (**struct iphdr**) or 348080867c5eSQuentin Monnet * **sizeof**\ (**struct ip6hdr**). 348139904084SLorenz Bauer * 348280867c5eSQuentin Monnet * *th* points to the start of the TCP header, while *th_len* 348380867c5eSQuentin Monnet * contains **sizeof**\ (**struct tcphdr**). 348439904084SLorenz Bauer * Return 348580867c5eSQuentin Monnet * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 348680867c5eSQuentin Monnet * error otherwise. 3487808649fbSAndrey Ignatov * 3488bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3489808649fbSAndrey Ignatov * Description 3490808649fbSAndrey Ignatov * Get name of sysctl in /proc/sys/ and copy it into provided by 3491808649fbSAndrey Ignatov * program buffer *buf* of size *buf_len*. 3492808649fbSAndrey Ignatov * 3493808649fbSAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 3494808649fbSAndrey Ignatov * 3495808649fbSAndrey Ignatov * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3496808649fbSAndrey Ignatov * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3497808649fbSAndrey Ignatov * only (e.g. "tcp_mem"). 3498808649fbSAndrey Ignatov * Return 3499808649fbSAndrey Ignatov * Number of character copied (not including the trailing NUL). 3500808649fbSAndrey Ignatov * 3501808649fbSAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3502808649fbSAndrey Ignatov * truncated name in this case). 35031d11b301SAndrey Ignatov * 3504bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 35051d11b301SAndrey Ignatov * Description 35061d11b301SAndrey Ignatov * Get current value of sysctl as it is presented in /proc/sys 35071d11b301SAndrey Ignatov * (incl. newline, etc), and copy it as a string into provided 35081d11b301SAndrey Ignatov * by program buffer *buf* of size *buf_len*. 35091d11b301SAndrey Ignatov * 35101d11b301SAndrey Ignatov * The whole value is copied, no matter what file position user 35111d11b301SAndrey Ignatov * space issued e.g. sys_read at. 35121d11b301SAndrey Ignatov * 35131d11b301SAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 35141d11b301SAndrey Ignatov * Return 35151d11b301SAndrey Ignatov * Number of character copied (not including the trailing NUL). 35161d11b301SAndrey Ignatov * 35171d11b301SAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 35181d11b301SAndrey Ignatov * truncated name in this case). 35191d11b301SAndrey Ignatov * 35201d11b301SAndrey Ignatov * **-EINVAL** if current value was unavailable, e.g. because 35211d11b301SAndrey Ignatov * sysctl is uninitialized and read returns -EIO for it. 35224e63acdfSAndrey Ignatov * 3523bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 35244e63acdfSAndrey Ignatov * Description 35254e63acdfSAndrey Ignatov * Get new value being written by user space to sysctl (before 35264e63acdfSAndrey Ignatov * the actual write happens) and copy it as a string into 35274e63acdfSAndrey Ignatov * provided by program buffer *buf* of size *buf_len*. 35284e63acdfSAndrey Ignatov * 35294e63acdfSAndrey Ignatov * User space may write new value at file position > 0. 35304e63acdfSAndrey Ignatov * 35314e63acdfSAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 35324e63acdfSAndrey Ignatov * Return 35334e63acdfSAndrey Ignatov * Number of character copied (not including the trailing NUL). 35344e63acdfSAndrey Ignatov * 35354e63acdfSAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 35364e63acdfSAndrey Ignatov * truncated name in this case). 35374e63acdfSAndrey Ignatov * 35384e63acdfSAndrey Ignatov * **-EINVAL** if sysctl is being read. 35394e63acdfSAndrey Ignatov * 3540bdb7b79bSAndrii Nakryiko * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 35414e63acdfSAndrey Ignatov * Description 35424e63acdfSAndrey Ignatov * Override new value being written by user space to sysctl with 35434e63acdfSAndrey Ignatov * value provided by program in buffer *buf* of size *buf_len*. 35444e63acdfSAndrey Ignatov * 35454e63acdfSAndrey Ignatov * *buf* should contain a string in same form as provided by user 35464e63acdfSAndrey Ignatov * space on sysctl write. 35474e63acdfSAndrey Ignatov * 35484e63acdfSAndrey Ignatov * User space may write new value at file position > 0. To override 35494e63acdfSAndrey Ignatov * the whole sysctl value file position should be set to zero. 35504e63acdfSAndrey Ignatov * Return 35514e63acdfSAndrey Ignatov * 0 on success. 35524e63acdfSAndrey Ignatov * 35534e63acdfSAndrey Ignatov * **-E2BIG** if the *buf_len* is too big. 35544e63acdfSAndrey Ignatov * 35554e63acdfSAndrey Ignatov * **-EINVAL** if sysctl is being read. 3556d7a4cb9bSAndrey Ignatov * 3557bdb7b79bSAndrii Nakryiko * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3558d7a4cb9bSAndrey Ignatov * Description 3559d7a4cb9bSAndrey Ignatov * Convert the initial part of the string from buffer *buf* of 3560d7a4cb9bSAndrey Ignatov * size *buf_len* to a long integer according to the given base 3561d7a4cb9bSAndrey Ignatov * and save the result in *res*. 3562d7a4cb9bSAndrey Ignatov * 3563d7a4cb9bSAndrey Ignatov * The string may begin with an arbitrary amount of white space 356480867c5eSQuentin Monnet * (as determined by **isspace**\ (3)) followed by a single 356580867c5eSQuentin Monnet * optional '**-**' sign. 3566d7a4cb9bSAndrey Ignatov * 3567d7a4cb9bSAndrey Ignatov * Five least significant bits of *flags* encode base, other bits 3568d7a4cb9bSAndrey Ignatov * are currently unused. 3569d7a4cb9bSAndrey Ignatov * 3570d7a4cb9bSAndrey Ignatov * Base must be either 8, 10, 16 or 0 to detect it automatically 357180867c5eSQuentin Monnet * similar to user space **strtol**\ (3). 3572d7a4cb9bSAndrey Ignatov * Return 3573d7a4cb9bSAndrey Ignatov * Number of characters consumed on success. Must be positive but 357480867c5eSQuentin Monnet * no more than *buf_len*. 3575d7a4cb9bSAndrey Ignatov * 3576d7a4cb9bSAndrey Ignatov * **-EINVAL** if no valid digits were found or unsupported base 3577d7a4cb9bSAndrey Ignatov * was provided. 3578d7a4cb9bSAndrey Ignatov * 3579d7a4cb9bSAndrey Ignatov * **-ERANGE** if resulting value was out of range. 3580d7a4cb9bSAndrey Ignatov * 3581bdb7b79bSAndrii Nakryiko * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3582d7a4cb9bSAndrey Ignatov * Description 3583d7a4cb9bSAndrey Ignatov * Convert the initial part of the string from buffer *buf* of 3584d7a4cb9bSAndrey Ignatov * size *buf_len* to an unsigned long integer according to the 3585d7a4cb9bSAndrey Ignatov * given base and save the result in *res*. 3586d7a4cb9bSAndrey Ignatov * 3587d7a4cb9bSAndrey Ignatov * The string may begin with an arbitrary amount of white space 358880867c5eSQuentin Monnet * (as determined by **isspace**\ (3)). 3589d7a4cb9bSAndrey Ignatov * 3590d7a4cb9bSAndrey Ignatov * Five least significant bits of *flags* encode base, other bits 3591d7a4cb9bSAndrey Ignatov * are currently unused. 3592d7a4cb9bSAndrey Ignatov * 3593d7a4cb9bSAndrey Ignatov * Base must be either 8, 10, 16 or 0 to detect it automatically 359480867c5eSQuentin Monnet * similar to user space **strtoul**\ (3). 3595d7a4cb9bSAndrey Ignatov * Return 3596d7a4cb9bSAndrey Ignatov * Number of characters consumed on success. Must be positive but 359780867c5eSQuentin Monnet * no more than *buf_len*. 3598d7a4cb9bSAndrey Ignatov * 3599d7a4cb9bSAndrey Ignatov * **-EINVAL** if no valid digits were found or unsupported base 3600d7a4cb9bSAndrey Ignatov * was provided. 3601d7a4cb9bSAndrey Ignatov * 3602d7a4cb9bSAndrey Ignatov * **-ERANGE** if resulting value was out of range. 36036ac99e8fSMartin KaFai Lau * 360430897832SKP Singh * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 36056ac99e8fSMartin KaFai Lau * Description 360680867c5eSQuentin Monnet * Get a bpf-local-storage from a *sk*. 36076ac99e8fSMartin KaFai Lau * 36086ac99e8fSMartin KaFai Lau * Logically, it could be thought of getting the value from 36096ac99e8fSMartin KaFai Lau * a *map* with *sk* as the **key**. From this 36106ac99e8fSMartin KaFai Lau * perspective, the usage is not much different from 361180867c5eSQuentin Monnet * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 361280867c5eSQuentin Monnet * helper enforces the key must be a full socket and the map must 361380867c5eSQuentin Monnet * be a **BPF_MAP_TYPE_SK_STORAGE** also. 36146ac99e8fSMartin KaFai Lau * 36156ac99e8fSMartin KaFai Lau * Underneath, the value is stored locally at *sk* instead of 361680867c5eSQuentin Monnet * the *map*. The *map* is used as the bpf-local-storage 361780867c5eSQuentin Monnet * "type". The bpf-local-storage "type" (i.e. the *map*) is 361880867c5eSQuentin Monnet * searched against all bpf-local-storages residing at *sk*. 36196ac99e8fSMartin KaFai Lau * 362030897832SKP Singh * *sk* is a kernel **struct sock** pointer for LSM program. 362130897832SKP Singh * *sk* is a **struct bpf_sock** pointer for other program types. 362230897832SKP Singh * 362380867c5eSQuentin Monnet * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 36246ac99e8fSMartin KaFai Lau * used such that a new bpf-local-storage will be 36256ac99e8fSMartin KaFai Lau * created if one does not exist. *value* can be used 362680867c5eSQuentin Monnet * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 36276ac99e8fSMartin KaFai Lau * the initial value of a bpf-local-storage. If *value* is 362880867c5eSQuentin Monnet * **NULL**, the new bpf-local-storage will be zero initialized. 36296ac99e8fSMartin KaFai Lau * Return 36306ac99e8fSMartin KaFai Lau * A bpf-local-storage pointer is returned on success. 36316ac99e8fSMartin KaFai Lau * 36326ac99e8fSMartin KaFai Lau * **NULL** if not found or there was an error in adding 36336ac99e8fSMartin KaFai Lau * a new bpf-local-storage. 36346ac99e8fSMartin KaFai Lau * 363530897832SKP Singh * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 36366ac99e8fSMartin KaFai Lau * Description 363780867c5eSQuentin Monnet * Delete a bpf-local-storage from a *sk*. 36386ac99e8fSMartin KaFai Lau * Return 36396ac99e8fSMartin KaFai Lau * 0 on success. 36406ac99e8fSMartin KaFai Lau * 36416ac99e8fSMartin KaFai Lau * **-ENOENT** if the bpf-local-storage cannot be found. 3642592a3498SMartin KaFai Lau * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 36438b401f9eSYonghong Song * 3644bdb7b79bSAndrii Nakryiko * long bpf_send_signal(u32 sig) 36458b401f9eSYonghong Song * Description 36468482941fSYonghong Song * Send signal *sig* to the process of the current task. 36478482941fSYonghong Song * The signal may be delivered to any of this process's threads. 36488b401f9eSYonghong Song * Return 36498b401f9eSYonghong Song * 0 on success or successfully queued. 36508b401f9eSYonghong Song * 36518b401f9eSYonghong Song * **-EBUSY** if work queue under nmi is full. 36528b401f9eSYonghong Song * 36538b401f9eSYonghong Song * **-EINVAL** if *sig* is invalid. 36548b401f9eSYonghong Song * 36558b401f9eSYonghong Song * **-EPERM** if no permission to send the *sig*. 36568b401f9eSYonghong Song * 36578b401f9eSYonghong Song * **-EAGAIN** if bpf program can try again. 365870d66244SPetar Penkov * 3659c0df236eSMartin KaFai Lau * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 366070d66244SPetar Penkov * Description 366170d66244SPetar Penkov * Try to issue a SYN cookie for the packet with corresponding 366270d66244SPetar Penkov * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 366370d66244SPetar Penkov * 366470d66244SPetar Penkov * *iph* points to the start of the IPv4 or IPv6 header, while 366570d66244SPetar Penkov * *iph_len* contains **sizeof**\ (**struct iphdr**) or 366670d66244SPetar Penkov * **sizeof**\ (**struct ip6hdr**). 366770d66244SPetar Penkov * 366870d66244SPetar Penkov * *th* points to the start of the TCP header, while *th_len* 366970d66244SPetar Penkov * contains the length of the TCP header. 367070d66244SPetar Penkov * Return 367170d66244SPetar Penkov * On success, lower 32 bits hold the generated SYN cookie in 367270d66244SPetar Penkov * followed by 16 bits which hold the MSS value for that cookie, 367370d66244SPetar Penkov * and the top 16 bits are unused. 367470d66244SPetar Penkov * 367570d66244SPetar Penkov * On failure, the returned value is one of the following: 367670d66244SPetar Penkov * 367770d66244SPetar Penkov * **-EINVAL** SYN cookie cannot be issued due to error 367870d66244SPetar Penkov * 367970d66244SPetar Penkov * **-ENOENT** SYN cookie should not be issued (no SYN flood) 368070d66244SPetar Penkov * 368170d66244SPetar Penkov * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 368270d66244SPetar Penkov * 368370d66244SPetar Penkov * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 3684a7658e1aSAlexei Starovoitov * 3685bdb7b79bSAndrii Nakryiko * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3686a7658e1aSAlexei Starovoitov * Description 3687a7658e1aSAlexei Starovoitov * Write raw *data* blob into a special BPF perf event held by 3688a7658e1aSAlexei Starovoitov * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3689a7658e1aSAlexei Starovoitov * event must have the following attributes: **PERF_SAMPLE_RAW** 3690a7658e1aSAlexei Starovoitov * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3691a7658e1aSAlexei Starovoitov * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3692a7658e1aSAlexei Starovoitov * 3693a7658e1aSAlexei Starovoitov * The *flags* are used to indicate the index in *map* for which 3694a7658e1aSAlexei Starovoitov * the value must be put, masked with **BPF_F_INDEX_MASK**. 3695a7658e1aSAlexei Starovoitov * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3696a7658e1aSAlexei Starovoitov * to indicate that the index of the current CPU core should be 3697a7658e1aSAlexei Starovoitov * used. 3698a7658e1aSAlexei Starovoitov * 3699a7658e1aSAlexei Starovoitov * The value to write, of *size*, is passed through eBPF stack and 3700a7658e1aSAlexei Starovoitov * pointed by *data*. 3701a7658e1aSAlexei Starovoitov * 3702a7658e1aSAlexei Starovoitov * *ctx* is a pointer to in-kernel struct sk_buff. 3703a7658e1aSAlexei Starovoitov * 3704a7658e1aSAlexei Starovoitov * This helper is similar to **bpf_perf_event_output**\ () but 3705a7658e1aSAlexei Starovoitov * restricted to raw_tracepoint bpf programs. 3706a7658e1aSAlexei Starovoitov * Return 3707a7658e1aSAlexei Starovoitov * 0 on success, or a negative error in case of failure. 37086ae08ae3SDaniel Borkmann * 3709bdb7b79bSAndrii Nakryiko * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 37106ae08ae3SDaniel Borkmann * Description 37116ae08ae3SDaniel Borkmann * Safely attempt to read *size* bytes from user space address 37126ae08ae3SDaniel Borkmann * *unsafe_ptr* and store the data in *dst*. 37136ae08ae3SDaniel Borkmann * Return 37146ae08ae3SDaniel Borkmann * 0 on success, or a negative error in case of failure. 37156ae08ae3SDaniel Borkmann * 3716bdb7b79bSAndrii Nakryiko * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 37176ae08ae3SDaniel Borkmann * Description 37186ae08ae3SDaniel Borkmann * Safely attempt to read *size* bytes from kernel space address 37196ae08ae3SDaniel Borkmann * *unsafe_ptr* and store the data in *dst*. 37206ae08ae3SDaniel Borkmann * Return 37216ae08ae3SDaniel Borkmann * 0 on success, or a negative error in case of failure. 37226ae08ae3SDaniel Borkmann * 3723bdb7b79bSAndrii Nakryiko * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 37246ae08ae3SDaniel Borkmann * Description 37256ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe user address 37266ae08ae3SDaniel Borkmann * *unsafe_ptr* to *dst*. The *size* should include the 37276ae08ae3SDaniel Borkmann * terminating NUL byte. In case the string length is smaller than 37286ae08ae3SDaniel Borkmann * *size*, the target is not padded with further NUL bytes. If the 37296ae08ae3SDaniel Borkmann * string length is larger than *size*, just *size*-1 bytes are 37306ae08ae3SDaniel Borkmann * copied and the last byte is set to NUL. 37316ae08ae3SDaniel Borkmann * 3732c6458e72SBrendan Jackman * On success, returns the number of bytes that were written, 3733c6458e72SBrendan Jackman * including the terminal NUL. This makes this helper useful in 3734c6458e72SBrendan Jackman * tracing programs for reading strings, and more importantly to 3735c6458e72SBrendan Jackman * get its length at runtime. See the following snippet: 37366ae08ae3SDaniel Borkmann * 37376ae08ae3SDaniel Borkmann * :: 37386ae08ae3SDaniel Borkmann * 37396ae08ae3SDaniel Borkmann * SEC("kprobe/sys_open") 37406ae08ae3SDaniel Borkmann * void bpf_sys_open(struct pt_regs *ctx) 37416ae08ae3SDaniel Borkmann * { 37426ae08ae3SDaniel Borkmann * char buf[PATHLEN]; // PATHLEN is defined to 256 37436ae08ae3SDaniel Borkmann * int res = bpf_probe_read_user_str(buf, sizeof(buf), 37446ae08ae3SDaniel Borkmann * ctx->di); 37456ae08ae3SDaniel Borkmann * 37466ae08ae3SDaniel Borkmann * // Consume buf, for example push it to 37476ae08ae3SDaniel Borkmann * // userspace via bpf_perf_event_output(); we 37486ae08ae3SDaniel Borkmann * // can use res (the string length) as event 37496ae08ae3SDaniel Borkmann * // size, after checking its boundaries. 37506ae08ae3SDaniel Borkmann * } 37516ae08ae3SDaniel Borkmann * 3752ab8d7809SQuentin Monnet * In comparison, using **bpf_probe_read_user**\ () helper here 37536ae08ae3SDaniel Borkmann * instead to read the string would require to estimate the length 37546ae08ae3SDaniel Borkmann * at compile time, and would often result in copying more memory 37556ae08ae3SDaniel Borkmann * than necessary. 37566ae08ae3SDaniel Borkmann * 37576ae08ae3SDaniel Borkmann * Another useful use case is when parsing individual process 37586ae08ae3SDaniel Borkmann * arguments or individual environment variables navigating 37596ae08ae3SDaniel Borkmann * *current*\ **->mm->arg_start** and *current*\ 37606ae08ae3SDaniel Borkmann * **->mm->env_start**: using this helper and the return value, 37616ae08ae3SDaniel Borkmann * one can quickly iterate at the right offset of the memory area. 37626ae08ae3SDaniel Borkmann * Return 3763c6458e72SBrendan Jackman * On success, the strictly positive length of the output string, 37646ae08ae3SDaniel Borkmann * including the trailing NUL character. On error, a negative 37656ae08ae3SDaniel Borkmann * value. 37666ae08ae3SDaniel Borkmann * 3767bdb7b79bSAndrii Nakryiko * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 37686ae08ae3SDaniel Borkmann * Description 37696ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 3770ab8d7809SQuentin Monnet * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 37716ae08ae3SDaniel Borkmann * Return 37726ae08ae3SDaniel Borkmann * On success, the strictly positive length of the string, including 37736ae08ae3SDaniel Borkmann * the trailing NUL character. On error, a negative value. 3774206057feSMartin KaFai Lau * 3775bdb7b79bSAndrii Nakryiko * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 3776206057feSMartin KaFai Lau * Description 3777ab8d7809SQuentin Monnet * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 3778206057feSMartin KaFai Lau * *rcv_nxt* is the ack_seq to be sent out. 3779206057feSMartin KaFai Lau * Return 3780206057feSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 3781206057feSMartin KaFai Lau * 3782bdb7b79bSAndrii Nakryiko * long bpf_send_signal_thread(u32 sig) 37838482941fSYonghong Song * Description 37848482941fSYonghong Song * Send signal *sig* to the thread corresponding to the current task. 37858482941fSYonghong Song * Return 37868482941fSYonghong Song * 0 on success or successfully queued. 37878482941fSYonghong Song * 37888482941fSYonghong Song * **-EBUSY** if work queue under nmi is full. 37898482941fSYonghong Song * 37908482941fSYonghong Song * **-EINVAL** if *sig* is invalid. 37918482941fSYonghong Song * 37928482941fSYonghong Song * **-EPERM** if no permission to send the *sig*. 37938482941fSYonghong Song * 37948482941fSYonghong Song * **-EAGAIN** if bpf program can try again. 37955576b991SMartin KaFai Lau * 37965576b991SMartin KaFai Lau * u64 bpf_jiffies64(void) 37975576b991SMartin KaFai Lau * Description 37985576b991SMartin KaFai Lau * Obtain the 64bit jiffies 37995576b991SMartin KaFai Lau * Return 38005576b991SMartin KaFai Lau * The 64 bit jiffies 3801fff7b643SDaniel Xu * 3802bdb7b79bSAndrii Nakryiko * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3803fff7b643SDaniel Xu * Description 3804fff7b643SDaniel Xu * For an eBPF program attached to a perf event, retrieve the 3805ab8d7809SQuentin Monnet * branch records (**struct perf_branch_entry**) associated to *ctx* 3806fff7b643SDaniel Xu * and store it in the buffer pointed by *buf* up to size 3807fff7b643SDaniel Xu * *size* bytes. 3808fff7b643SDaniel Xu * Return 3809fff7b643SDaniel Xu * On success, number of bytes written to *buf*. On error, a 3810fff7b643SDaniel Xu * negative value. 3811fff7b643SDaniel Xu * 3812fff7b643SDaniel Xu * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3813fff7b643SDaniel Xu * instead return the number of bytes required to store all the 3814fff7b643SDaniel Xu * branch entries. If this flag is set, *buf* may be NULL. 3815fff7b643SDaniel Xu * 3816fff7b643SDaniel Xu * **-EINVAL** if arguments invalid or **size** not a multiple 3817ab8d7809SQuentin Monnet * of **sizeof**\ (**struct perf_branch_entry**\ ). 3818fff7b643SDaniel Xu * 3819fff7b643SDaniel Xu * **-ENOENT** if architecture does not support branch records. 3820b4490c5cSCarlos Neira * 3821bdb7b79bSAndrii Nakryiko * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3822b4490c5cSCarlos Neira * Description 3823b4490c5cSCarlos Neira * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3824b4490c5cSCarlos Neira * *namespace* will be returned in *nsdata*. 3825ab8d7809SQuentin Monnet * Return 3826ab8d7809SQuentin Monnet * 0 on success, or one of the following in case of failure: 3827b4490c5cSCarlos Neira * 3828b4490c5cSCarlos Neira * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3829b4490c5cSCarlos Neira * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3830b4490c5cSCarlos Neira * 3831b4490c5cSCarlos Neira * **-ENOENT** if pidns does not exists for the current task. 3832b4490c5cSCarlos Neira * 3833bdb7b79bSAndrii Nakryiko * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3834d831ee84SEelco Chaudron * Description 3835d831ee84SEelco Chaudron * Write raw *data* blob into a special BPF perf event held by 3836d831ee84SEelco Chaudron * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3837d831ee84SEelco Chaudron * event must have the following attributes: **PERF_SAMPLE_RAW** 3838d831ee84SEelco Chaudron * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3839d831ee84SEelco Chaudron * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3840d831ee84SEelco Chaudron * 3841d831ee84SEelco Chaudron * The *flags* are used to indicate the index in *map* for which 3842d831ee84SEelco Chaudron * the value must be put, masked with **BPF_F_INDEX_MASK**. 3843d831ee84SEelco Chaudron * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3844d831ee84SEelco Chaudron * to indicate that the index of the current CPU core should be 3845d831ee84SEelco Chaudron * used. 3846d831ee84SEelco Chaudron * 3847d831ee84SEelco Chaudron * The value to write, of *size*, is passed through eBPF stack and 3848d831ee84SEelco Chaudron * pointed by *data*. 3849d831ee84SEelco Chaudron * 3850d831ee84SEelco Chaudron * *ctx* is a pointer to in-kernel struct xdp_buff. 3851d831ee84SEelco Chaudron * 3852d831ee84SEelco Chaudron * This helper is similar to **bpf_perf_eventoutput**\ () but 3853d831ee84SEelco Chaudron * restricted to raw_tracepoint bpf programs. 3854d831ee84SEelco Chaudron * Return 3855d831ee84SEelco Chaudron * 0 on success, or a negative error in case of failure. 3856f318903cSDaniel Borkmann * 3857f318903cSDaniel Borkmann * u64 bpf_get_netns_cookie(void *ctx) 3858f318903cSDaniel Borkmann * Description 3859f318903cSDaniel Borkmann * Retrieve the cookie (generated by the kernel) of the network 3860f318903cSDaniel Borkmann * namespace the input *ctx* is associated with. The network 3861f318903cSDaniel Borkmann * namespace cookie remains stable for its lifetime and provides 3862f318903cSDaniel Borkmann * a global identifier that can be assumed unique. If *ctx* is 3863f318903cSDaniel Borkmann * NULL, then the helper returns the cookie for the initial 3864f318903cSDaniel Borkmann * network namespace. The cookie itself is very similar to that 3865ab8d7809SQuentin Monnet * of **bpf_get_socket_cookie**\ () helper, but for network 3866ab8d7809SQuentin Monnet * namespaces instead of sockets. 3867f318903cSDaniel Borkmann * Return 3868f318903cSDaniel Borkmann * A 8-byte long opaque number. 38690f09abd1SDaniel Borkmann * 38700f09abd1SDaniel Borkmann * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 38710f09abd1SDaniel Borkmann * Description 38720f09abd1SDaniel Borkmann * Return id of cgroup v2 that is ancestor of the cgroup associated 38730f09abd1SDaniel Borkmann * with the current task at the *ancestor_level*. The root cgroup 38740f09abd1SDaniel Borkmann * is at *ancestor_level* zero and each step down the hierarchy 38750f09abd1SDaniel Borkmann * increments the level. If *ancestor_level* == level of cgroup 38760f09abd1SDaniel Borkmann * associated with the current task, then return value will be the 38770f09abd1SDaniel Borkmann * same as that of **bpf_get_current_cgroup_id**\ (). 38780f09abd1SDaniel Borkmann * 38790f09abd1SDaniel Borkmann * The helper is useful to implement policies based on cgroups 38800f09abd1SDaniel Borkmann * that are upper in hierarchy than immediate cgroup associated 38810f09abd1SDaniel Borkmann * with the current task. 38820f09abd1SDaniel Borkmann * 38830f09abd1SDaniel Borkmann * The format of returned id and helper limitations are same as in 38840f09abd1SDaniel Borkmann * **bpf_get_current_cgroup_id**\ (). 38850f09abd1SDaniel Borkmann * Return 38860f09abd1SDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 3887cf7fbe66SJoe Stringer * 388827e5203bSMartin KaFai Lau * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 3889cf7fbe66SJoe Stringer * Description 3890e9ddbb77SJakub Sitnicki * Helper is overloaded depending on BPF program type. This 3891e9ddbb77SJakub Sitnicki * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 3892e9ddbb77SJakub Sitnicki * **BPF_PROG_TYPE_SCHED_ACT** programs. 3893e9ddbb77SJakub Sitnicki * 3894cf7fbe66SJoe Stringer * Assign the *sk* to the *skb*. When combined with appropriate 3895cf7fbe66SJoe Stringer * routing configuration to receive the packet towards the socket, 3896cf7fbe66SJoe Stringer * will cause *skb* to be delivered to the specified socket. 3897cf7fbe66SJoe Stringer * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3898cf7fbe66SJoe Stringer * **bpf_clone_redirect**\ () or other methods outside of BPF may 3899cf7fbe66SJoe Stringer * interfere with successful delivery to the socket. 3900cf7fbe66SJoe Stringer * 3901cf7fbe66SJoe Stringer * This operation is only valid from TC ingress path. 3902cf7fbe66SJoe Stringer * 3903cf7fbe66SJoe Stringer * The *flags* argument must be zero. 3904cf7fbe66SJoe Stringer * Return 3905ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 3906cf7fbe66SJoe Stringer * 3907ab8d7809SQuentin Monnet * **-EINVAL** if specified *flags* are not supported. 3908ab8d7809SQuentin Monnet * 3909ab8d7809SQuentin Monnet * **-ENOENT** if the socket is unavailable for assignment. 3910ab8d7809SQuentin Monnet * 3911ab8d7809SQuentin Monnet * **-ENETUNREACH** if the socket is unreachable (wrong netns). 3912ab8d7809SQuentin Monnet * 3913ab8d7809SQuentin Monnet * **-EOPNOTSUPP** if the operation is not supported, for example 3914ab8d7809SQuentin Monnet * a call from outside of TC ingress. 3915ab8d7809SQuentin Monnet * 3916ab8d7809SQuentin Monnet * **-ESOCKTNOSUPPORT** if the socket type is not supported 3917ab8d7809SQuentin Monnet * (reuseport). 391871d19214SMaciej Żenczykowski * 3919e9ddbb77SJakub Sitnicki * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 3920e9ddbb77SJakub Sitnicki * Description 3921e9ddbb77SJakub Sitnicki * Helper is overloaded depending on BPF program type. This 3922e9ddbb77SJakub Sitnicki * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 3923e9ddbb77SJakub Sitnicki * 3924e9ddbb77SJakub Sitnicki * Select the *sk* as a result of a socket lookup. 3925e9ddbb77SJakub Sitnicki * 3926e9ddbb77SJakub Sitnicki * For the operation to succeed passed socket must be compatible 3927e9ddbb77SJakub Sitnicki * with the packet description provided by the *ctx* object. 3928e9ddbb77SJakub Sitnicki * 3929e9ddbb77SJakub Sitnicki * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 3930e9ddbb77SJakub Sitnicki * be an exact match. While IP family (**AF_INET** or 3931e9ddbb77SJakub Sitnicki * **AF_INET6**) must be compatible, that is IPv6 sockets 3932e9ddbb77SJakub Sitnicki * that are not v6-only can be selected for IPv4 packets. 3933e9ddbb77SJakub Sitnicki * 3934e9ddbb77SJakub Sitnicki * Only TCP listeners and UDP unconnected sockets can be 3935e9ddbb77SJakub Sitnicki * selected. *sk* can also be NULL to reset any previous 3936e9ddbb77SJakub Sitnicki * selection. 3937e9ddbb77SJakub Sitnicki * 3938e9ddbb77SJakub Sitnicki * *flags* argument can combination of following values: 3939e9ddbb77SJakub Sitnicki * 3940e9ddbb77SJakub Sitnicki * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 3941e9ddbb77SJakub Sitnicki * socket selection, potentially done by a BPF program 3942e9ddbb77SJakub Sitnicki * that ran before us. 3943e9ddbb77SJakub Sitnicki * 3944e9ddbb77SJakub Sitnicki * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 3945e9ddbb77SJakub Sitnicki * load-balancing within reuseport group for the socket 3946e9ddbb77SJakub Sitnicki * being selected. 3947e9ddbb77SJakub Sitnicki * 3948e9ddbb77SJakub Sitnicki * On success *ctx->sk* will point to the selected socket. 3949e9ddbb77SJakub Sitnicki * 3950e9ddbb77SJakub Sitnicki * Return 3951e9ddbb77SJakub Sitnicki * 0 on success, or a negative errno in case of failure. 3952e9ddbb77SJakub Sitnicki * 3953e9ddbb77SJakub Sitnicki * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 3954e9ddbb77SJakub Sitnicki * not compatible with packet family (*ctx->family*). 3955e9ddbb77SJakub Sitnicki * 3956e9ddbb77SJakub Sitnicki * * **-EEXIST** if socket has been already selected, 3957e9ddbb77SJakub Sitnicki * potentially by another program, and 3958e9ddbb77SJakub Sitnicki * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 3959e9ddbb77SJakub Sitnicki * 3960e9ddbb77SJakub Sitnicki * * **-EINVAL** if unsupported flags were specified. 3961e9ddbb77SJakub Sitnicki * 3962e9ddbb77SJakub Sitnicki * * **-EPROTOTYPE** if socket L4 protocol 3963e9ddbb77SJakub Sitnicki * (*sk->protocol*) doesn't match packet protocol 3964e9ddbb77SJakub Sitnicki * (*ctx->protocol*). 3965e9ddbb77SJakub Sitnicki * 3966e9ddbb77SJakub Sitnicki * * **-ESOCKTNOSUPPORT** if socket is not in allowed 3967e9ddbb77SJakub Sitnicki * state (TCP listening or UDP unconnected). 3968e9ddbb77SJakub Sitnicki * 396971d19214SMaciej Żenczykowski * u64 bpf_ktime_get_boot_ns(void) 397071d19214SMaciej Żenczykowski * Description 397171d19214SMaciej Żenczykowski * Return the time elapsed since system boot, in nanoseconds. 397271d19214SMaciej Żenczykowski * Does include the time the system was suspended. 3973ab8d7809SQuentin Monnet * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 397471d19214SMaciej Żenczykowski * Return 397571d19214SMaciej Żenczykowski * Current *ktime*. 3976492e639fSYonghong Song * 3977bdb7b79bSAndrii Nakryiko * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 3978492e639fSYonghong Song * Description 3979ab8d7809SQuentin Monnet * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 3980ab8d7809SQuentin Monnet * out the format string. 3981492e639fSYonghong Song * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 3982492e639fSYonghong Song * the format string itself. The *data* and *data_len* are format string 3983ab8d7809SQuentin Monnet * arguments. The *data* are a **u64** array and corresponding format string 3984492e639fSYonghong Song * values are stored in the array. For strings and pointers where pointees 3985492e639fSYonghong Song * are accessed, only the pointer values are stored in the *data* array. 3986ab8d7809SQuentin Monnet * The *data_len* is the size of *data* in bytes. 3987492e639fSYonghong Song * 3988492e639fSYonghong Song * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 3989492e639fSYonghong Song * Reading kernel memory may fail due to either invalid address or 3990492e639fSYonghong Song * valid address but requiring a major memory fault. If reading kernel memory 3991492e639fSYonghong Song * fails, the string for **%s** will be an empty string, and the ip 3992492e639fSYonghong Song * address for **%p{i,I}{4,6}** will be 0. Not returning error to 3993ab8d7809SQuentin Monnet * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 3994492e639fSYonghong Song * Return 3995ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 3996492e639fSYonghong Song * 3997ab8d7809SQuentin Monnet * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 3998492e639fSYonghong Song * by returning 1 from bpf program. 3999ab8d7809SQuentin Monnet * 4000ab8d7809SQuentin Monnet * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4001ab8d7809SQuentin Monnet * 4002ab8d7809SQuentin Monnet * **-E2BIG** if *fmt* contains too many format specifiers. 4003ab8d7809SQuentin Monnet * 4004ab8d7809SQuentin Monnet * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4005492e639fSYonghong Song * 4006bdb7b79bSAndrii Nakryiko * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4007492e639fSYonghong Song * Description 4008ab8d7809SQuentin Monnet * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4009492e639fSYonghong Song * The *m* represents the seq_file. The *data* and *len* represent the 4010492e639fSYonghong Song * data to write in bytes. 4011492e639fSYonghong Song * Return 4012ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 4013492e639fSYonghong Song * 4014ab8d7809SQuentin Monnet * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4015f307fa2cSAndrey Ignatov * 4016a5fa25adSMartin KaFai Lau * u64 bpf_sk_cgroup_id(void *sk) 4017f307fa2cSAndrey Ignatov * Description 4018f307fa2cSAndrey Ignatov * Return the cgroup v2 id of the socket *sk*. 4019f307fa2cSAndrey Ignatov * 4020a5fa25adSMartin KaFai Lau * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4021f307fa2cSAndrey Ignatov * returned from **bpf_sk_lookup_xxx**\ (), 4022f307fa2cSAndrey Ignatov * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4023f307fa2cSAndrey Ignatov * same as in **bpf_skb_cgroup_id**\ (). 4024f307fa2cSAndrey Ignatov * 4025f307fa2cSAndrey Ignatov * This helper is available only if the kernel was compiled with 4026f307fa2cSAndrey Ignatov * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4027f307fa2cSAndrey Ignatov * Return 4028f307fa2cSAndrey Ignatov * The id is returned or 0 in case the id could not be retrieved. 4029f307fa2cSAndrey Ignatov * 4030a5fa25adSMartin KaFai Lau * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4031f307fa2cSAndrey Ignatov * Description 4032f307fa2cSAndrey Ignatov * Return id of cgroup v2 that is ancestor of cgroup associated 4033f307fa2cSAndrey Ignatov * with the *sk* at the *ancestor_level*. The root cgroup is at 4034f307fa2cSAndrey Ignatov * *ancestor_level* zero and each step down the hierarchy 4035f307fa2cSAndrey Ignatov * increments the level. If *ancestor_level* == level of cgroup 4036f307fa2cSAndrey Ignatov * associated with *sk*, then return value will be same as that 4037f307fa2cSAndrey Ignatov * of **bpf_sk_cgroup_id**\ (). 4038f307fa2cSAndrey Ignatov * 4039f307fa2cSAndrey Ignatov * The helper is useful to implement policies based on cgroups 4040f307fa2cSAndrey Ignatov * that are upper in hierarchy than immediate cgroup associated 4041f307fa2cSAndrey Ignatov * with *sk*. 4042f307fa2cSAndrey Ignatov * 4043f307fa2cSAndrey Ignatov * The format of returned id and helper limitations are same as in 4044f307fa2cSAndrey Ignatov * **bpf_sk_cgroup_id**\ (). 4045f307fa2cSAndrey Ignatov * Return 4046f307fa2cSAndrey Ignatov * The id is returned or 0 in case the id could not be retrieved. 4047457f4436SAndrii Nakryiko * 4048e1613b57SAndrii Nakryiko * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4049457f4436SAndrii Nakryiko * Description 4050457f4436SAndrii Nakryiko * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4051bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4052bcc7f554SQuentin Monnet * of new data availability is sent. 4053bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4054bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 4055457f4436SAndrii Nakryiko * Return 4056bcc7f554SQuentin Monnet * 0 on success, or a negative error in case of failure. 4057457f4436SAndrii Nakryiko * 4058457f4436SAndrii Nakryiko * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4059457f4436SAndrii Nakryiko * Description 4060457f4436SAndrii Nakryiko * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4061457f4436SAndrii Nakryiko * Return 4062457f4436SAndrii Nakryiko * Valid pointer with *size* bytes of memory available; NULL, 4063457f4436SAndrii Nakryiko * otherwise. 4064457f4436SAndrii Nakryiko * 4065457f4436SAndrii Nakryiko * void bpf_ringbuf_submit(void *data, u64 flags) 4066457f4436SAndrii Nakryiko * Description 4067457f4436SAndrii Nakryiko * Submit reserved ring buffer sample, pointed to by *data*. 4068bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4069bcc7f554SQuentin Monnet * of new data availability is sent. 4070bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4071bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 4072457f4436SAndrii Nakryiko * Return 4073457f4436SAndrii Nakryiko * Nothing. Always succeeds. 4074457f4436SAndrii Nakryiko * 4075457f4436SAndrii Nakryiko * void bpf_ringbuf_discard(void *data, u64 flags) 4076457f4436SAndrii Nakryiko * Description 4077457f4436SAndrii Nakryiko * Discard reserved ring buffer sample, pointed to by *data*. 4078bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4079bcc7f554SQuentin Monnet * of new data availability is sent. 4080bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4081bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 4082457f4436SAndrii Nakryiko * Return 4083457f4436SAndrii Nakryiko * Nothing. Always succeeds. 4084457f4436SAndrii Nakryiko * 4085457f4436SAndrii Nakryiko * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4086457f4436SAndrii Nakryiko * Description 4087457f4436SAndrii Nakryiko * Query various characteristics of provided ring buffer. What 4088457f4436SAndrii Nakryiko * exactly is queries is determined by *flags*: 4089bcc7f554SQuentin Monnet * 4090bcc7f554SQuentin Monnet * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4091bcc7f554SQuentin Monnet * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4092bcc7f554SQuentin Monnet * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4093bcc7f554SQuentin Monnet * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4094bcc7f554SQuentin Monnet * 4095bcc7f554SQuentin Monnet * Data returned is just a momentary snapshot of actual values 4096457f4436SAndrii Nakryiko * and could be inaccurate, so this facility should be used to 4097457f4436SAndrii Nakryiko * power heuristics and for reporting, not to make 100% correct 4098457f4436SAndrii Nakryiko * calculation. 4099457f4436SAndrii Nakryiko * Return 4100bcc7f554SQuentin Monnet * Requested value, or 0, if *flags* are not recognized. 41017cdec54fSDaniel Borkmann * 4102bdb7b79bSAndrii Nakryiko * long bpf_csum_level(struct sk_buff *skb, u64 level) 41037cdec54fSDaniel Borkmann * Description 41047cdec54fSDaniel Borkmann * Change the skbs checksum level by one layer up or down, or 41057cdec54fSDaniel Borkmann * reset it entirely to none in order to have the stack perform 41067cdec54fSDaniel Borkmann * checksum validation. The level is applicable to the following 41077cdec54fSDaniel Borkmann * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 41087cdec54fSDaniel Borkmann * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 41097cdec54fSDaniel Borkmann * through **bpf_skb_adjust_room**\ () helper with passing in 41107cdec54fSDaniel Borkmann * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 41117cdec54fSDaniel Borkmann * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 41127cdec54fSDaniel Borkmann * the UDP header is removed. Similarly, an encap of the latter 41137cdec54fSDaniel Borkmann * into the former could be accompanied by a helper call to 41147cdec54fSDaniel Borkmann * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 41157cdec54fSDaniel Borkmann * skb is still intended to be processed in higher layers of the 41167cdec54fSDaniel Borkmann * stack instead of just egressing at tc. 41177cdec54fSDaniel Borkmann * 41187cdec54fSDaniel Borkmann * There are three supported level settings at this time: 41197cdec54fSDaniel Borkmann * 41207cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 41217cdec54fSDaniel Borkmann * with CHECKSUM_UNNECESSARY. 41227cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 41237cdec54fSDaniel Borkmann * with CHECKSUM_UNNECESSARY. 41247cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 41257cdec54fSDaniel Borkmann * sets CHECKSUM_NONE to force checksum validation by the stack. 41267cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 41277cdec54fSDaniel Borkmann * skb->csum_level. 41287cdec54fSDaniel Borkmann * Return 41297cdec54fSDaniel Borkmann * 0 on success, or a negative error in case of failure. In the 41307cdec54fSDaniel Borkmann * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 41317cdec54fSDaniel Borkmann * is returned or the error code -EACCES in case the skb is not 41327cdec54fSDaniel Borkmann * subject to CHECKSUM_UNNECESSARY. 4133af7ec138SYonghong Song * 4134af7ec138SYonghong Song * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4135af7ec138SYonghong Song * Description 4136af7ec138SYonghong Song * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4137af7ec138SYonghong Song * Return 4138938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4139478cfbdfSYonghong Song * 4140478cfbdfSYonghong Song * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4141478cfbdfSYonghong Song * Description 4142478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4143478cfbdfSYonghong Song * Return 4144938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4145478cfbdfSYonghong Song * 4146478cfbdfSYonghong Song * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4147478cfbdfSYonghong Song * Description 4148478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4149478cfbdfSYonghong Song * Return 4150938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4151478cfbdfSYonghong Song * 4152478cfbdfSYonghong Song * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4153478cfbdfSYonghong Song * Description 4154478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4155478cfbdfSYonghong Song * Return 4156938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 41570d4fad3eSYonghong Song * 41580d4fad3eSYonghong Song * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 41590d4fad3eSYonghong Song * Description 41600d4fad3eSYonghong Song * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 41610d4fad3eSYonghong Song * Return 4162938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4163fa28dcb8SSong Liu * 4164fa28dcb8SSong Liu * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4165fa28dcb8SSong Liu * Description 4166fa28dcb8SSong Liu * Return a user or a kernel stack in bpf program provided buffer. 4167fa28dcb8SSong Liu * To achieve this, the helper needs *task*, which is a valid 4168938c3efdSQuentin Monnet * pointer to **struct task_struct**. To store the stacktrace, the 4169fa28dcb8SSong Liu * bpf program provides *buf* with a nonnegative *size*. 4170fa28dcb8SSong Liu * 4171fa28dcb8SSong Liu * The last argument, *flags*, holds the number of stack frames to 4172fa28dcb8SSong Liu * skip (from 0 to 255), masked with 4173fa28dcb8SSong Liu * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4174fa28dcb8SSong Liu * the following flags: 4175fa28dcb8SSong Liu * 4176fa28dcb8SSong Liu * **BPF_F_USER_STACK** 4177fa28dcb8SSong Liu * Collect a user space stack instead of a kernel stack. 4178fa28dcb8SSong Liu * **BPF_F_USER_BUILD_ID** 4179fa28dcb8SSong Liu * Collect buildid+offset instead of ips for user stack, 4180fa28dcb8SSong Liu * only valid if **BPF_F_USER_STACK** is also specified. 4181fa28dcb8SSong Liu * 4182fa28dcb8SSong Liu * **bpf_get_task_stack**\ () can collect up to 4183fa28dcb8SSong Liu * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4184fa28dcb8SSong Liu * to sufficient large buffer size. Note that 4185fa28dcb8SSong Liu * this limit can be controlled with the **sysctl** program, and 4186fa28dcb8SSong Liu * that it should be manually increased in order to profile long 4187fa28dcb8SSong Liu * user stacks (such as stacks for Java programs). To do so, use: 4188fa28dcb8SSong Liu * 4189fa28dcb8SSong Liu * :: 4190fa28dcb8SSong Liu * 4191fa28dcb8SSong Liu * # sysctl kernel.perf_event_max_stack=<new value> 4192fa28dcb8SSong Liu * Return 4193fa28dcb8SSong Liu * A non-negative value equal to or less than *size* on success, 4194fa28dcb8SSong Liu * or a negative error in case of failure. 4195fa28dcb8SSong Liu * 41960813a841SMartin KaFai Lau * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 41970813a841SMartin KaFai Lau * Description 41980813a841SMartin KaFai Lau * Load header option. Support reading a particular TCP header 4199938c3efdSQuentin Monnet * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 42000813a841SMartin KaFai Lau * 42010813a841SMartin KaFai Lau * If *flags* is 0, it will search the option from the 4202938c3efdSQuentin Monnet * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 42030813a841SMartin KaFai Lau * has details on what skb_data contains under different 4204938c3efdSQuentin Monnet * *skops*\ **->op**. 42050813a841SMartin KaFai Lau * 42060813a841SMartin KaFai Lau * The first byte of the *searchby_res* specifies the 42070813a841SMartin KaFai Lau * kind that it wants to search. 42080813a841SMartin KaFai Lau * 42090813a841SMartin KaFai Lau * If the searching kind is an experimental kind 42100813a841SMartin KaFai Lau * (i.e. 253 or 254 according to RFC6994). It also 42110813a841SMartin KaFai Lau * needs to specify the "magic" which is either 42120813a841SMartin KaFai Lau * 2 bytes or 4 bytes. It then also needs to 42130813a841SMartin KaFai Lau * specify the size of the magic by using 42140813a841SMartin KaFai Lau * the 2nd byte which is "kind-length" of a TCP 42150813a841SMartin KaFai Lau * header option and the "kind-length" also 42160813a841SMartin KaFai Lau * includes the first 2 bytes "kind" and "kind-length" 42170813a841SMartin KaFai Lau * itself as a normal TCP header option also does. 42180813a841SMartin KaFai Lau * 42190813a841SMartin KaFai Lau * For example, to search experimental kind 254 with 42200813a841SMartin KaFai Lau * 2 byte magic 0xeB9F, the searchby_res should be 42210813a841SMartin KaFai Lau * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 42220813a841SMartin KaFai Lau * 42230813a841SMartin KaFai Lau * To search for the standard window scale option (3), 4224938c3efdSQuentin Monnet * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 42250813a841SMartin KaFai Lau * Note, kind-length must be 0 for regular option. 42260813a841SMartin KaFai Lau * 42270813a841SMartin KaFai Lau * Searching for No-Op (0) and End-of-Option-List (1) are 42280813a841SMartin KaFai Lau * not supported. 42290813a841SMartin KaFai Lau * 42300813a841SMartin KaFai Lau * *len* must be at least 2 bytes which is the minimal size 42310813a841SMartin KaFai Lau * of a header option. 42320813a841SMartin KaFai Lau * 42330813a841SMartin KaFai Lau * Supported flags: 4234938c3efdSQuentin Monnet * 42350813a841SMartin KaFai Lau * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 42360813a841SMartin KaFai Lau * saved_syn packet or the just-received syn packet. 42370813a841SMartin KaFai Lau * 42380813a841SMartin KaFai Lau * Return 42390813a841SMartin KaFai Lau * > 0 when found, the header option is copied to *searchby_res*. 4240938c3efdSQuentin Monnet * The return value is the total length copied. On failure, a 4241938c3efdSQuentin Monnet * negative error code is returned: 42420813a841SMartin KaFai Lau * 4243938c3efdSQuentin Monnet * **-EINVAL** if a parameter is invalid. 42440813a841SMartin KaFai Lau * 4245938c3efdSQuentin Monnet * **-ENOMSG** if the option is not found. 42460813a841SMartin KaFai Lau * 4247938c3efdSQuentin Monnet * **-ENOENT** if no syn packet is available when 4248938c3efdSQuentin Monnet * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 42490813a841SMartin KaFai Lau * 4250938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space. Only *len* number of 42510813a841SMartin KaFai Lau * bytes are copied. 42520813a841SMartin KaFai Lau * 4253938c3efdSQuentin Monnet * **-EFAULT** on failure to parse the header options in the 4254938c3efdSQuentin Monnet * packet. 42550813a841SMartin KaFai Lau * 4256938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4257938c3efdSQuentin Monnet * *skops*\ **->op**. 42580813a841SMartin KaFai Lau * 42590813a841SMartin KaFai Lau * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 42600813a841SMartin KaFai Lau * Description 42610813a841SMartin KaFai Lau * Store header option. The data will be copied 42620813a841SMartin KaFai Lau * from buffer *from* with length *len* to the TCP header. 42630813a841SMartin KaFai Lau * 42640813a841SMartin KaFai Lau * The buffer *from* should have the whole option that 42650813a841SMartin KaFai Lau * includes the kind, kind-length, and the actual 42660813a841SMartin KaFai Lau * option data. The *len* must be at least kind-length 42670813a841SMartin KaFai Lau * long. The kind-length does not have to be 4 byte 42680813a841SMartin KaFai Lau * aligned. The kernel will take care of the padding 42690813a841SMartin KaFai Lau * and setting the 4 bytes aligned value to th->doff. 42700813a841SMartin KaFai Lau * 42710813a841SMartin KaFai Lau * This helper will check for duplicated option 42720813a841SMartin KaFai Lau * by searching the same option in the outgoing skb. 42730813a841SMartin KaFai Lau * 42740813a841SMartin KaFai Lau * This helper can only be called during 4275938c3efdSQuentin Monnet * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 42760813a841SMartin KaFai Lau * 42770813a841SMartin KaFai Lau * Return 42780813a841SMartin KaFai Lau * 0 on success, or negative error in case of failure: 42790813a841SMartin KaFai Lau * 4280938c3efdSQuentin Monnet * **-EINVAL** If param is invalid. 42810813a841SMartin KaFai Lau * 4282938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space in the header. 42830813a841SMartin KaFai Lau * Nothing has been written 42840813a841SMartin KaFai Lau * 4285938c3efdSQuentin Monnet * **-EEXIST** if the option already exists. 42860813a841SMartin KaFai Lau * 4287938c3efdSQuentin Monnet * **-EFAULT** on failrue to parse the existing header options. 42880813a841SMartin KaFai Lau * 4289938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4290938c3efdSQuentin Monnet * *skops*\ **->op**. 42910813a841SMartin KaFai Lau * 42920813a841SMartin KaFai Lau * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 42930813a841SMartin KaFai Lau * Description 42940813a841SMartin KaFai Lau * Reserve *len* bytes for the bpf header option. The 4295938c3efdSQuentin Monnet * space will be used by **bpf_store_hdr_opt**\ () later in 4296938c3efdSQuentin Monnet * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 42970813a841SMartin KaFai Lau * 4298938c3efdSQuentin Monnet * If **bpf_reserve_hdr_opt**\ () is called multiple times, 42990813a841SMartin KaFai Lau * the total number of bytes will be reserved. 43000813a841SMartin KaFai Lau * 43010813a841SMartin KaFai Lau * This helper can only be called during 4302938c3efdSQuentin Monnet * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 43030813a841SMartin KaFai Lau * 43040813a841SMartin KaFai Lau * Return 43050813a841SMartin KaFai Lau * 0 on success, or negative error in case of failure: 43060813a841SMartin KaFai Lau * 4307938c3efdSQuentin Monnet * **-EINVAL** if a parameter is invalid. 43080813a841SMartin KaFai Lau * 4309938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space in the header. 43100813a841SMartin KaFai Lau * 4311938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4312938c3efdSQuentin Monnet * *skops*\ **->op**. 43136e22ab9dSJiri Olsa * 43148ea63684SKP Singh * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 43158ea63684SKP Singh * Description 43168ea63684SKP Singh * Get a bpf_local_storage from an *inode*. 43178ea63684SKP Singh * 43188ea63684SKP Singh * Logically, it could be thought of as getting the value from 43198ea63684SKP Singh * a *map* with *inode* as the **key**. From this 43208ea63684SKP Singh * perspective, the usage is not much different from 43218ea63684SKP Singh * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 43228ea63684SKP Singh * helper enforces the key must be an inode and the map must also 43238ea63684SKP Singh * be a **BPF_MAP_TYPE_INODE_STORAGE**. 43248ea63684SKP Singh * 43258ea63684SKP Singh * Underneath, the value is stored locally at *inode* instead of 43268ea63684SKP Singh * the *map*. The *map* is used as the bpf-local-storage 43278ea63684SKP Singh * "type". The bpf-local-storage "type" (i.e. the *map*) is 43288ea63684SKP Singh * searched against all bpf_local_storage residing at *inode*. 43298ea63684SKP Singh * 43308ea63684SKP Singh * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 43318ea63684SKP Singh * used such that a new bpf_local_storage will be 43328ea63684SKP Singh * created if one does not exist. *value* can be used 43338ea63684SKP Singh * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 43348ea63684SKP Singh * the initial value of a bpf_local_storage. If *value* is 43358ea63684SKP Singh * **NULL**, the new bpf_local_storage will be zero initialized. 43368ea63684SKP Singh * Return 43378ea63684SKP Singh * A bpf_local_storage pointer is returned on success. 43388ea63684SKP Singh * 43398ea63684SKP Singh * **NULL** if not found or there was an error in adding 43408ea63684SKP Singh * a new bpf_local_storage. 43418ea63684SKP Singh * 43428ea63684SKP Singh * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 43438ea63684SKP Singh * Description 43448ea63684SKP Singh * Delete a bpf_local_storage from an *inode*. 43458ea63684SKP Singh * Return 43468ea63684SKP Singh * 0 on success. 43478ea63684SKP Singh * 43488ea63684SKP Singh * **-ENOENT** if the bpf_local_storage cannot be found. 43496e22ab9dSJiri Olsa * 43506e22ab9dSJiri Olsa * long bpf_d_path(struct path *path, char *buf, u32 sz) 43516e22ab9dSJiri Olsa * Description 4352938c3efdSQuentin Monnet * Return full path for given **struct path** object, which 4353938c3efdSQuentin Monnet * needs to be the kernel BTF *path* object. The path is 4354938c3efdSQuentin Monnet * returned in the provided buffer *buf* of size *sz* and 43556e22ab9dSJiri Olsa * is zero terminated. 43566e22ab9dSJiri Olsa * 43576e22ab9dSJiri Olsa * Return 43586e22ab9dSJiri Olsa * On success, the strictly positive length of the string, 43596e22ab9dSJiri Olsa * including the trailing NUL character. On error, a negative 43606e22ab9dSJiri Olsa * value. 436107be4c4aSAlexei Starovoitov * 436207be4c4aSAlexei Starovoitov * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 436307be4c4aSAlexei Starovoitov * Description 436407be4c4aSAlexei Starovoitov * Read *size* bytes from user space address *user_ptr* and store 4365938c3efdSQuentin Monnet * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 436607be4c4aSAlexei Starovoitov * Return 436707be4c4aSAlexei Starovoitov * 0 on success, or a negative error in case of failure. 4368c4d0bfb4SAlan Maguire * 4369c4d0bfb4SAlan Maguire * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4370c4d0bfb4SAlan Maguire * Description 4371c4d0bfb4SAlan Maguire * Use BTF to store a string representation of *ptr*->ptr in *str*, 4372c4d0bfb4SAlan Maguire * using *ptr*->type_id. This value should specify the type 4373c4d0bfb4SAlan Maguire * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4374c4d0bfb4SAlan Maguire * can be used to look up vmlinux BTF type ids. Traversing the 4375c4d0bfb4SAlan Maguire * data structure using BTF, the type information and values are 4376c4d0bfb4SAlan Maguire * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4377c4d0bfb4SAlan Maguire * the pointer data is carried out to avoid kernel crashes during 4378c4d0bfb4SAlan Maguire * operation. Smaller types can use string space on the stack; 4379c4d0bfb4SAlan Maguire * larger programs can use map data to store the string 4380c4d0bfb4SAlan Maguire * representation. 4381c4d0bfb4SAlan Maguire * 4382c4d0bfb4SAlan Maguire * The string can be subsequently shared with userspace via 4383c4d0bfb4SAlan Maguire * bpf_perf_event_output() or ring buffer interfaces. 4384c4d0bfb4SAlan Maguire * bpf_trace_printk() is to be avoided as it places too small 4385c4d0bfb4SAlan Maguire * a limit on string size to be useful. 4386c4d0bfb4SAlan Maguire * 4387c4d0bfb4SAlan Maguire * *flags* is a combination of 4388c4d0bfb4SAlan Maguire * 4389c4d0bfb4SAlan Maguire * **BTF_F_COMPACT** 4390c4d0bfb4SAlan Maguire * no formatting around type information 4391c4d0bfb4SAlan Maguire * **BTF_F_NONAME** 4392c4d0bfb4SAlan Maguire * no struct/union member names/types 4393c4d0bfb4SAlan Maguire * **BTF_F_PTR_RAW** 4394c4d0bfb4SAlan Maguire * show raw (unobfuscated) pointer values; 4395c4d0bfb4SAlan Maguire * equivalent to printk specifier %px. 4396c4d0bfb4SAlan Maguire * **BTF_F_ZERO** 4397c4d0bfb4SAlan Maguire * show zero-valued struct/union members; they 4398c4d0bfb4SAlan Maguire * are not displayed by default 4399c4d0bfb4SAlan Maguire * 4400c4d0bfb4SAlan Maguire * Return 4401c4d0bfb4SAlan Maguire * The number of bytes that were written (or would have been 4402c4d0bfb4SAlan Maguire * written if output had to be truncated due to string size), 4403c4d0bfb4SAlan Maguire * or a negative error in cases of failure. 4404eb411377SAlan Maguire * 4405eb411377SAlan Maguire * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4406eb411377SAlan Maguire * Description 4407eb411377SAlan Maguire * Use BTF to write to seq_write a string representation of 4408eb411377SAlan Maguire * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4409eb411377SAlan Maguire * *flags* are identical to those used for bpf_snprintf_btf. 4410eb411377SAlan Maguire * Return 4411eb411377SAlan Maguire * 0 on success or a negative error in case of failure. 4412b426ce83SDaniel Borkmann * 4413b426ce83SDaniel Borkmann * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4414b426ce83SDaniel Borkmann * Description 4415b426ce83SDaniel Borkmann * See **bpf_get_cgroup_classid**\ () for the main description. 4416b426ce83SDaniel Borkmann * This helper differs from **bpf_get_cgroup_classid**\ () in that 4417b426ce83SDaniel Borkmann * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4418b426ce83SDaniel Borkmann * associated socket instead of the current process. 4419b426ce83SDaniel Borkmann * Return 4420b426ce83SDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 4421b4ab3141SDaniel Borkmann * 4422ba452c9eSToke Høiland-Jørgensen * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4423b4ab3141SDaniel Borkmann * Description 4424b4ab3141SDaniel Borkmann * Redirect the packet to another net device of index *ifindex* 4425b4ab3141SDaniel Borkmann * and fill in L2 addresses from neighboring subsystem. This helper 4426b4ab3141SDaniel Borkmann * is somewhat similar to **bpf_redirect**\ (), except that it 4427dd2ce6a5SDaniel Borkmann * populates L2 addresses as well, meaning, internally, the helper 4428ba452c9eSToke Høiland-Jørgensen * relies on the neighbor lookup for the L2 address of the nexthop. 4429ba452c9eSToke Høiland-Jørgensen * 4430ba452c9eSToke Høiland-Jørgensen * The helper will perform a FIB lookup based on the skb's 4431ba452c9eSToke Høiland-Jørgensen * networking header to get the address of the next hop, unless 4432ba452c9eSToke Høiland-Jørgensen * this is supplied by the caller in the *params* argument. The 4433ba452c9eSToke Høiland-Jørgensen * *plen* argument indicates the len of *params* and should be set 4434ba452c9eSToke Høiland-Jørgensen * to 0 if *params* is NULL. 4435dd2ce6a5SDaniel Borkmann * 4436b4ab3141SDaniel Borkmann * The *flags* argument is reserved and must be 0. The helper is 4437dd2ce6a5SDaniel Borkmann * currently only supported for tc BPF program types, and enabled 4438dd2ce6a5SDaniel Borkmann * for IPv4 and IPv6 protocols. 4439b4ab3141SDaniel Borkmann * Return 4440b4ab3141SDaniel Borkmann * The helper returns **TC_ACT_REDIRECT** on success or 4441b4ab3141SDaniel Borkmann * **TC_ACT_SHOT** on error. 4442eaa6bcb7SHao Luo * 4443eaa6bcb7SHao Luo * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4444eaa6bcb7SHao Luo * Description 4445eaa6bcb7SHao Luo * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4446eaa6bcb7SHao Luo * pointer to the percpu kernel variable on *cpu*. A ksym is an 4447eaa6bcb7SHao Luo * extern variable decorated with '__ksym'. For ksym, there is a 4448eaa6bcb7SHao Luo * global var (either static or global) defined of the same name 4449eaa6bcb7SHao Luo * in the kernel. The ksym is percpu if the global var is percpu. 4450eaa6bcb7SHao Luo * The returned pointer points to the global percpu var on *cpu*. 4451eaa6bcb7SHao Luo * 4452eaa6bcb7SHao Luo * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4453eaa6bcb7SHao Luo * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4454eaa6bcb7SHao Luo * happens if *cpu* is larger than nr_cpu_ids. The caller of 4455eaa6bcb7SHao Luo * bpf_per_cpu_ptr() must check the returned value. 4456eaa6bcb7SHao Luo * Return 4457eaa6bcb7SHao Luo * A pointer pointing to the kernel percpu variable on *cpu*, or 4458eaa6bcb7SHao Luo * NULL, if *cpu* is invalid. 445963d9b80dSHao Luo * 446063d9b80dSHao Luo * void *bpf_this_cpu_ptr(const void *percpu_ptr) 446163d9b80dSHao Luo * Description 446263d9b80dSHao Luo * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 446363d9b80dSHao Luo * pointer to the percpu kernel variable on this cpu. See the 446463d9b80dSHao Luo * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 446563d9b80dSHao Luo * 446663d9b80dSHao Luo * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 446763d9b80dSHao Luo * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 446863d9b80dSHao Luo * never return NULL. 446963d9b80dSHao Luo * Return 447063d9b80dSHao Luo * A pointer pointing to the kernel percpu variable on this cpu. 44719aa1206eSDaniel Borkmann * 44729aa1206eSDaniel Borkmann * long bpf_redirect_peer(u32 ifindex, u64 flags) 44739aa1206eSDaniel Borkmann * Description 44749aa1206eSDaniel Borkmann * Redirect the packet to another net device of index *ifindex*. 44759aa1206eSDaniel Borkmann * This helper is somewhat similar to **bpf_redirect**\ (), except 44769aa1206eSDaniel Borkmann * that the redirection happens to the *ifindex*' peer device and 44779aa1206eSDaniel Borkmann * the netns switch takes place from ingress to ingress without 44789aa1206eSDaniel Borkmann * going through the CPU's backlog queue. 44799aa1206eSDaniel Borkmann * 44809aa1206eSDaniel Borkmann * The *flags* argument is reserved and must be 0. The helper is 44819aa1206eSDaniel Borkmann * currently only supported for tc BPF program types at the ingress 44829aa1206eSDaniel Borkmann * hook and for veth device types. The peer device must reside in a 44839aa1206eSDaniel Borkmann * different network namespace. 44849aa1206eSDaniel Borkmann * Return 44859aa1206eSDaniel Borkmann * The helper returns **TC_ACT_REDIRECT** on success or 44869aa1206eSDaniel Borkmann * **TC_ACT_SHOT** on error. 44874cf1bc1fSKP Singh * 44884cf1bc1fSKP Singh * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 44894cf1bc1fSKP Singh * Description 44904cf1bc1fSKP Singh * Get a bpf_local_storage from the *task*. 44914cf1bc1fSKP Singh * 44924cf1bc1fSKP Singh * Logically, it could be thought of as getting the value from 44934cf1bc1fSKP Singh * a *map* with *task* as the **key**. From this 44944cf1bc1fSKP Singh * perspective, the usage is not much different from 44954cf1bc1fSKP Singh * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 44964cf1bc1fSKP Singh * helper enforces the key must be an task_struct and the map must also 44974cf1bc1fSKP Singh * be a **BPF_MAP_TYPE_TASK_STORAGE**. 44984cf1bc1fSKP Singh * 44994cf1bc1fSKP Singh * Underneath, the value is stored locally at *task* instead of 45004cf1bc1fSKP Singh * the *map*. The *map* is used as the bpf-local-storage 45014cf1bc1fSKP Singh * "type". The bpf-local-storage "type" (i.e. the *map*) is 45024cf1bc1fSKP Singh * searched against all bpf_local_storage residing at *task*. 45034cf1bc1fSKP Singh * 45044cf1bc1fSKP Singh * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 45054cf1bc1fSKP Singh * used such that a new bpf_local_storage will be 45064cf1bc1fSKP Singh * created if one does not exist. *value* can be used 45074cf1bc1fSKP Singh * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 45084cf1bc1fSKP Singh * the initial value of a bpf_local_storage. If *value* is 45094cf1bc1fSKP Singh * **NULL**, the new bpf_local_storage will be zero initialized. 45104cf1bc1fSKP Singh * Return 45114cf1bc1fSKP Singh * A bpf_local_storage pointer is returned on success. 45124cf1bc1fSKP Singh * 45134cf1bc1fSKP Singh * **NULL** if not found or there was an error in adding 45144cf1bc1fSKP Singh * a new bpf_local_storage. 45154cf1bc1fSKP Singh * 45164cf1bc1fSKP Singh * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 45174cf1bc1fSKP Singh * Description 45184cf1bc1fSKP Singh * Delete a bpf_local_storage from a *task*. 45194cf1bc1fSKP Singh * Return 45204cf1bc1fSKP Singh * 0 on success. 45214cf1bc1fSKP Singh * 45224cf1bc1fSKP Singh * **-ENOENT** if the bpf_local_storage cannot be found. 45233ca1032aSKP Singh * 45243ca1032aSKP Singh * struct task_struct *bpf_get_current_task_btf(void) 45253ca1032aSKP Singh * Description 45263ca1032aSKP Singh * Return a BTF pointer to the "current" task. 45273ca1032aSKP Singh * This pointer can also be used in helpers that accept an 45283ca1032aSKP Singh * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 45293ca1032aSKP Singh * Return 45303ca1032aSKP Singh * Pointer to the current task. 45313f6719c7SKP Singh * 45323f6719c7SKP Singh * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 45333f6719c7SKP Singh * Description 45343f6719c7SKP Singh * Set or clear certain options on *bprm*: 45353f6719c7SKP Singh * 45363f6719c7SKP Singh * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 45373f6719c7SKP Singh * which sets the **AT_SECURE** auxv for glibc. The bit 45383f6719c7SKP Singh * is cleared if the flag is not specified. 45393f6719c7SKP Singh * Return 45403f6719c7SKP Singh * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4541d0551261SDmitrii Banshchikov * 4542d0551261SDmitrii Banshchikov * u64 bpf_ktime_get_coarse_ns(void) 4543d0551261SDmitrii Banshchikov * Description 4544d0551261SDmitrii Banshchikov * Return a coarse-grained version of the time elapsed since 4545d0551261SDmitrii Banshchikov * system boot, in nanoseconds. Does not include time the system 4546d0551261SDmitrii Banshchikov * was suspended. 4547d0551261SDmitrii Banshchikov * 4548d0551261SDmitrii Banshchikov * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4549d0551261SDmitrii Banshchikov * Return 4550d0551261SDmitrii Banshchikov * Current *ktime*. 455127672f0dSKP Singh * 455227672f0dSKP Singh * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 455327672f0dSKP Singh * Description 455427672f0dSKP Singh * Returns the stored IMA hash of the *inode* (if it's avaialable). 455527672f0dSKP Singh * If the hash is larger than *size*, then only *size* 455627672f0dSKP Singh * bytes will be copied to *dst* 455727672f0dSKP Singh * Return 455827672f0dSKP Singh * The **hash_algo** is returned on success, 455927672f0dSKP Singh * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 456027672f0dSKP Singh * invalid arguments are passed. 45614f19cab7SFlorent Revest * 45624f19cab7SFlorent Revest * struct socket *bpf_sock_from_file(struct file *file) 45634f19cab7SFlorent Revest * Description 45644f19cab7SFlorent Revest * If the given file represents a socket, returns the associated 45654f19cab7SFlorent Revest * socket. 45664f19cab7SFlorent Revest * Return 45674f19cab7SFlorent Revest * A pointer to a struct socket on success or NULL if the file is 45684f19cab7SFlorent Revest * not a socket. 456934b2021cSJesper Dangaard Brouer * 457034b2021cSJesper Dangaard Brouer * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 457134b2021cSJesper Dangaard Brouer * Description 457234b2021cSJesper Dangaard Brouer * Check ctx packet size against exceeding MTU of net device (based 457334b2021cSJesper Dangaard Brouer * on *ifindex*). This helper will likely be used in combination 457434b2021cSJesper Dangaard Brouer * with helpers that adjust/change the packet size. 457534b2021cSJesper Dangaard Brouer * 457634b2021cSJesper Dangaard Brouer * The argument *len_diff* can be used for querying with a planned 457734b2021cSJesper Dangaard Brouer * size change. This allows to check MTU prior to changing packet 457834b2021cSJesper Dangaard Brouer * ctx. Providing an *len_diff* adjustment that is larger than the 457934b2021cSJesper Dangaard Brouer * actual packet size (resulting in negative packet size) will in 458034b2021cSJesper Dangaard Brouer * principle not exceed the MTU, why it is not considered a 458134b2021cSJesper Dangaard Brouer * failure. Other BPF-helpers are needed for performing the 458234b2021cSJesper Dangaard Brouer * planned size change, why the responsability for catch a negative 458334b2021cSJesper Dangaard Brouer * packet size belong in those helpers. 458434b2021cSJesper Dangaard Brouer * 458534b2021cSJesper Dangaard Brouer * Specifying *ifindex* zero means the MTU check is performed 458634b2021cSJesper Dangaard Brouer * against the current net device. This is practical if this isn't 458734b2021cSJesper Dangaard Brouer * used prior to redirect. 458834b2021cSJesper Dangaard Brouer * 458934b2021cSJesper Dangaard Brouer * The Linux kernel route table can configure MTUs on a more 459034b2021cSJesper Dangaard Brouer * specific per route level, which is not provided by this helper. 459134b2021cSJesper Dangaard Brouer * For route level MTU checks use the **bpf_fib_lookup**\ () 459234b2021cSJesper Dangaard Brouer * helper. 459334b2021cSJesper Dangaard Brouer * 459434b2021cSJesper Dangaard Brouer * *ctx* is either **struct xdp_md** for XDP programs or 459534b2021cSJesper Dangaard Brouer * **struct sk_buff** for tc cls_act programs. 459634b2021cSJesper Dangaard Brouer * 459734b2021cSJesper Dangaard Brouer * The *flags* argument can be a combination of one or more of the 459834b2021cSJesper Dangaard Brouer * following values: 459934b2021cSJesper Dangaard Brouer * 460034b2021cSJesper Dangaard Brouer * **BPF_MTU_CHK_SEGS** 460134b2021cSJesper Dangaard Brouer * This flag will only works for *ctx* **struct sk_buff**. 460234b2021cSJesper Dangaard Brouer * If packet context contains extra packet segment buffers 460334b2021cSJesper Dangaard Brouer * (often knows as GSO skb), then MTU check is harder to 460434b2021cSJesper Dangaard Brouer * check at this point, because in transmit path it is 460534b2021cSJesper Dangaard Brouer * possible for the skb packet to get re-segmented 460634b2021cSJesper Dangaard Brouer * (depending on net device features). This could still be 460734b2021cSJesper Dangaard Brouer * a MTU violation, so this flag enables performing MTU 460834b2021cSJesper Dangaard Brouer * check against segments, with a different violation 460934b2021cSJesper Dangaard Brouer * return code to tell it apart. Check cannot use len_diff. 461034b2021cSJesper Dangaard Brouer * 461134b2021cSJesper Dangaard Brouer * On return *mtu_len* pointer contains the MTU value of the net 461234b2021cSJesper Dangaard Brouer * device. Remember the net device configured MTU is the L3 size, 461334b2021cSJesper Dangaard Brouer * which is returned here and XDP and TX length operate at L2. 461434b2021cSJesper Dangaard Brouer * Helper take this into account for you, but remember when using 461534b2021cSJesper Dangaard Brouer * MTU value in your BPF-code. On input *mtu_len* must be a valid 461634b2021cSJesper Dangaard Brouer * pointer and be initialized (to zero), else verifier will reject 461734b2021cSJesper Dangaard Brouer * BPF program. 461834b2021cSJesper Dangaard Brouer * 461934b2021cSJesper Dangaard Brouer * Return 462034b2021cSJesper Dangaard Brouer * * 0 on success, and populate MTU value in *mtu_len* pointer. 462134b2021cSJesper Dangaard Brouer * 462234b2021cSJesper Dangaard Brouer * * < 0 if any input argument is invalid (*mtu_len* not updated) 462334b2021cSJesper Dangaard Brouer * 462434b2021cSJesper Dangaard Brouer * MTU violations return positive values, but also populate MTU 462534b2021cSJesper Dangaard Brouer * value in *mtu_len* pointer, as this can be needed for 462634b2021cSJesper Dangaard Brouer * implementing PMTU handing: 462734b2021cSJesper Dangaard Brouer * 462834b2021cSJesper Dangaard Brouer * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 462934b2021cSJesper Dangaard Brouer * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 463034b2021cSJesper Dangaard Brouer * 463169c087baSYonghong Song * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 463269c087baSYonghong Song * Description 463369c087baSYonghong Song * For each element in **map**, call **callback_fn** function with 463469c087baSYonghong Song * **map**, **callback_ctx** and other map-specific parameters. 463569c087baSYonghong Song * The **callback_fn** should be a static function and 463669c087baSYonghong Song * the **callback_ctx** should be a pointer to the stack. 463769c087baSYonghong Song * The **flags** is used to control certain aspects of the helper. 463869c087baSYonghong Song * Currently, the **flags** must be 0. 463969c087baSYonghong Song * 464069c087baSYonghong Song * The following are a list of supported map types and their 464169c087baSYonghong Song * respective expected callback signatures: 464269c087baSYonghong Song * 464369c087baSYonghong Song * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 464469c087baSYonghong Song * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 464569c087baSYonghong Song * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 464669c087baSYonghong Song * 464769c087baSYonghong Song * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 464869c087baSYonghong Song * 464969c087baSYonghong Song * For per_cpu maps, the map_value is the value on the cpu where the 465069c087baSYonghong Song * bpf_prog is running. 465169c087baSYonghong Song * 465269c087baSYonghong Song * If **callback_fn** return 0, the helper will continue to the next 465369c087baSYonghong Song * element. If return value is 1, the helper will skip the rest of 465469c087baSYonghong Song * elements and return. Other return values are not used now. 465569c087baSYonghong Song * 465669c087baSYonghong Song * Return 465769c087baSYonghong Song * The number of traversed map elements for success, **-EINVAL** for 465869c087baSYonghong Song * invalid **flags**. 46597a4b28c6SDaniel Borkmann */ 4660ebb676daSThomas Graf #define __BPF_FUNC_MAPPER(FN) \ 4661ebb676daSThomas Graf FN(unspec), \ 4662ebb676daSThomas Graf FN(map_lookup_elem), \ 4663ebb676daSThomas Graf FN(map_update_elem), \ 4664ebb676daSThomas Graf FN(map_delete_elem), \ 4665ebb676daSThomas Graf FN(probe_read), \ 4666ebb676daSThomas Graf FN(ktime_get_ns), \ 4667ebb676daSThomas Graf FN(trace_printk), \ 4668ebb676daSThomas Graf FN(get_prandom_u32), \ 4669ebb676daSThomas Graf FN(get_smp_processor_id), \ 4670ebb676daSThomas Graf FN(skb_store_bytes), \ 4671ebb676daSThomas Graf FN(l3_csum_replace), \ 4672ebb676daSThomas Graf FN(l4_csum_replace), \ 4673ebb676daSThomas Graf FN(tail_call), \ 4674ebb676daSThomas Graf FN(clone_redirect), \ 4675ebb676daSThomas Graf FN(get_current_pid_tgid), \ 4676ebb676daSThomas Graf FN(get_current_uid_gid), \ 4677ebb676daSThomas Graf FN(get_current_comm), \ 4678ebb676daSThomas Graf FN(get_cgroup_classid), \ 4679ebb676daSThomas Graf FN(skb_vlan_push), \ 4680ebb676daSThomas Graf FN(skb_vlan_pop), \ 4681ebb676daSThomas Graf FN(skb_get_tunnel_key), \ 4682ebb676daSThomas Graf FN(skb_set_tunnel_key), \ 4683ebb676daSThomas Graf FN(perf_event_read), \ 4684ebb676daSThomas Graf FN(redirect), \ 4685ebb676daSThomas Graf FN(get_route_realm), \ 4686ebb676daSThomas Graf FN(perf_event_output), \ 4687ebb676daSThomas Graf FN(skb_load_bytes), \ 4688ebb676daSThomas Graf FN(get_stackid), \ 4689ebb676daSThomas Graf FN(csum_diff), \ 4690ebb676daSThomas Graf FN(skb_get_tunnel_opt), \ 4691ebb676daSThomas Graf FN(skb_set_tunnel_opt), \ 4692ebb676daSThomas Graf FN(skb_change_proto), \ 4693ebb676daSThomas Graf FN(skb_change_type), \ 4694ebb676daSThomas Graf FN(skb_under_cgroup), \ 4695ebb676daSThomas Graf FN(get_hash_recalc), \ 4696ebb676daSThomas Graf FN(get_current_task), \ 4697ebb676daSThomas Graf FN(probe_write_user), \ 4698ebb676daSThomas Graf FN(current_task_under_cgroup), \ 4699ebb676daSThomas Graf FN(skb_change_tail), \ 4700ebb676daSThomas Graf FN(skb_pull_data), \ 4701ebb676daSThomas Graf FN(csum_update), \ 4702ebb676daSThomas Graf FN(set_hash_invalid), \ 47033a0af8fdSThomas Graf FN(get_numa_node_id), \ 470417bedab2SMartin KaFai Lau FN(skb_change_head), \ 4705a5e8c070SGianluca Borello FN(xdp_adjust_head), \ 470691b8270fSChenbo Feng FN(probe_read_str), \ 47076acc5c29SChenbo Feng FN(get_socket_cookie), \ 4708ded092cdSDaniel Borkmann FN(get_socket_uid), \ 47098c4b4c7eSLawrence Brakmo FN(set_hash), \ 47102be7e212SDaniel Borkmann FN(setsockopt), \ 471197f91a7cSJohn Fastabend FN(skb_adjust_room), \ 4712174a79ffSJohn Fastabend FN(redirect_map), \ 4713174a79ffSJohn Fastabend FN(sk_redirect_map), \ 4714174a79ffSJohn Fastabend FN(sock_map_update), \ 4715908432caSYonghong Song FN(xdp_adjust_meta), \ 47164bebdc7aSYonghong Song FN(perf_event_read_value), \ 4717cd86d1fdSLawrence Brakmo FN(perf_prog_read_value), \ 47189802d865SJosef Bacik FN(getsockopt), \ 4719b13d8807SLawrence Brakmo FN(override_return), \ 47204f738adbSJohn Fastabend FN(sock_ops_cb_flags_set), \ 47212a100317SJohn Fastabend FN(msg_redirect_map), \ 472291843d54SJohn Fastabend FN(msg_apply_bytes), \ 4723015632bbSJohn Fastabend FN(msg_cork_bytes), \ 4724d74bad4eSAndrey Ignatov FN(msg_pull_data), \ 4725b32cc5b9SNikita V. Shirokov FN(bind), \ 472612bed760SEyal Birger FN(xdp_adjust_tail), \ 4727c195651eSYonghong Song FN(skb_get_xfrm_state), \ 47284e1ec56cSDaniel Borkmann FN(get_stack), \ 472987f5fc7eSDavid Ahern FN(skb_load_bytes_relative), \ 473081110384SJohn Fastabend FN(fib_lookup), \ 473181110384SJohn Fastabend FN(sock_hash_update), \ 473281110384SJohn Fastabend FN(msg_redirect_hash), \ 4733fe94cc29SMathieu Xhonneux FN(sk_redirect_hash), \ 4734fe94cc29SMathieu Xhonneux FN(lwt_push_encap), \ 4735fe94cc29SMathieu Xhonneux FN(lwt_seg6_store_bytes), \ 4736fe94cc29SMathieu Xhonneux FN(lwt_seg6_adjust_srh), \ 4737f4364dcfSSean Young FN(lwt_seg6_action), \ 4738f4364dcfSSean Young FN(rc_repeat), \ 4739cb20b08eSDaniel Borkmann FN(rc_keydown), \ 4740bf6fa2c8SYonghong Song FN(skb_cgroup_id), \ 4741cd339431SRoman Gushchin FN(get_current_cgroup_id), \ 47422dbb9b9eSMartin KaFai Lau FN(get_local_storage), \ 474377236281SAndrey Ignatov FN(sk_select_reuseport), \ 47446acc9b43SJoe Stringer FN(skb_ancestor_cgroup_id), \ 47456acc9b43SJoe Stringer FN(sk_lookup_tcp), \ 47466acc9b43SJoe Stringer FN(sk_lookup_udp), \ 4747f1a2e44aSMauricio Vasquez B FN(sk_release), \ 4748f1a2e44aSMauricio Vasquez B FN(map_push_elem), \ 4749f1a2e44aSMauricio Vasquez B FN(map_pop_elem), \ 47506fff607eSJohn Fastabend FN(map_peek_elem), \ 47517246d8edSJohn Fastabend FN(msg_push_data), \ 475201d3240aSSean Young FN(msg_pop_data), \ 4753d83525caSAlexei Starovoitov FN(rc_pointer_rel), \ 4754d83525caSAlexei Starovoitov FN(spin_lock), \ 475546f8bc92SMartin KaFai Lau FN(spin_unlock), \ 4756655a51e5SMartin KaFai Lau FN(sk_fullsock), \ 4757f7c917baSbrakmo FN(tcp_sock), \ 4758dbafd7ddSMartin KaFai Lau FN(skb_ecn_set_ce), \ 4759edbf8c01SLorenz Bauer FN(get_listener_sock), \ 476039904084SLorenz Bauer FN(skc_lookup_tcp), \ 4761808649fbSAndrey Ignatov FN(tcp_check_syncookie), \ 47621d11b301SAndrey Ignatov FN(sysctl_get_name), \ 47634e63acdfSAndrey Ignatov FN(sysctl_get_current_value), \ 47644e63acdfSAndrey Ignatov FN(sysctl_get_new_value), \ 4765d7a4cb9bSAndrey Ignatov FN(sysctl_set_new_value), \ 4766d7a4cb9bSAndrey Ignatov FN(strtol), \ 47676ac99e8fSMartin KaFai Lau FN(strtoul), \ 47686ac99e8fSMartin KaFai Lau FN(sk_storage_get), \ 47698b401f9eSYonghong Song FN(sk_storage_delete), \ 477070d66244SPetar Penkov FN(send_signal), \ 4771a7658e1aSAlexei Starovoitov FN(tcp_gen_syncookie), \ 47726ae08ae3SDaniel Borkmann FN(skb_output), \ 47736ae08ae3SDaniel Borkmann FN(probe_read_user), \ 47746ae08ae3SDaniel Borkmann FN(probe_read_kernel), \ 47756ae08ae3SDaniel Borkmann FN(probe_read_user_str), \ 4776206057feSMartin KaFai Lau FN(probe_read_kernel_str), \ 47778482941fSYonghong Song FN(tcp_send_ack), \ 47785576b991SMartin KaFai Lau FN(send_signal_thread), \ 4779fff7b643SDaniel Xu FN(jiffies64), \ 4780b4490c5cSCarlos Neira FN(read_branch_records), \ 4781d831ee84SEelco Chaudron FN(get_ns_current_pid_tgid), \ 4782f318903cSDaniel Borkmann FN(xdp_output), \ 47830f09abd1SDaniel Borkmann FN(get_netns_cookie), \ 4784cf7fbe66SJoe Stringer FN(get_current_ancestor_cgroup_id), \ 478571d19214SMaciej Żenczykowski FN(sk_assign), \ 4786492e639fSYonghong Song FN(ktime_get_boot_ns), \ 4787492e639fSYonghong Song FN(seq_printf), \ 4788f307fa2cSAndrey Ignatov FN(seq_write), \ 4789f307fa2cSAndrey Ignatov FN(sk_cgroup_id), \ 4790457f4436SAndrii Nakryiko FN(sk_ancestor_cgroup_id), \ 4791457f4436SAndrii Nakryiko FN(ringbuf_output), \ 4792457f4436SAndrii Nakryiko FN(ringbuf_reserve), \ 4793457f4436SAndrii Nakryiko FN(ringbuf_submit), \ 4794457f4436SAndrii Nakryiko FN(ringbuf_discard), \ 47957cdec54fSDaniel Borkmann FN(ringbuf_query), \ 4796af7ec138SYonghong Song FN(csum_level), \ 4797478cfbdfSYonghong Song FN(skc_to_tcp6_sock), \ 4798478cfbdfSYonghong Song FN(skc_to_tcp_sock), \ 4799478cfbdfSYonghong Song FN(skc_to_tcp_timewait_sock), \ 48000d4fad3eSYonghong Song FN(skc_to_tcp_request_sock), \ 4801fa28dcb8SSong Liu FN(skc_to_udp6_sock), \ 4802fa28dcb8SSong Liu FN(get_task_stack), \ 48030813a841SMartin KaFai Lau FN(load_hdr_opt), \ 48040813a841SMartin KaFai Lau FN(store_hdr_opt), \ 48058ea63684SKP Singh FN(reserve_hdr_opt), \ 48068ea63684SKP Singh FN(inode_storage_get), \ 48078ea63684SKP Singh FN(inode_storage_delete), \ 48086e22ab9dSJiri Olsa FN(d_path), \ 480907be4c4aSAlexei Starovoitov FN(copy_from_user), \ 4810c4d0bfb4SAlan Maguire FN(snprintf_btf), \ 4811eb411377SAlan Maguire FN(seq_printf_btf), \ 4812b426ce83SDaniel Borkmann FN(skb_cgroup_classid), \ 4813b4ab3141SDaniel Borkmann FN(redirect_neigh), \ 4814b7906b70SAndrii Nakryiko FN(per_cpu_ptr), \ 4815b7906b70SAndrii Nakryiko FN(this_cpu_ptr), \ 48169aa1206eSDaniel Borkmann FN(redirect_peer), \ 48174cf1bc1fSKP Singh FN(task_storage_get), \ 48184cf1bc1fSKP Singh FN(task_storage_delete), \ 48193ca1032aSKP Singh FN(get_current_task_btf), \ 48203f6719c7SKP Singh FN(bprm_opts_set), \ 4821d0551261SDmitrii Banshchikov FN(ktime_get_coarse_ns), \ 482227672f0dSKP Singh FN(ima_inode_hash), \ 48234f19cab7SFlorent Revest FN(sock_from_file), \ 482434b2021cSJesper Dangaard Brouer FN(check_mtu), \ 482569c087baSYonghong Song FN(for_each_map_elem), \ 4826fa28dcb8SSong Liu /* */ 48277a4b28c6SDaniel Borkmann 4828ebb676daSThomas Graf /* integer value in 'imm' field of BPF_CALL instruction selects which helper 4829ebb676daSThomas Graf * function eBPF program intends to call 48302d0e30c3SDaniel Borkmann */ 4831ebb676daSThomas Graf #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 4832ebb676daSThomas Graf enum bpf_func_id { 4833ebb676daSThomas Graf __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 483409756af4SAlexei Starovoitov __BPF_FUNC_MAX_ID, 483509756af4SAlexei Starovoitov }; 4836ebb676daSThomas Graf #undef __BPF_ENUM_FN 483709756af4SAlexei Starovoitov 4838781c53bcSDaniel Borkmann /* All flags used by eBPF helper functions, placed here. */ 4839781c53bcSDaniel Borkmann 4840781c53bcSDaniel Borkmann /* BPF_FUNC_skb_store_bytes flags. */ 48411aae4bddSAndrii Nakryiko enum { 48421aae4bddSAndrii Nakryiko BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 48431aae4bddSAndrii Nakryiko BPF_F_INVALIDATE_HASH = (1ULL << 1), 48441aae4bddSAndrii Nakryiko }; 4845781c53bcSDaniel Borkmann 4846781c53bcSDaniel Borkmann /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 4847781c53bcSDaniel Borkmann * First 4 bits are for passing the header field size. 4848781c53bcSDaniel Borkmann */ 48491aae4bddSAndrii Nakryiko enum { 48501aae4bddSAndrii Nakryiko BPF_F_HDR_FIELD_MASK = 0xfULL, 48511aae4bddSAndrii Nakryiko }; 4852781c53bcSDaniel Borkmann 4853781c53bcSDaniel Borkmann /* BPF_FUNC_l4_csum_replace flags. */ 48541aae4bddSAndrii Nakryiko enum { 48551aae4bddSAndrii Nakryiko BPF_F_PSEUDO_HDR = (1ULL << 4), 48561aae4bddSAndrii Nakryiko BPF_F_MARK_MANGLED_0 = (1ULL << 5), 48571aae4bddSAndrii Nakryiko BPF_F_MARK_ENFORCE = (1ULL << 6), 48581aae4bddSAndrii Nakryiko }; 4859781c53bcSDaniel Borkmann 4860781c53bcSDaniel Borkmann /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 48611aae4bddSAndrii Nakryiko enum { 48621aae4bddSAndrii Nakryiko BPF_F_INGRESS = (1ULL << 0), 48631aae4bddSAndrii Nakryiko }; 4864781c53bcSDaniel Borkmann 4865c6c33454SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 48661aae4bddSAndrii Nakryiko enum { 48671aae4bddSAndrii Nakryiko BPF_F_TUNINFO_IPV6 = (1ULL << 0), 48681aae4bddSAndrii Nakryiko }; 4869c6c33454SDaniel Borkmann 4870c195651eSYonghong Song /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 48711aae4bddSAndrii Nakryiko enum { 48721aae4bddSAndrii Nakryiko BPF_F_SKIP_FIELD_MASK = 0xffULL, 48731aae4bddSAndrii Nakryiko BPF_F_USER_STACK = (1ULL << 8), 4874c195651eSYonghong Song /* flags used by BPF_FUNC_get_stackid only. */ 48751aae4bddSAndrii Nakryiko BPF_F_FAST_STACK_CMP = (1ULL << 9), 48761aae4bddSAndrii Nakryiko BPF_F_REUSE_STACKID = (1ULL << 10), 4877c195651eSYonghong Song /* flags used by BPF_FUNC_get_stack only. */ 48781aae4bddSAndrii Nakryiko BPF_F_USER_BUILD_ID = (1ULL << 11), 48791aae4bddSAndrii Nakryiko }; 4880d5a3b1f6SAlexei Starovoitov 48812da897e5SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key flags. */ 48821aae4bddSAndrii Nakryiko enum { 48831aae4bddSAndrii Nakryiko BPF_F_ZERO_CSUM_TX = (1ULL << 1), 48841aae4bddSAndrii Nakryiko BPF_F_DONT_FRAGMENT = (1ULL << 2), 48851aae4bddSAndrii Nakryiko BPF_F_SEQ_NUMBER = (1ULL << 3), 48861aae4bddSAndrii Nakryiko }; 48872da897e5SDaniel Borkmann 4888908432caSYonghong Song /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 4889908432caSYonghong Song * BPF_FUNC_perf_event_read_value flags. 4890908432caSYonghong Song */ 48911aae4bddSAndrii Nakryiko enum { 48921aae4bddSAndrii Nakryiko BPF_F_INDEX_MASK = 0xffffffffULL, 48931aae4bddSAndrii Nakryiko BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 4894555c8a86SDaniel Borkmann /* BPF_FUNC_perf_event_output for sk_buff input context. */ 48951aae4bddSAndrii Nakryiko BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 48961aae4bddSAndrii Nakryiko }; 48971e33759cSDaniel Borkmann 4898f71c6143SJoe Stringer /* Current network namespace */ 48991aae4bddSAndrii Nakryiko enum { 49001aae4bddSAndrii Nakryiko BPF_F_CURRENT_NETNS = (-1L), 49011aae4bddSAndrii Nakryiko }; 4902f71c6143SJoe Stringer 49037cdec54fSDaniel Borkmann /* BPF_FUNC_csum_level level values. */ 49047cdec54fSDaniel Borkmann enum { 49057cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_QUERY, 49067cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_INC, 49077cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_DEC, 49087cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_RESET, 49097cdec54fSDaniel Borkmann }; 49107cdec54fSDaniel Borkmann 49112278f6ccSWillem de Bruijn /* BPF_FUNC_skb_adjust_room flags. */ 49121aae4bddSAndrii Nakryiko enum { 49131aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 49141aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 49151aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 49161aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 49171aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 4918836e66c2SDaniel Borkmann BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 49191aae4bddSAndrii Nakryiko }; 49202278f6ccSWillem de Bruijn 49211aae4bddSAndrii Nakryiko enum { 49221aae4bddSAndrii Nakryiko BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 49231aae4bddSAndrii Nakryiko BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 49241aae4bddSAndrii Nakryiko }; 492558dfc900SAlan Maguire 492658dfc900SAlan Maguire #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 492758dfc900SAlan Maguire BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 492858dfc900SAlan Maguire << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 4929868d5235SWillem de Bruijn 4930808649fbSAndrey Ignatov /* BPF_FUNC_sysctl_get_name flags. */ 49311aae4bddSAndrii Nakryiko enum { 49321aae4bddSAndrii Nakryiko BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 49331aae4bddSAndrii Nakryiko }; 4934808649fbSAndrey Ignatov 4935f836a56eSKP Singh /* BPF_FUNC_<kernel_obj>_storage_get flags */ 49361aae4bddSAndrii Nakryiko enum { 4937f836a56eSKP Singh BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 4938f836a56eSKP Singh /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 4939f836a56eSKP Singh * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 4940f836a56eSKP Singh */ 4941f836a56eSKP Singh BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 49421aae4bddSAndrii Nakryiko }; 49436ac99e8fSMartin KaFai Lau 4944fff7b643SDaniel Xu /* BPF_FUNC_read_branch_records flags. */ 49451aae4bddSAndrii Nakryiko enum { 49461aae4bddSAndrii Nakryiko BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 49471aae4bddSAndrii Nakryiko }; 4948fff7b643SDaniel Xu 4949457f4436SAndrii Nakryiko /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 4950457f4436SAndrii Nakryiko * BPF_FUNC_bpf_ringbuf_output flags. 4951457f4436SAndrii Nakryiko */ 4952457f4436SAndrii Nakryiko enum { 4953457f4436SAndrii Nakryiko BPF_RB_NO_WAKEUP = (1ULL << 0), 4954457f4436SAndrii Nakryiko BPF_RB_FORCE_WAKEUP = (1ULL << 1), 4955457f4436SAndrii Nakryiko }; 4956457f4436SAndrii Nakryiko 4957457f4436SAndrii Nakryiko /* BPF_FUNC_bpf_ringbuf_query flags */ 4958457f4436SAndrii Nakryiko enum { 4959457f4436SAndrii Nakryiko BPF_RB_AVAIL_DATA = 0, 4960457f4436SAndrii Nakryiko BPF_RB_RING_SIZE = 1, 4961457f4436SAndrii Nakryiko BPF_RB_CONS_POS = 2, 4962457f4436SAndrii Nakryiko BPF_RB_PROD_POS = 3, 4963457f4436SAndrii Nakryiko }; 4964457f4436SAndrii Nakryiko 4965457f4436SAndrii Nakryiko /* BPF ring buffer constants */ 4966457f4436SAndrii Nakryiko enum { 4967457f4436SAndrii Nakryiko BPF_RINGBUF_BUSY_BIT = (1U << 31), 4968457f4436SAndrii Nakryiko BPF_RINGBUF_DISCARD_BIT = (1U << 30), 4969457f4436SAndrii Nakryiko BPF_RINGBUF_HDR_SZ = 8, 4970457f4436SAndrii Nakryiko }; 4971457f4436SAndrii Nakryiko 4972e9ddbb77SJakub Sitnicki /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 4973e9ddbb77SJakub Sitnicki enum { 4974e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 4975e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 4976e9ddbb77SJakub Sitnicki }; 4977e9ddbb77SJakub Sitnicki 49782be7e212SDaniel Borkmann /* Mode for BPF_FUNC_skb_adjust_room helper. */ 49792be7e212SDaniel Borkmann enum bpf_adj_room_mode { 49802be7e212SDaniel Borkmann BPF_ADJ_ROOM_NET, 498114aa3192SWillem de Bruijn BPF_ADJ_ROOM_MAC, 49822be7e212SDaniel Borkmann }; 49832be7e212SDaniel Borkmann 49844e1ec56cSDaniel Borkmann /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 49854e1ec56cSDaniel Borkmann enum bpf_hdr_start_off { 49864e1ec56cSDaniel Borkmann BPF_HDR_START_MAC, 49874e1ec56cSDaniel Borkmann BPF_HDR_START_NET, 49884e1ec56cSDaniel Borkmann }; 49894e1ec56cSDaniel Borkmann 4990fe94cc29SMathieu Xhonneux /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 4991fe94cc29SMathieu Xhonneux enum bpf_lwt_encap_mode { 4992fe94cc29SMathieu Xhonneux BPF_LWT_ENCAP_SEG6, 49933e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_SEG6_INLINE, 49943e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_IP, 4995fe94cc29SMathieu Xhonneux }; 4996fe94cc29SMathieu Xhonneux 49973f6719c7SKP Singh /* Flags for bpf_bprm_opts_set helper */ 49983f6719c7SKP Singh enum { 49993f6719c7SKP Singh BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 50003f6719c7SKP Singh }; 50013f6719c7SKP Singh 5002b7df9adaSDaniel Borkmann #define __bpf_md_ptr(type, name) \ 5003b7df9adaSDaniel Borkmann union { \ 5004b7df9adaSDaniel Borkmann type name; \ 5005b7df9adaSDaniel Borkmann __u64 :64; \ 5006b7df9adaSDaniel Borkmann } __attribute__((aligned(8))) 5007b7df9adaSDaniel Borkmann 50089bac3d6dSAlexei Starovoitov /* user accessible mirror of in-kernel sk_buff. 50099bac3d6dSAlexei Starovoitov * new fields can only be added to the end of this structure 50109bac3d6dSAlexei Starovoitov */ 50119bac3d6dSAlexei Starovoitov struct __sk_buff { 50129bac3d6dSAlexei Starovoitov __u32 len; 50139bac3d6dSAlexei Starovoitov __u32 pkt_type; 50149bac3d6dSAlexei Starovoitov __u32 mark; 50159bac3d6dSAlexei Starovoitov __u32 queue_mapping; 5016c2497395SAlexei Starovoitov __u32 protocol; 5017c2497395SAlexei Starovoitov __u32 vlan_present; 5018c2497395SAlexei Starovoitov __u32 vlan_tci; 501927cd5452SMichal Sekletar __u32 vlan_proto; 5020bcad5718SDaniel Borkmann __u32 priority; 502137e82c2fSAlexei Starovoitov __u32 ingress_ifindex; 502237e82c2fSAlexei Starovoitov __u32 ifindex; 5023d691f9e8SAlexei Starovoitov __u32 tc_index; 5024d691f9e8SAlexei Starovoitov __u32 cb[5]; 5025ba7591d8SDaniel Borkmann __u32 hash; 5026045efa82SDaniel Borkmann __u32 tc_classid; 5027969bf05eSAlexei Starovoitov __u32 data; 5028969bf05eSAlexei Starovoitov __u32 data_end; 5029b1d9fc41SDaniel Borkmann __u32 napi_id; 50308a31db56SJohn Fastabend 5031de8f3a83SDaniel Borkmann /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 50328a31db56SJohn Fastabend __u32 family; 50338a31db56SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 50348a31db56SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 50358a31db56SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 50368a31db56SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 50378a31db56SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 50388a31db56SJohn Fastabend __u32 local_port; /* stored in host byte order */ 5039de8f3a83SDaniel Borkmann /* ... here. */ 5040de8f3a83SDaniel Borkmann 5041de8f3a83SDaniel Borkmann __u32 data_meta; 5042b7df9adaSDaniel Borkmann __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 5043f11216b2SVlad Dumitrescu __u64 tstamp; 5044e3da08d0SPetar Penkov __u32 wire_len; 5045d9ff286aSEric Dumazet __u32 gso_segs; 504646f8bc92SMartin KaFai Lau __bpf_md_ptr(struct bpf_sock *, sk); 5047cf62089bSWillem de Bruijn __u32 gso_size; 50489bac3d6dSAlexei Starovoitov }; 50499bac3d6dSAlexei Starovoitov 5050d3aa45ceSAlexei Starovoitov struct bpf_tunnel_key { 5051d3aa45ceSAlexei Starovoitov __u32 tunnel_id; 5052c6c33454SDaniel Borkmann union { 5053d3aa45ceSAlexei Starovoitov __u32 remote_ipv4; 5054c6c33454SDaniel Borkmann __u32 remote_ipv6[4]; 5055c6c33454SDaniel Borkmann }; 5056c6c33454SDaniel Borkmann __u8 tunnel_tos; 5057c6c33454SDaniel Borkmann __u8 tunnel_ttl; 50581fbc2e0cSDaniel Borkmann __u16 tunnel_ext; /* Padding, future use. */ 50594018ab18SDaniel Borkmann __u32 tunnel_label; 5060d3aa45ceSAlexei Starovoitov }; 5061d3aa45ceSAlexei Starovoitov 506212bed760SEyal Birger /* user accessible mirror of in-kernel xfrm_state. 506312bed760SEyal Birger * new fields can only be added to the end of this structure 506412bed760SEyal Birger */ 506512bed760SEyal Birger struct bpf_xfrm_state { 506612bed760SEyal Birger __u32 reqid; 506712bed760SEyal Birger __u32 spi; /* Stored in network byte order */ 506812bed760SEyal Birger __u16 family; 50691fbc2e0cSDaniel Borkmann __u16 ext; /* Padding, future use. */ 507012bed760SEyal Birger union { 507112bed760SEyal Birger __u32 remote_ipv4; /* Stored in network byte order */ 507212bed760SEyal Birger __u32 remote_ipv6[4]; /* Stored in network byte order */ 507312bed760SEyal Birger }; 507412bed760SEyal Birger }; 507512bed760SEyal Birger 50763a0af8fdSThomas Graf /* Generic BPF return codes which all BPF program types may support. 50773a0af8fdSThomas Graf * The values are binary compatible with their TC_ACT_* counter-part to 50783a0af8fdSThomas Graf * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 50793a0af8fdSThomas Graf * programs. 50803a0af8fdSThomas Graf * 50813a0af8fdSThomas Graf * XDP is handled seprately, see XDP_*. 50823a0af8fdSThomas Graf */ 50833a0af8fdSThomas Graf enum bpf_ret_code { 50843a0af8fdSThomas Graf BPF_OK = 0, 50853a0af8fdSThomas Graf /* 1 reserved */ 50863a0af8fdSThomas Graf BPF_DROP = 2, 50873a0af8fdSThomas Graf /* 3-6 reserved */ 50883a0af8fdSThomas Graf BPF_REDIRECT = 7, 50893e0bd37cSPeter Oskolkov /* >127 are reserved for prog type specific return codes. 50903e0bd37cSPeter Oskolkov * 50913e0bd37cSPeter Oskolkov * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 50923e0bd37cSPeter Oskolkov * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 50933e0bd37cSPeter Oskolkov * changed and should be routed based on its new L3 header. 50943e0bd37cSPeter Oskolkov * (This is an L3 redirect, as opposed to L2 redirect 50953e0bd37cSPeter Oskolkov * represented by BPF_REDIRECT above). 50963e0bd37cSPeter Oskolkov */ 50973e0bd37cSPeter Oskolkov BPF_LWT_REROUTE = 128, 50983a0af8fdSThomas Graf }; 50993a0af8fdSThomas Graf 510061023658SDavid Ahern struct bpf_sock { 510161023658SDavid Ahern __u32 bound_dev_if; 5102aa4c1037SDavid Ahern __u32 family; 5103aa4c1037SDavid Ahern __u32 type; 5104aa4c1037SDavid Ahern __u32 protocol; 5105482dca93SDavid Ahern __u32 mark; 5106482dca93SDavid Ahern __u32 priority; 5107aa65d696SMartin KaFai Lau /* IP address also allows 1 and 2 bytes access */ 5108aa65d696SMartin KaFai Lau __u32 src_ip4; 5109aa65d696SMartin KaFai Lau __u32 src_ip6[4]; 5110aa65d696SMartin KaFai Lau __u32 src_port; /* host byte order */ 5111aa65d696SMartin KaFai Lau __u32 dst_port; /* network byte order */ 5112aa65d696SMartin KaFai Lau __u32 dst_ip4; 5113aa65d696SMartin KaFai Lau __u32 dst_ip6[4]; 5114aa65d696SMartin KaFai Lau __u32 state; 5115c3c16f2eSAmritha Nambiar __s32 rx_queue_mapping; 511661023658SDavid Ahern }; 511761023658SDavid Ahern 5118655a51e5SMartin KaFai Lau struct bpf_tcp_sock { 5119655a51e5SMartin KaFai Lau __u32 snd_cwnd; /* Sending congestion window */ 5120655a51e5SMartin KaFai Lau __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 5121655a51e5SMartin KaFai Lau __u32 rtt_min; 5122655a51e5SMartin KaFai Lau __u32 snd_ssthresh; /* Slow start size threshold */ 5123655a51e5SMartin KaFai Lau __u32 rcv_nxt; /* What we want to receive next */ 5124655a51e5SMartin KaFai Lau __u32 snd_nxt; /* Next sequence we send */ 5125655a51e5SMartin KaFai Lau __u32 snd_una; /* First byte we want an ack for */ 5126655a51e5SMartin KaFai Lau __u32 mss_cache; /* Cached effective mss, not including SACKS */ 5127655a51e5SMartin KaFai Lau __u32 ecn_flags; /* ECN status bits. */ 5128655a51e5SMartin KaFai Lau __u32 rate_delivered; /* saved rate sample: packets delivered */ 5129655a51e5SMartin KaFai Lau __u32 rate_interval_us; /* saved rate sample: time elapsed */ 5130655a51e5SMartin KaFai Lau __u32 packets_out; /* Packets which are "in flight" */ 5131655a51e5SMartin KaFai Lau __u32 retrans_out; /* Retransmitted packets out */ 5132655a51e5SMartin KaFai Lau __u32 total_retrans; /* Total retransmits for entire connection */ 5133655a51e5SMartin KaFai Lau __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 5134655a51e5SMartin KaFai Lau * total number of segments in. 5135655a51e5SMartin KaFai Lau */ 5136655a51e5SMartin KaFai Lau __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 5137655a51e5SMartin KaFai Lau * total number of data segments in. 5138655a51e5SMartin KaFai Lau */ 5139655a51e5SMartin KaFai Lau __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 5140655a51e5SMartin KaFai Lau * The total number of segments sent. 5141655a51e5SMartin KaFai Lau */ 5142655a51e5SMartin KaFai Lau __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 5143655a51e5SMartin KaFai Lau * total number of data segments sent. 5144655a51e5SMartin KaFai Lau */ 5145655a51e5SMartin KaFai Lau __u32 lost_out; /* Lost packets */ 5146655a51e5SMartin KaFai Lau __u32 sacked_out; /* SACK'd packets */ 5147655a51e5SMartin KaFai Lau __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 5148655a51e5SMartin KaFai Lau * sum(delta(rcv_nxt)), or how many bytes 5149655a51e5SMartin KaFai Lau * were acked. 5150655a51e5SMartin KaFai Lau */ 5151655a51e5SMartin KaFai Lau __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 5152655a51e5SMartin KaFai Lau * sum(delta(snd_una)), or how many bytes 5153655a51e5SMartin KaFai Lau * were acked. 5154655a51e5SMartin KaFai Lau */ 51550357746dSStanislav Fomichev __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 51560357746dSStanislav Fomichev * total number of DSACK blocks received 51570357746dSStanislav Fomichev */ 51580357746dSStanislav Fomichev __u32 delivered; /* Total data packets delivered incl. rexmits */ 51590357746dSStanislav Fomichev __u32 delivered_ce; /* Like the above but only ECE marked packets */ 5160c2cb5e82SStanislav Fomichev __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 5161655a51e5SMartin KaFai Lau }; 5162655a51e5SMartin KaFai Lau 51636acc9b43SJoe Stringer struct bpf_sock_tuple { 51646acc9b43SJoe Stringer union { 51656acc9b43SJoe Stringer struct { 51666acc9b43SJoe Stringer __be32 saddr; 51676acc9b43SJoe Stringer __be32 daddr; 51686acc9b43SJoe Stringer __be16 sport; 51696acc9b43SJoe Stringer __be16 dport; 51706acc9b43SJoe Stringer } ipv4; 51716acc9b43SJoe Stringer struct { 51726acc9b43SJoe Stringer __be32 saddr[4]; 51736acc9b43SJoe Stringer __be32 daddr[4]; 51746acc9b43SJoe Stringer __be16 sport; 51756acc9b43SJoe Stringer __be16 dport; 51766acc9b43SJoe Stringer } ipv6; 51776acc9b43SJoe Stringer }; 51786acc9b43SJoe Stringer }; 51796acc9b43SJoe Stringer 5180fada7fdcSJonathan Lemon struct bpf_xdp_sock { 5181fada7fdcSJonathan Lemon __u32 queue_id; 5182fada7fdcSJonathan Lemon }; 5183fada7fdcSJonathan Lemon 518417bedab2SMartin KaFai Lau #define XDP_PACKET_HEADROOM 256 518517bedab2SMartin KaFai Lau 51866a773a15SBrenden Blanco /* User return codes for XDP prog type. 51876a773a15SBrenden Blanco * A valid XDP program must return one of these defined values. All other 51889beb8bedSDaniel Borkmann * return codes are reserved for future use. Unknown return codes will 51899beb8bedSDaniel Borkmann * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 51906a773a15SBrenden Blanco */ 51916a773a15SBrenden Blanco enum xdp_action { 51926a773a15SBrenden Blanco XDP_ABORTED = 0, 51936a773a15SBrenden Blanco XDP_DROP, 51946a773a15SBrenden Blanco XDP_PASS, 51956ce96ca3SBrenden Blanco XDP_TX, 5196814abfabSJohn Fastabend XDP_REDIRECT, 51976a773a15SBrenden Blanco }; 51986a773a15SBrenden Blanco 51996a773a15SBrenden Blanco /* user accessible metadata for XDP packet hook 52006a773a15SBrenden Blanco * new fields must be added to the end of this structure 52016a773a15SBrenden Blanco */ 52026a773a15SBrenden Blanco struct xdp_md { 52036a773a15SBrenden Blanco __u32 data; 52046a773a15SBrenden Blanco __u32 data_end; 5205de8f3a83SDaniel Borkmann __u32 data_meta; 5206daaf24c6SJesper Dangaard Brouer /* Below access go through struct xdp_rxq_info */ 520702dd3291SJesper Dangaard Brouer __u32 ingress_ifindex; /* rxq->dev->ifindex */ 520802dd3291SJesper Dangaard Brouer __u32 rx_queue_index; /* rxq->queue_index */ 520964b59025SDavid Ahern 521064b59025SDavid Ahern __u32 egress_ifindex; /* txq->dev->ifindex */ 52116a773a15SBrenden Blanco }; 52126a773a15SBrenden Blanco 5213281920b7SJesper Dangaard Brouer /* DEVMAP map-value layout 5214281920b7SJesper Dangaard Brouer * 5215281920b7SJesper Dangaard Brouer * The struct data-layout of map-value is a configuration interface. 5216281920b7SJesper Dangaard Brouer * New members can only be added to the end of this structure. 5217281920b7SJesper Dangaard Brouer */ 5218281920b7SJesper Dangaard Brouer struct bpf_devmap_val { 5219281920b7SJesper Dangaard Brouer __u32 ifindex; /* device index */ 5220281920b7SJesper Dangaard Brouer union { 5221281920b7SJesper Dangaard Brouer int fd; /* prog fd on map write */ 5222281920b7SJesper Dangaard Brouer __u32 id; /* prog id on map read */ 5223281920b7SJesper Dangaard Brouer } bpf_prog; 5224281920b7SJesper Dangaard Brouer }; 5225281920b7SJesper Dangaard Brouer 5226644bfe51SLorenzo Bianconi /* CPUMAP map-value layout 5227644bfe51SLorenzo Bianconi * 5228644bfe51SLorenzo Bianconi * The struct data-layout of map-value is a configuration interface. 5229644bfe51SLorenzo Bianconi * New members can only be added to the end of this structure. 5230644bfe51SLorenzo Bianconi */ 5231644bfe51SLorenzo Bianconi struct bpf_cpumap_val { 5232644bfe51SLorenzo Bianconi __u32 qsize; /* queue size to remote target CPU */ 523392164774SLorenzo Bianconi union { 523492164774SLorenzo Bianconi int fd; /* prog fd on map write */ 523592164774SLorenzo Bianconi __u32 id; /* prog id on map read */ 523692164774SLorenzo Bianconi } bpf_prog; 5237644bfe51SLorenzo Bianconi }; 5238644bfe51SLorenzo Bianconi 5239174a79ffSJohn Fastabend enum sk_action { 5240bfa64075SJohn Fastabend SK_DROP = 0, 5241bfa64075SJohn Fastabend SK_PASS, 5242174a79ffSJohn Fastabend }; 5243174a79ffSJohn Fastabend 52444f738adbSJohn Fastabend /* user accessible metadata for SK_MSG packet hook, new fields must 52454f738adbSJohn Fastabend * be added to the end of this structure 52464f738adbSJohn Fastabend */ 52474f738adbSJohn Fastabend struct sk_msg_md { 5248b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 5249b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 5250303def35SJohn Fastabend 5251303def35SJohn Fastabend __u32 family; 5252303def35SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 5253303def35SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 5254303def35SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 5255303def35SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 5256303def35SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 5257303def35SJohn Fastabend __u32 local_port; /* stored in host byte order */ 52583bdbd022SJohn Fastabend __u32 size; /* Total size of sk_msg */ 525913d70f5aSJohn Fastabend 526013d70f5aSJohn Fastabend __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 52614f738adbSJohn Fastabend }; 52624f738adbSJohn Fastabend 52632dbb9b9eSMartin KaFai Lau struct sk_reuseport_md { 52642dbb9b9eSMartin KaFai Lau /* 52652dbb9b9eSMartin KaFai Lau * Start of directly accessible data. It begins from 52662dbb9b9eSMartin KaFai Lau * the tcp/udp header. 52672dbb9b9eSMartin KaFai Lau */ 5268b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 5269b7df9adaSDaniel Borkmann /* End of directly accessible data */ 5270b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 52712dbb9b9eSMartin KaFai Lau /* 52722dbb9b9eSMartin KaFai Lau * Total length of packet (starting from the tcp/udp header). 52732dbb9b9eSMartin KaFai Lau * Note that the directly accessible bytes (data_end - data) 52742dbb9b9eSMartin KaFai Lau * could be less than this "len". Those bytes could be 52752dbb9b9eSMartin KaFai Lau * indirectly read by a helper "bpf_skb_load_bytes()". 52762dbb9b9eSMartin KaFai Lau */ 52772dbb9b9eSMartin KaFai Lau __u32 len; 52782dbb9b9eSMartin KaFai Lau /* 52792dbb9b9eSMartin KaFai Lau * Eth protocol in the mac header (network byte order). e.g. 52802dbb9b9eSMartin KaFai Lau * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 52812dbb9b9eSMartin KaFai Lau */ 52822dbb9b9eSMartin KaFai Lau __u32 eth_protocol; 52832dbb9b9eSMartin KaFai Lau __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 52842dbb9b9eSMartin KaFai Lau __u32 bind_inany; /* Is sock bound to an INANY address? */ 52852dbb9b9eSMartin KaFai Lau __u32 hash; /* A hash of the packet 4 tuples */ 52862dbb9b9eSMartin KaFai Lau }; 52872dbb9b9eSMartin KaFai Lau 52881e270976SMartin KaFai Lau #define BPF_TAG_SIZE 8 52891e270976SMartin KaFai Lau 52901e270976SMartin KaFai Lau struct bpf_prog_info { 52911e270976SMartin KaFai Lau __u32 type; 52921e270976SMartin KaFai Lau __u32 id; 52931e270976SMartin KaFai Lau __u8 tag[BPF_TAG_SIZE]; 52941e270976SMartin KaFai Lau __u32 jited_prog_len; 52951e270976SMartin KaFai Lau __u32 xlated_prog_len; 52961e270976SMartin KaFai Lau __aligned_u64 jited_prog_insns; 52971e270976SMartin KaFai Lau __aligned_u64 xlated_prog_insns; 5298cb4d2b3fSMartin KaFai Lau __u64 load_time; /* ns since boottime */ 5299cb4d2b3fSMartin KaFai Lau __u32 created_by_uid; 5300cb4d2b3fSMartin KaFai Lau __u32 nr_map_ids; 5301cb4d2b3fSMartin KaFai Lau __aligned_u64 map_ids; 5302067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 5303675fc275SJakub Kicinski __u32 ifindex; 5304b85fab0eSJiri Olsa __u32 gpl_compatible:1; 53050472301aSBaruch Siach __u32 :31; /* alignment pad */ 5306675fc275SJakub Kicinski __u64 netns_dev; 5307675fc275SJakub Kicinski __u64 netns_ino; 5308dbecd738SSandipan Das __u32 nr_jited_ksyms; 5309815581c1SSandipan Das __u32 nr_jited_func_lens; 5310dbecd738SSandipan Das __aligned_u64 jited_ksyms; 5311815581c1SSandipan Das __aligned_u64 jited_func_lens; 5312838e9690SYonghong Song __u32 btf_id; 5313838e9690SYonghong Song __u32 func_info_rec_size; 5314838e9690SYonghong Song __aligned_u64 func_info; 531511d8b82dSYonghong Song __u32 nr_func_info; 531611d8b82dSYonghong Song __u32 nr_line_info; 5317c454a46bSMartin KaFai Lau __aligned_u64 line_info; 5318c454a46bSMartin KaFai Lau __aligned_u64 jited_line_info; 531911d8b82dSYonghong Song __u32 nr_jited_line_info; 5320c454a46bSMartin KaFai Lau __u32 line_info_rec_size; 5321c454a46bSMartin KaFai Lau __u32 jited_line_info_rec_size; 5322c872bdb3SSong Liu __u32 nr_prog_tags; 5323c872bdb3SSong Liu __aligned_u64 prog_tags; 53245f8f8b93SAlexei Starovoitov __u64 run_time_ns; 53255f8f8b93SAlexei Starovoitov __u64 run_cnt; 53269ed9e9baSAlexei Starovoitov __u64 recursion_misses; 53271e270976SMartin KaFai Lau } __attribute__((aligned(8))); 53281e270976SMartin KaFai Lau 53291e270976SMartin KaFai Lau struct bpf_map_info { 53301e270976SMartin KaFai Lau __u32 type; 53311e270976SMartin KaFai Lau __u32 id; 53321e270976SMartin KaFai Lau __u32 key_size; 53331e270976SMartin KaFai Lau __u32 value_size; 53341e270976SMartin KaFai Lau __u32 max_entries; 53351e270976SMartin KaFai Lau __u32 map_flags; 5336067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 533752775b33SJakub Kicinski __u32 ifindex; 533885d33df3SMartin KaFai Lau __u32 btf_vmlinux_value_type_id; 533952775b33SJakub Kicinski __u64 netns_dev; 534052775b33SJakub Kicinski __u64 netns_ino; 534178958fcaSMartin KaFai Lau __u32 btf_id; 53429b2cf328SMartin KaFai Lau __u32 btf_key_type_id; 53439b2cf328SMartin KaFai Lau __u32 btf_value_type_id; 53441e270976SMartin KaFai Lau } __attribute__((aligned(8))); 53451e270976SMartin KaFai Lau 534662dab84cSMartin KaFai Lau struct bpf_btf_info { 534762dab84cSMartin KaFai Lau __aligned_u64 btf; 534862dab84cSMartin KaFai Lau __u32 btf_size; 534962dab84cSMartin KaFai Lau __u32 id; 535053297220SAndrii Nakryiko __aligned_u64 name; 535153297220SAndrii Nakryiko __u32 name_len; 535253297220SAndrii Nakryiko __u32 kernel_btf; 535362dab84cSMartin KaFai Lau } __attribute__((aligned(8))); 535462dab84cSMartin KaFai Lau 5355f2e10bffSAndrii Nakryiko struct bpf_link_info { 5356f2e10bffSAndrii Nakryiko __u32 type; 5357f2e10bffSAndrii Nakryiko __u32 id; 5358f2e10bffSAndrii Nakryiko __u32 prog_id; 5359f2e10bffSAndrii Nakryiko union { 5360f2e10bffSAndrii Nakryiko struct { 5361f2e10bffSAndrii Nakryiko __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 5362f2e10bffSAndrii Nakryiko __u32 tp_name_len; /* in/out: tp_name buffer len */ 5363f2e10bffSAndrii Nakryiko } raw_tracepoint; 5364f2e10bffSAndrii Nakryiko struct { 5365f2e10bffSAndrii Nakryiko __u32 attach_type; 5366f2e10bffSAndrii Nakryiko } tracing; 5367f2e10bffSAndrii Nakryiko struct { 5368f2e10bffSAndrii Nakryiko __u64 cgroup_id; 5369f2e10bffSAndrii Nakryiko __u32 attach_type; 5370f2e10bffSAndrii Nakryiko } cgroup; 53717f045a49SJakub Sitnicki struct { 53726b0a249aSYonghong Song __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 53736b0a249aSYonghong Song __u32 target_name_len; /* in/out: target_name buffer len */ 53746b0a249aSYonghong Song union { 5375b0c9eb37SYonghong Song struct { 53766b0a249aSYonghong Song __u32 map_id; 53776b0a249aSYonghong Song } map; 5378b0c9eb37SYonghong Song }; 53796b0a249aSYonghong Song } iter; 53806b0a249aSYonghong Song struct { 53817f045a49SJakub Sitnicki __u32 netns_ino; 53827f045a49SJakub Sitnicki __u32 attach_type; 53837f045a49SJakub Sitnicki } netns; 5384c1931c97SAndrii Nakryiko struct { 5385c1931c97SAndrii Nakryiko __u32 ifindex; 5386c1931c97SAndrii Nakryiko } xdp; 5387f2e10bffSAndrii Nakryiko }; 5388f2e10bffSAndrii Nakryiko } __attribute__((aligned(8))); 5389f2e10bffSAndrii Nakryiko 53904fbac77dSAndrey Ignatov /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 53914fbac77dSAndrey Ignatov * by user and intended to be used by socket (e.g. to bind to, depends on 5392bfdfa517SRandy Dunlap * attach type). 53934fbac77dSAndrey Ignatov */ 53944fbac77dSAndrey Ignatov struct bpf_sock_addr { 53954fbac77dSAndrey Ignatov __u32 user_family; /* Allows 4-byte read, but no write. */ 53964fbac77dSAndrey Ignatov __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 53974fbac77dSAndrey Ignatov * Stored in network byte order. 53984fbac77dSAndrey Ignatov */ 5399d4ecfeb1SStanislav Fomichev __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 54004fbac77dSAndrey Ignatov * Stored in network byte order. 54014fbac77dSAndrey Ignatov */ 54027aebfa1bSAndrey Ignatov __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 54034fbac77dSAndrey Ignatov * Stored in network byte order 54044fbac77dSAndrey Ignatov */ 54054fbac77dSAndrey Ignatov __u32 family; /* Allows 4-byte read, but no write */ 54064fbac77dSAndrey Ignatov __u32 type; /* Allows 4-byte read, but no write */ 54074fbac77dSAndrey Ignatov __u32 protocol; /* Allows 4-byte read, but no write */ 5408600c70baSStanislav Fomichev __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 54091cedee13SAndrey Ignatov * Stored in network byte order. 54101cedee13SAndrey Ignatov */ 5411d4ecfeb1SStanislav Fomichev __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 54121cedee13SAndrey Ignatov * Stored in network byte order. 54131cedee13SAndrey Ignatov */ 5414fb85c4a7SStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 54154fbac77dSAndrey Ignatov }; 54164fbac77dSAndrey Ignatov 541740304b2aSLawrence Brakmo /* User bpf_sock_ops struct to access socket values and specify request ops 541840304b2aSLawrence Brakmo * and their replies. 541940304b2aSLawrence Brakmo * Some of this fields are in network (bigendian) byte order and may need 542040304b2aSLawrence Brakmo * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 542140304b2aSLawrence Brakmo * New fields can only be added at the end of this structure 542240304b2aSLawrence Brakmo */ 542340304b2aSLawrence Brakmo struct bpf_sock_ops { 542440304b2aSLawrence Brakmo __u32 op; 542540304b2aSLawrence Brakmo union { 5426de525be2SLawrence Brakmo __u32 args[4]; /* Optionally passed to bpf program */ 5427de525be2SLawrence Brakmo __u32 reply; /* Returned by bpf program */ 5428de525be2SLawrence Brakmo __u32 replylong[4]; /* Optionally returned by bpf prog */ 542940304b2aSLawrence Brakmo }; 543040304b2aSLawrence Brakmo __u32 family; 543140304b2aSLawrence Brakmo __u32 remote_ip4; /* Stored in network byte order */ 543240304b2aSLawrence Brakmo __u32 local_ip4; /* Stored in network byte order */ 543340304b2aSLawrence Brakmo __u32 remote_ip6[4]; /* Stored in network byte order */ 543440304b2aSLawrence Brakmo __u32 local_ip6[4]; /* Stored in network byte order */ 543540304b2aSLawrence Brakmo __u32 remote_port; /* Stored in network byte order */ 543640304b2aSLawrence Brakmo __u32 local_port; /* stored in host byte order */ 5437f19397a5SLawrence Brakmo __u32 is_fullsock; /* Some TCP fields are only valid if 5438f19397a5SLawrence Brakmo * there is a full socket. If not, the 5439f19397a5SLawrence Brakmo * fields read as zero. 5440f19397a5SLawrence Brakmo */ 5441f19397a5SLawrence Brakmo __u32 snd_cwnd; 5442f19397a5SLawrence Brakmo __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 5443b13d8807SLawrence Brakmo __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 544444f0e430SLawrence Brakmo __u32 state; 544544f0e430SLawrence Brakmo __u32 rtt_min; 544644f0e430SLawrence Brakmo __u32 snd_ssthresh; 544744f0e430SLawrence Brakmo __u32 rcv_nxt; 544844f0e430SLawrence Brakmo __u32 snd_nxt; 544944f0e430SLawrence Brakmo __u32 snd_una; 545044f0e430SLawrence Brakmo __u32 mss_cache; 545144f0e430SLawrence Brakmo __u32 ecn_flags; 545244f0e430SLawrence Brakmo __u32 rate_delivered; 545344f0e430SLawrence Brakmo __u32 rate_interval_us; 545444f0e430SLawrence Brakmo __u32 packets_out; 545544f0e430SLawrence Brakmo __u32 retrans_out; 545644f0e430SLawrence Brakmo __u32 total_retrans; 545744f0e430SLawrence Brakmo __u32 segs_in; 545844f0e430SLawrence Brakmo __u32 data_segs_in; 545944f0e430SLawrence Brakmo __u32 segs_out; 546044f0e430SLawrence Brakmo __u32 data_segs_out; 546144f0e430SLawrence Brakmo __u32 lost_out; 546244f0e430SLawrence Brakmo __u32 sacked_out; 546344f0e430SLawrence Brakmo __u32 sk_txhash; 546444f0e430SLawrence Brakmo __u64 bytes_received; 546544f0e430SLawrence Brakmo __u64 bytes_acked; 54661314ef56SStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 54670813a841SMartin KaFai Lau /* [skb_data, skb_data_end) covers the whole TCP header. 54680813a841SMartin KaFai Lau * 54690813a841SMartin KaFai Lau * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 54700813a841SMartin KaFai Lau * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 54710813a841SMartin KaFai Lau * header has not been written. 54720813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 54730813a841SMartin KaFai Lau * been written so far. 54740813a841SMartin KaFai Lau * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 54750813a841SMartin KaFai Lau * the 3WHS. 54760813a841SMartin KaFai Lau * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 54770813a841SMartin KaFai Lau * the 3WHS. 54780813a841SMartin KaFai Lau * 54790813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also be used to read a particular option. 54800813a841SMartin KaFai Lau */ 54810813a841SMartin KaFai Lau __bpf_md_ptr(void *, skb_data); 54820813a841SMartin KaFai Lau __bpf_md_ptr(void *, skb_data_end); 54830813a841SMartin KaFai Lau __u32 skb_len; /* The total length of a packet. 54840813a841SMartin KaFai Lau * It includes the header, options, 54850813a841SMartin KaFai Lau * and payload. 54860813a841SMartin KaFai Lau */ 54870813a841SMartin KaFai Lau __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 54880813a841SMartin KaFai Lau * an easy way to check for tcp_flags 54890813a841SMartin KaFai Lau * without parsing skb_data. 54900813a841SMartin KaFai Lau * 54910813a841SMartin KaFai Lau * In particular, the skb_tcp_flags 54920813a841SMartin KaFai Lau * will still be available in 54930813a841SMartin KaFai Lau * BPF_SOCK_OPS_HDR_OPT_LEN even though 54940813a841SMartin KaFai Lau * the outgoing header has not 54950813a841SMartin KaFai Lau * been written yet. 54960813a841SMartin KaFai Lau */ 549740304b2aSLawrence Brakmo }; 549840304b2aSLawrence Brakmo 5499b13d8807SLawrence Brakmo /* Definitions for bpf_sock_ops_cb_flags */ 55001aae4bddSAndrii Nakryiko enum { 55011aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 55021aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 55031aae4bddSAndrii Nakryiko BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 55041aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 55050813a841SMartin KaFai Lau /* Call bpf for all received TCP headers. The bpf prog will be 55060813a841SMartin KaFai Lau * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 55070813a841SMartin KaFai Lau * 55080813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 55090813a841SMartin KaFai Lau * for the header option related helpers that will be useful 55100813a841SMartin KaFai Lau * to the bpf programs. 55110813a841SMartin KaFai Lau * 55120813a841SMartin KaFai Lau * It could be used at the client/active side (i.e. connect() side) 55130813a841SMartin KaFai Lau * when the server told it that the server was in syncookie 55140813a841SMartin KaFai Lau * mode and required the active side to resend the bpf-written 55150813a841SMartin KaFai Lau * options. The active side can keep writing the bpf-options until 55160813a841SMartin KaFai Lau * it received a valid packet from the server side to confirm 55170813a841SMartin KaFai Lau * the earlier packet (and options) has been received. The later 55180813a841SMartin KaFai Lau * example patch is using it like this at the active side when the 55190813a841SMartin KaFai Lau * server is in syncookie mode. 55200813a841SMartin KaFai Lau * 55210813a841SMartin KaFai Lau * The bpf prog will usually turn this off in the common cases. 55220813a841SMartin KaFai Lau */ 552300d211a4SMartin KaFai Lau BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 55240813a841SMartin KaFai Lau /* Call bpf when kernel has received a header option that 55250813a841SMartin KaFai Lau * the kernel cannot handle. The bpf prog will be called under 55260813a841SMartin KaFai Lau * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 55270813a841SMartin KaFai Lau * 55280813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 55290813a841SMartin KaFai Lau * for the header option related helpers that will be useful 55300813a841SMartin KaFai Lau * to the bpf programs. 55310813a841SMartin KaFai Lau */ 553200d211a4SMartin KaFai Lau BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 55330813a841SMartin KaFai Lau /* Call bpf when the kernel is writing header options for the 55340813a841SMartin KaFai Lau * outgoing packet. The bpf prog will first be called 55350813a841SMartin KaFai Lau * to reserve space in a skb under 55360813a841SMartin KaFai Lau * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 55370813a841SMartin KaFai Lau * the bpf prog will be called to write the header option(s) 55380813a841SMartin KaFai Lau * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 55390813a841SMartin KaFai Lau * 55400813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 55410813a841SMartin KaFai Lau * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 55420813a841SMartin KaFai Lau * related helpers that will be useful to the bpf programs. 55430813a841SMartin KaFai Lau * 55440813a841SMartin KaFai Lau * The kernel gets its chance to reserve space and write 55450813a841SMartin KaFai Lau * options first before the BPF program does. 55460813a841SMartin KaFai Lau */ 5547331fca43SMartin KaFai Lau BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 55481aae4bddSAndrii Nakryiko /* Mask of all currently supported cb flags */ 5549331fca43SMartin KaFai Lau BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 55501aae4bddSAndrii Nakryiko }; 5551b13d8807SLawrence Brakmo 555240304b2aSLawrence Brakmo /* List of known BPF sock_ops operators. 555340304b2aSLawrence Brakmo * New entries can only be added at the end 555440304b2aSLawrence Brakmo */ 555540304b2aSLawrence Brakmo enum { 555640304b2aSLawrence Brakmo BPF_SOCK_OPS_VOID, 55578550f328SLawrence Brakmo BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 55588550f328SLawrence Brakmo * -1 if default value should be used 55598550f328SLawrence Brakmo */ 556013d3b1ebSLawrence Brakmo BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 556113d3b1ebSLawrence Brakmo * window (in packets) or -1 if default 556213d3b1ebSLawrence Brakmo * value should be used 556313d3b1ebSLawrence Brakmo */ 55649872a4bdSLawrence Brakmo BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 55659872a4bdSLawrence Brakmo * active connection is initialized 55669872a4bdSLawrence Brakmo */ 55679872a4bdSLawrence Brakmo BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 55689872a4bdSLawrence Brakmo * active connection is 55699872a4bdSLawrence Brakmo * established 55709872a4bdSLawrence Brakmo */ 55719872a4bdSLawrence Brakmo BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 55729872a4bdSLawrence Brakmo * passive connection is 55739872a4bdSLawrence Brakmo * established 55749872a4bdSLawrence Brakmo */ 557591b5b21cSLawrence Brakmo BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 557691b5b21cSLawrence Brakmo * needs ECN 557791b5b21cSLawrence Brakmo */ 5578e6546ef6SLawrence Brakmo BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 5579e6546ef6SLawrence Brakmo * based on the path and may be 5580e6546ef6SLawrence Brakmo * dependent on the congestion control 5581e6546ef6SLawrence Brakmo * algorithm. In general it indicates 5582e6546ef6SLawrence Brakmo * a congestion threshold. RTTs above 5583e6546ef6SLawrence Brakmo * this indicate congestion 5584e6546ef6SLawrence Brakmo */ 5585f89013f6SLawrence Brakmo BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 5586f89013f6SLawrence Brakmo * Arg1: value of icsk_retransmits 5587f89013f6SLawrence Brakmo * Arg2: value of icsk_rto 5588f89013f6SLawrence Brakmo * Arg3: whether RTO has expired 5589f89013f6SLawrence Brakmo */ 5590a31ad29eSLawrence Brakmo BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 5591a31ad29eSLawrence Brakmo * Arg1: sequence number of 1st byte 5592a31ad29eSLawrence Brakmo * Arg2: # segments 5593a31ad29eSLawrence Brakmo * Arg3: return value of 5594a31ad29eSLawrence Brakmo * tcp_transmit_skb (0 => success) 5595a31ad29eSLawrence Brakmo */ 5596d4487491SLawrence Brakmo BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 5597d4487491SLawrence Brakmo * Arg1: old_state 5598d4487491SLawrence Brakmo * Arg2: new_state 5599d4487491SLawrence Brakmo */ 5600f333ee0cSAndrey Ignatov BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 5601f333ee0cSAndrey Ignatov * socket transition to LISTEN state. 5602f333ee0cSAndrey Ignatov */ 560323729ff2SStanislav Fomichev BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 560423729ff2SStanislav Fomichev */ 56050813a841SMartin KaFai Lau BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 56060813a841SMartin KaFai Lau * It will be called to handle 56070813a841SMartin KaFai Lau * the packets received at 56080813a841SMartin KaFai Lau * an already established 56090813a841SMartin KaFai Lau * connection. 56100813a841SMartin KaFai Lau * 56110813a841SMartin KaFai Lau * sock_ops->skb_data: 56120813a841SMartin KaFai Lau * Referring to the received skb. 56130813a841SMartin KaFai Lau * It covers the TCP header only. 56140813a841SMartin KaFai Lau * 56150813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also 56160813a841SMartin KaFai Lau * be used to search for a 56170813a841SMartin KaFai Lau * particular option. 56180813a841SMartin KaFai Lau */ 56190813a841SMartin KaFai Lau BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 56200813a841SMartin KaFai Lau * header option later in 56210813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 56220813a841SMartin KaFai Lau * Arg1: bool want_cookie. (in 56230813a841SMartin KaFai Lau * writing SYNACK only) 56240813a841SMartin KaFai Lau * 56250813a841SMartin KaFai Lau * sock_ops->skb_data: 56260813a841SMartin KaFai Lau * Not available because no header has 56270813a841SMartin KaFai Lau * been written yet. 56280813a841SMartin KaFai Lau * 56290813a841SMartin KaFai Lau * sock_ops->skb_tcp_flags: 56300813a841SMartin KaFai Lau * The tcp_flags of the 56310813a841SMartin KaFai Lau * outgoing skb. (e.g. SYN, ACK, FIN). 56320813a841SMartin KaFai Lau * 56330813a841SMartin KaFai Lau * bpf_reserve_hdr_opt() should 56340813a841SMartin KaFai Lau * be used to reserve space. 56350813a841SMartin KaFai Lau */ 56360813a841SMartin KaFai Lau BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 56370813a841SMartin KaFai Lau * Arg1: bool want_cookie. (in 56380813a841SMartin KaFai Lau * writing SYNACK only) 56390813a841SMartin KaFai Lau * 56400813a841SMartin KaFai Lau * sock_ops->skb_data: 56410813a841SMartin KaFai Lau * Referring to the outgoing skb. 56420813a841SMartin KaFai Lau * It covers the TCP header 56430813a841SMartin KaFai Lau * that has already been written 56440813a841SMartin KaFai Lau * by the kernel and the 56450813a841SMartin KaFai Lau * earlier bpf-progs. 56460813a841SMartin KaFai Lau * 56470813a841SMartin KaFai Lau * sock_ops->skb_tcp_flags: 56480813a841SMartin KaFai Lau * The tcp_flags of the outgoing 56490813a841SMartin KaFai Lau * skb. (e.g. SYN, ACK, FIN). 56500813a841SMartin KaFai Lau * 56510813a841SMartin KaFai Lau * bpf_store_hdr_opt() should 56520813a841SMartin KaFai Lau * be used to write the 56530813a841SMartin KaFai Lau * option. 56540813a841SMartin KaFai Lau * 56550813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also 56560813a841SMartin KaFai Lau * be used to search for a 56570813a841SMartin KaFai Lau * particular option that 56580813a841SMartin KaFai Lau * has already been written 56590813a841SMartin KaFai Lau * by the kernel or the 56600813a841SMartin KaFai Lau * earlier bpf-progs. 56610813a841SMartin KaFai Lau */ 5662d4487491SLawrence Brakmo }; 5663d4487491SLawrence Brakmo 5664d4487491SLawrence Brakmo /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 5665d4487491SLawrence Brakmo * changes between the TCP and BPF versions. Ideally this should never happen. 5666d4487491SLawrence Brakmo * If it does, we need to add code to convert them before calling 5667d4487491SLawrence Brakmo * the BPF sock_ops function. 5668d4487491SLawrence Brakmo */ 5669d4487491SLawrence Brakmo enum { 5670d4487491SLawrence Brakmo BPF_TCP_ESTABLISHED = 1, 5671d4487491SLawrence Brakmo BPF_TCP_SYN_SENT, 5672d4487491SLawrence Brakmo BPF_TCP_SYN_RECV, 5673d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT1, 5674d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT2, 5675d4487491SLawrence Brakmo BPF_TCP_TIME_WAIT, 5676d4487491SLawrence Brakmo BPF_TCP_CLOSE, 5677d4487491SLawrence Brakmo BPF_TCP_CLOSE_WAIT, 5678d4487491SLawrence Brakmo BPF_TCP_LAST_ACK, 5679d4487491SLawrence Brakmo BPF_TCP_LISTEN, 5680d4487491SLawrence Brakmo BPF_TCP_CLOSING, /* Now a valid state */ 5681d4487491SLawrence Brakmo BPF_TCP_NEW_SYN_RECV, 5682d4487491SLawrence Brakmo 5683d4487491SLawrence Brakmo BPF_TCP_MAX_STATES /* Leave at the end! */ 568440304b2aSLawrence Brakmo }; 568540304b2aSLawrence Brakmo 56861aae4bddSAndrii Nakryiko enum { 56871aae4bddSAndrii Nakryiko TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 56881aae4bddSAndrii Nakryiko TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 56892b8ee4f0SMartin KaFai Lau TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 5690ca584ba0SMartin KaFai Lau TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 56910813a841SMartin KaFai Lau /* Copy the SYN pkt to optval 56920813a841SMartin KaFai Lau * 56930813a841SMartin KaFai Lau * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 56940813a841SMartin KaFai Lau * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 56950813a841SMartin KaFai Lau * to only getting from the saved_syn. It can either get the 56960813a841SMartin KaFai Lau * syn packet from: 56970813a841SMartin KaFai Lau * 56980813a841SMartin KaFai Lau * 1. the just-received SYN packet (only available when writing the 56990813a841SMartin KaFai Lau * SYNACK). It will be useful when it is not necessary to 57000813a841SMartin KaFai Lau * save the SYN packet for latter use. It is also the only way 57010813a841SMartin KaFai Lau * to get the SYN during syncookie mode because the syn 57020813a841SMartin KaFai Lau * packet cannot be saved during syncookie. 57030813a841SMartin KaFai Lau * 57040813a841SMartin KaFai Lau * OR 57050813a841SMartin KaFai Lau * 57060813a841SMartin KaFai Lau * 2. the earlier saved syn which was done by 57070813a841SMartin KaFai Lau * bpf_setsockopt(TCP_SAVE_SYN). 57080813a841SMartin KaFai Lau * 57090813a841SMartin KaFai Lau * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 57100813a841SMartin KaFai Lau * SYN packet is obtained. 57110813a841SMartin KaFai Lau * 57120813a841SMartin KaFai Lau * If the bpf-prog does not need the IP[46] header, the 57130813a841SMartin KaFai Lau * bpf-prog can avoid parsing the IP header by using 57140813a841SMartin KaFai Lau * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 57150813a841SMartin KaFai Lau * IP[46] and TCP header by using TCP_BPF_SYN_IP. 57160813a841SMartin KaFai Lau * 57170813a841SMartin KaFai Lau * >0: Total number of bytes copied 57180813a841SMartin KaFai Lau * -ENOSPC: Not enough space in optval. Only optlen number of 57190813a841SMartin KaFai Lau * bytes is copied. 57200813a841SMartin KaFai Lau * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 57210813a841SMartin KaFai Lau * is not saved by setsockopt(TCP_SAVE_SYN). 57220813a841SMartin KaFai Lau */ 57230813a841SMartin KaFai Lau TCP_BPF_SYN = 1005, /* Copy the TCP header */ 57240813a841SMartin KaFai Lau TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 5725267cf9faSMartin KaFai Lau TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 57260813a841SMartin KaFai Lau }; 57270813a841SMartin KaFai Lau 57280813a841SMartin KaFai Lau enum { 57290813a841SMartin KaFai Lau BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 57300813a841SMartin KaFai Lau }; 57310813a841SMartin KaFai Lau 57320813a841SMartin KaFai Lau /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 57330813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 57340813a841SMartin KaFai Lau */ 57350813a841SMartin KaFai Lau enum { 57360813a841SMartin KaFai Lau BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 57370813a841SMartin KaFai Lau * total option spaces 57380813a841SMartin KaFai Lau * required for an established 57390813a841SMartin KaFai Lau * sk in order to calculate the 57400813a841SMartin KaFai Lau * MSS. No skb is actually 57410813a841SMartin KaFai Lau * sent. 57420813a841SMartin KaFai Lau */ 57430813a841SMartin KaFai Lau BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 57440813a841SMartin KaFai Lau * when sending a SYN. 57450813a841SMartin KaFai Lau */ 57461aae4bddSAndrii Nakryiko }; 5747fc747810SLawrence Brakmo 5748908432caSYonghong Song struct bpf_perf_event_value { 5749908432caSYonghong Song __u64 counter; 5750908432caSYonghong Song __u64 enabled; 5751908432caSYonghong Song __u64 running; 5752908432caSYonghong Song }; 5753908432caSYonghong Song 57541aae4bddSAndrii Nakryiko enum { 57551aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 57561aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_READ = (1ULL << 1), 57571aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_WRITE = (1ULL << 2), 57581aae4bddSAndrii Nakryiko }; 5759ebc614f6SRoman Gushchin 57601aae4bddSAndrii Nakryiko enum { 57611aae4bddSAndrii Nakryiko BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 57621aae4bddSAndrii Nakryiko BPF_DEVCG_DEV_CHAR = (1ULL << 1), 57631aae4bddSAndrii Nakryiko }; 5764ebc614f6SRoman Gushchin 5765ebc614f6SRoman Gushchin struct bpf_cgroup_dev_ctx { 576606ef0ccbSYonghong Song /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 576706ef0ccbSYonghong Song __u32 access_type; 5768ebc614f6SRoman Gushchin __u32 major; 5769ebc614f6SRoman Gushchin __u32 minor; 5770ebc614f6SRoman Gushchin }; 5771ebc614f6SRoman Gushchin 5772c4f6699dSAlexei Starovoitov struct bpf_raw_tracepoint_args { 5773c4f6699dSAlexei Starovoitov __u64 args[0]; 5774c4f6699dSAlexei Starovoitov }; 5775c4f6699dSAlexei Starovoitov 577687f5fc7eSDavid Ahern /* DIRECT: Skip the FIB rules and go to FIB table associated with device 577787f5fc7eSDavid Ahern * OUTPUT: Do lookup from egress perspective; default is ingress 577887f5fc7eSDavid Ahern */ 57791aae4bddSAndrii Nakryiko enum { 57801aae4bddSAndrii Nakryiko BPF_FIB_LOOKUP_DIRECT = (1U << 0), 57811aae4bddSAndrii Nakryiko BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 57821aae4bddSAndrii Nakryiko }; 578387f5fc7eSDavid Ahern 57844c79579bSDavid Ahern enum { 57854c79579bSDavid Ahern BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 57864c79579bSDavid Ahern BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 57874c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 57884c79579bSDavid Ahern BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 57894c79579bSDavid Ahern BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 57904c79579bSDavid Ahern BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 57914c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 57924c79579bSDavid Ahern BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 57934c79579bSDavid Ahern BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 57944c79579bSDavid Ahern }; 57954c79579bSDavid Ahern 579687f5fc7eSDavid Ahern struct bpf_fib_lookup { 5797fa898d76SDavid Ahern /* input: network family for lookup (AF_INET, AF_INET6) 5798fa898d76SDavid Ahern * output: network family of egress nexthop 5799fa898d76SDavid Ahern */ 5800fa898d76SDavid Ahern __u8 family; 580187f5fc7eSDavid Ahern 580287f5fc7eSDavid Ahern /* set if lookup is to consider L4 data - e.g., FIB rules */ 580387f5fc7eSDavid Ahern __u8 l4_protocol; 580487f5fc7eSDavid Ahern __be16 sport; 580587f5fc7eSDavid Ahern __be16 dport; 580687f5fc7eSDavid Ahern 5807e1850ea9SJesper Dangaard Brouer union { /* used for MTU check */ 5808e1850ea9SJesper Dangaard Brouer /* input to lookup */ 5809e1850ea9SJesper Dangaard Brouer __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 58104c79579bSDavid Ahern 5811e1850ea9SJesper Dangaard Brouer /* output: MTU value */ 5812e1850ea9SJesper Dangaard Brouer __u16 mtu_result; 5813e1850ea9SJesper Dangaard Brouer }; 58144c79579bSDavid Ahern /* input: L3 device index for lookup 58154c79579bSDavid Ahern * output: device index from FIB lookup 58164c79579bSDavid Ahern */ 58174c79579bSDavid Ahern __u32 ifindex; 581887f5fc7eSDavid Ahern 581987f5fc7eSDavid Ahern union { 582087f5fc7eSDavid Ahern /* inputs to lookup */ 582187f5fc7eSDavid Ahern __u8 tos; /* AF_INET */ 5822bd3a08aaSDavid Ahern __be32 flowinfo; /* AF_INET6, flow_label + priority */ 582387f5fc7eSDavid Ahern 5824fa898d76SDavid Ahern /* output: metric of fib result (IPv4/IPv6 only) */ 582587f5fc7eSDavid Ahern __u32 rt_metric; 582687f5fc7eSDavid Ahern }; 582787f5fc7eSDavid Ahern 582887f5fc7eSDavid Ahern union { 582987f5fc7eSDavid Ahern __be32 ipv4_src; 583087f5fc7eSDavid Ahern __u32 ipv6_src[4]; /* in6_addr; network order */ 583187f5fc7eSDavid Ahern }; 583287f5fc7eSDavid Ahern 5833fa898d76SDavid Ahern /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 5834fa898d76SDavid Ahern * network header. output: bpf_fib_lookup sets to gateway address 5835fa898d76SDavid Ahern * if FIB lookup returns gateway route 583687f5fc7eSDavid Ahern */ 583787f5fc7eSDavid Ahern union { 583887f5fc7eSDavid Ahern __be32 ipv4_dst; 583987f5fc7eSDavid Ahern __u32 ipv6_dst[4]; /* in6_addr; network order */ 584087f5fc7eSDavid Ahern }; 584187f5fc7eSDavid Ahern 584287f5fc7eSDavid Ahern /* output */ 584387f5fc7eSDavid Ahern __be16 h_vlan_proto; 584487f5fc7eSDavid Ahern __be16 h_vlan_TCI; 584587f5fc7eSDavid Ahern __u8 smac[6]; /* ETH_ALEN */ 584687f5fc7eSDavid Ahern __u8 dmac[6]; /* ETH_ALEN */ 584787f5fc7eSDavid Ahern }; 584887f5fc7eSDavid Ahern 5849ba452c9eSToke Høiland-Jørgensen struct bpf_redir_neigh { 5850ba452c9eSToke Høiland-Jørgensen /* network family for lookup (AF_INET, AF_INET6) */ 5851ba452c9eSToke Høiland-Jørgensen __u32 nh_family; 5852ba452c9eSToke Høiland-Jørgensen /* network address of nexthop; skips fib lookup to find gateway */ 5853ba452c9eSToke Høiland-Jørgensen union { 5854ba452c9eSToke Høiland-Jørgensen __be32 ipv4_nh; 5855ba452c9eSToke Høiland-Jørgensen __u32 ipv6_nh[4]; /* in6_addr; network order */ 5856ba452c9eSToke Høiland-Jørgensen }; 5857ba452c9eSToke Høiland-Jørgensen }; 5858ba452c9eSToke Høiland-Jørgensen 585934b2021cSJesper Dangaard Brouer /* bpf_check_mtu flags*/ 586034b2021cSJesper Dangaard Brouer enum bpf_check_mtu_flags { 586134b2021cSJesper Dangaard Brouer BPF_MTU_CHK_SEGS = (1U << 0), 586234b2021cSJesper Dangaard Brouer }; 586334b2021cSJesper Dangaard Brouer 586434b2021cSJesper Dangaard Brouer enum bpf_check_mtu_ret { 586534b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 586634b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 586734b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 586834b2021cSJesper Dangaard Brouer }; 586934b2021cSJesper Dangaard Brouer 587041bdc4b4SYonghong Song enum bpf_task_fd_type { 587141bdc4b4SYonghong Song BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 587241bdc4b4SYonghong Song BPF_FD_TYPE_TRACEPOINT, /* tp name */ 587341bdc4b4SYonghong Song BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 587441bdc4b4SYonghong Song BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 587541bdc4b4SYonghong Song BPF_FD_TYPE_UPROBE, /* filename + offset */ 587641bdc4b4SYonghong Song BPF_FD_TYPE_URETPROBE, /* filename + offset */ 587741bdc4b4SYonghong Song }; 587841bdc4b4SYonghong Song 58791aae4bddSAndrii Nakryiko enum { 58801aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 58811aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 58821aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 58831aae4bddSAndrii Nakryiko }; 5884086f9568SStanislav Fomichev 5885d58e468bSPetar Penkov struct bpf_flow_keys { 5886d58e468bSPetar Penkov __u16 nhoff; 5887d58e468bSPetar Penkov __u16 thoff; 5888d58e468bSPetar Penkov __u16 addr_proto; /* ETH_P_* of valid addrs */ 5889d58e468bSPetar Penkov __u8 is_frag; 5890d58e468bSPetar Penkov __u8 is_first_frag; 5891d58e468bSPetar Penkov __u8 is_encap; 5892d58e468bSPetar Penkov __u8 ip_proto; 5893d58e468bSPetar Penkov __be16 n_proto; 5894d58e468bSPetar Penkov __be16 sport; 5895d58e468bSPetar Penkov __be16 dport; 5896d58e468bSPetar Penkov union { 5897d58e468bSPetar Penkov struct { 5898d58e468bSPetar Penkov __be32 ipv4_src; 5899d58e468bSPetar Penkov __be32 ipv4_dst; 5900d58e468bSPetar Penkov }; 5901d58e468bSPetar Penkov struct { 5902d58e468bSPetar Penkov __u32 ipv6_src[4]; /* in6_addr; network order */ 5903d58e468bSPetar Penkov __u32 ipv6_dst[4]; /* in6_addr; network order */ 5904d58e468bSPetar Penkov }; 5905d58e468bSPetar Penkov }; 5906086f9568SStanislav Fomichev __u32 flags; 590771c99e32SStanislav Fomichev __be32 flow_label; 5908d58e468bSPetar Penkov }; 5909d58e468bSPetar Penkov 5910838e9690SYonghong Song struct bpf_func_info { 5911d30d42e0SMartin KaFai Lau __u32 insn_off; 5912838e9690SYonghong Song __u32 type_id; 5913838e9690SYonghong Song }; 5914838e9690SYonghong Song 5915c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 5916c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 5917c454a46bSMartin KaFai Lau 5918c454a46bSMartin KaFai Lau struct bpf_line_info { 5919c454a46bSMartin KaFai Lau __u32 insn_off; 5920c454a46bSMartin KaFai Lau __u32 file_name_off; 5921c454a46bSMartin KaFai Lau __u32 line_off; 5922c454a46bSMartin KaFai Lau __u32 line_col; 5923c454a46bSMartin KaFai Lau }; 5924c454a46bSMartin KaFai Lau 5925d83525caSAlexei Starovoitov struct bpf_spin_lock { 5926d83525caSAlexei Starovoitov __u32 val; 5927d83525caSAlexei Starovoitov }; 59287b146cebSAndrey Ignatov 59297b146cebSAndrey Ignatov struct bpf_sysctl { 59307b146cebSAndrey Ignatov __u32 write; /* Sysctl is being read (= 0) or written (= 1). 59317b146cebSAndrey Ignatov * Allows 1,2,4-byte read, but no write. 59327b146cebSAndrey Ignatov */ 5933e1550bfeSAndrey Ignatov __u32 file_pos; /* Sysctl file position to read from, write to. 5934e1550bfeSAndrey Ignatov * Allows 1,2,4-byte read an 4-byte write. 5935e1550bfeSAndrey Ignatov */ 59367b146cebSAndrey Ignatov }; 59377b146cebSAndrey Ignatov 59380d01da6aSStanislav Fomichev struct bpf_sockopt { 59390d01da6aSStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 59400d01da6aSStanislav Fomichev __bpf_md_ptr(void *, optval); 59410d01da6aSStanislav Fomichev __bpf_md_ptr(void *, optval_end); 59420d01da6aSStanislav Fomichev 59430d01da6aSStanislav Fomichev __s32 level; 59440d01da6aSStanislav Fomichev __s32 optname; 59450d01da6aSStanislav Fomichev __s32 optlen; 59460d01da6aSStanislav Fomichev __s32 retval; 59470d01da6aSStanislav Fomichev }; 59480d01da6aSStanislav Fomichev 5949b4490c5cSCarlos Neira struct bpf_pidns_info { 5950b4490c5cSCarlos Neira __u32 pid; 5951b4490c5cSCarlos Neira __u32 tgid; 5952b4490c5cSCarlos Neira }; 5953e9ddbb77SJakub Sitnicki 5954e9ddbb77SJakub Sitnicki /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 5955e9ddbb77SJakub Sitnicki struct bpf_sk_lookup { 5956e9ddbb77SJakub Sitnicki __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 5957e9ddbb77SJakub Sitnicki 5958e9ddbb77SJakub Sitnicki __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 5959e9ddbb77SJakub Sitnicki __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 5960e9ddbb77SJakub Sitnicki __u32 remote_ip4; /* Network byte order */ 5961e9ddbb77SJakub Sitnicki __u32 remote_ip6[4]; /* Network byte order */ 5962e9ddbb77SJakub Sitnicki __u32 remote_port; /* Network byte order */ 5963e9ddbb77SJakub Sitnicki __u32 local_ip4; /* Network byte order */ 5964e9ddbb77SJakub Sitnicki __u32 local_ip6[4]; /* Network byte order */ 5965e9ddbb77SJakub Sitnicki __u32 local_port; /* Host byte order */ 5966e9ddbb77SJakub Sitnicki }; 5967e9ddbb77SJakub Sitnicki 5968c4d0bfb4SAlan Maguire /* 5969c4d0bfb4SAlan Maguire * struct btf_ptr is used for typed pointer representation; the 5970c4d0bfb4SAlan Maguire * type id is used to render the pointer data as the appropriate type 5971c4d0bfb4SAlan Maguire * via the bpf_snprintf_btf() helper described above. A flags field - 5972c4d0bfb4SAlan Maguire * potentially to specify additional details about the BTF pointer 5973c4d0bfb4SAlan Maguire * (rather than its mode of display) - is included for future use. 5974c4d0bfb4SAlan Maguire * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 5975c4d0bfb4SAlan Maguire */ 5976c4d0bfb4SAlan Maguire struct btf_ptr { 5977c4d0bfb4SAlan Maguire void *ptr; 5978c4d0bfb4SAlan Maguire __u32 type_id; 5979c4d0bfb4SAlan Maguire __u32 flags; /* BTF ptr flags; unused at present. */ 5980c4d0bfb4SAlan Maguire }; 5981c4d0bfb4SAlan Maguire 5982c4d0bfb4SAlan Maguire /* 5983c4d0bfb4SAlan Maguire * Flags to control bpf_snprintf_btf() behaviour. 5984c4d0bfb4SAlan Maguire * - BTF_F_COMPACT: no formatting around type information 5985c4d0bfb4SAlan Maguire * - BTF_F_NONAME: no struct/union member names/types 5986c4d0bfb4SAlan Maguire * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 5987c4d0bfb4SAlan Maguire * equivalent to %px. 5988c4d0bfb4SAlan Maguire * - BTF_F_ZERO: show zero-valued struct/union members; they 5989c4d0bfb4SAlan Maguire * are not displayed by default 5990c4d0bfb4SAlan Maguire */ 5991c4d0bfb4SAlan Maguire enum { 5992c4d0bfb4SAlan Maguire BTF_F_COMPACT = (1ULL << 0), 5993c4d0bfb4SAlan Maguire BTF_F_NONAME = (1ULL << 1), 5994c4d0bfb4SAlan Maguire BTF_F_PTR_RAW = (1ULL << 2), 5995c4d0bfb4SAlan Maguire BTF_F_ZERO = (1ULL << 3), 5996c4d0bfb4SAlan Maguire }; 5997c4d0bfb4SAlan Maguire 5998daedfb22SAlexei Starovoitov #endif /* _UAPI__LINUX_BPF_H__ */ 5999