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 * 315f3c45326SJoe Stringer * The sizes of the buffers provided as input and output 316f3c45326SJoe Stringer * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 317f3c45326SJoe Stringer * be provided in the corresponding variables *ctx_size_in*, 318f3c45326SJoe Stringer * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 319f3c45326SJoe Stringer * of these parameters are not provided (ie set to NULL), the 320f3c45326SJoe Stringer * corresponding size field must be zero. 321f3c45326SJoe Stringer * 322f3c45326SJoe Stringer * Some program types have particular requirements: 323f3c45326SJoe Stringer * 324f3c45326SJoe Stringer * **BPF_PROG_TYPE_SK_LOOKUP** 325f3c45326SJoe Stringer * *data_in* and *data_out* must be NULL. 326f3c45326SJoe Stringer * 327f3c45326SJoe Stringer * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 328f3c45326SJoe Stringer * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 329f3c45326SJoe Stringer * 330f3c45326SJoe Stringer * *ctx_out*, *data_in* and *data_out* must be NULL. 331f3c45326SJoe Stringer * *repeat* must be zero. 332f3c45326SJoe Stringer * 333f67c9cbfSJoe Stringer * Return 334f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 335f67c9cbfSJoe Stringer * is set appropriately. 336f67c9cbfSJoe Stringer * 3372a3fdca4SJoe Stringer * **ENOSPC** 3382a3fdca4SJoe Stringer * Either *data_size_out* or *ctx_size_out* is too small. 3392a3fdca4SJoe Stringer * **ENOTSUPP** 3402a3fdca4SJoe Stringer * This command is not supported by the program type of 3412a3fdca4SJoe Stringer * the program referred to by *prog_fd*. 3422a3fdca4SJoe Stringer * 343f67c9cbfSJoe Stringer * BPF_PROG_GET_NEXT_ID 344f67c9cbfSJoe Stringer * Description 345f67c9cbfSJoe Stringer * Fetch the next eBPF program currently loaded into the kernel. 346f67c9cbfSJoe Stringer * 347f67c9cbfSJoe Stringer * Looks for the eBPF program with an id greater than *start_id* 348f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF programs 349f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 350f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 351f67c9cbfSJoe Stringer * 352f67c9cbfSJoe Stringer * Return 353f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 354f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 355f67c9cbfSJoe Stringer * 356f67c9cbfSJoe Stringer * BPF_MAP_GET_NEXT_ID 357f67c9cbfSJoe Stringer * Description 358f67c9cbfSJoe Stringer * Fetch the next eBPF map currently loaded into the kernel. 359f67c9cbfSJoe Stringer * 360f67c9cbfSJoe Stringer * Looks for the eBPF map with an id greater than *start_id* 361f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF maps 362f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 363f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 364f67c9cbfSJoe Stringer * 365f67c9cbfSJoe Stringer * Return 366f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 367f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 368f67c9cbfSJoe Stringer * 369f67c9cbfSJoe Stringer * BPF_PROG_GET_FD_BY_ID 370f67c9cbfSJoe Stringer * Description 371f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF program corresponding to 372f67c9cbfSJoe Stringer * *prog_id*. 373f67c9cbfSJoe Stringer * 374f67c9cbfSJoe Stringer * Return 375f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 376f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 377f67c9cbfSJoe Stringer * 378f67c9cbfSJoe Stringer * BPF_MAP_GET_FD_BY_ID 379f67c9cbfSJoe Stringer * Description 380f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF map corresponding to 381f67c9cbfSJoe Stringer * *map_id*. 382f67c9cbfSJoe Stringer * 383f67c9cbfSJoe Stringer * Return 384f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 385f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 386f67c9cbfSJoe Stringer * 387f67c9cbfSJoe Stringer * BPF_OBJ_GET_INFO_BY_FD 388f67c9cbfSJoe Stringer * Description 389f67c9cbfSJoe Stringer * Obtain information about the eBPF object corresponding to 390f67c9cbfSJoe Stringer * *bpf_fd*. 391f67c9cbfSJoe Stringer * 392f67c9cbfSJoe Stringer * Populates up to *info_len* bytes of *info*, which will be in 393f67c9cbfSJoe Stringer * one of the following formats depending on the eBPF object type 394f67c9cbfSJoe Stringer * of *bpf_fd*: 395f67c9cbfSJoe Stringer * 396f67c9cbfSJoe Stringer * * **struct bpf_prog_info** 397f67c9cbfSJoe Stringer * * **struct bpf_map_info** 398f67c9cbfSJoe Stringer * * **struct bpf_btf_info** 399f67c9cbfSJoe Stringer * * **struct bpf_link_info** 400f67c9cbfSJoe Stringer * 401f67c9cbfSJoe Stringer * Return 402f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 403f67c9cbfSJoe Stringer * is set appropriately. 404f67c9cbfSJoe Stringer * 405f67c9cbfSJoe Stringer * BPF_PROG_QUERY 406f67c9cbfSJoe Stringer * Description 407f67c9cbfSJoe Stringer * Obtain information about eBPF programs associated with the 408f67c9cbfSJoe Stringer * specified *attach_type* hook. 409f67c9cbfSJoe Stringer * 4105d999994SJoe Stringer * The *target_fd* must be a valid file descriptor for a kernel 4115d999994SJoe Stringer * object which depends on the attach type of *attach_bpf_fd*: 4125d999994SJoe Stringer * 4135d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_DEVICE**, 4145d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SKB**, 4155d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK**, 4165d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 4175d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 4185d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 4195d999994SJoe Stringer * **BPF_PROG_TYPE_SOCK_OPS** 4205d999994SJoe Stringer * 4215d999994SJoe Stringer * Control Group v2 hierarchy with the eBPF controller 4225d999994SJoe Stringer * enabled. Requires the kernel to be compiled with 4235d999994SJoe Stringer * **CONFIG_CGROUP_BPF**. 4245d999994SJoe Stringer * 4255d999994SJoe Stringer * **BPF_PROG_TYPE_FLOW_DISSECTOR** 4265d999994SJoe Stringer * 4275d999994SJoe Stringer * Network namespace (eg /proc/self/ns/net). 4285d999994SJoe Stringer * 4295d999994SJoe Stringer * **BPF_PROG_TYPE_LIRC_MODE2** 4305d999994SJoe Stringer * 4315d999994SJoe Stringer * LIRC device path (eg /dev/lircN). Requires the kernel 4325d999994SJoe Stringer * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 4335d999994SJoe Stringer * 4345d999994SJoe Stringer * **BPF_PROG_QUERY** always fetches the number of programs 4355d999994SJoe Stringer * attached and the *attach_flags* which were used to attach those 4365d999994SJoe Stringer * programs. Additionally, if *prog_ids* is nonzero and the number 4375d999994SJoe Stringer * of attached programs is less than *prog_cnt*, populates 4385d999994SJoe Stringer * *prog_ids* with the eBPF program ids of the programs attached 4395d999994SJoe Stringer * at *target_fd*. 4405d999994SJoe Stringer * 4415d999994SJoe Stringer * The following flags may alter the result: 4425d999994SJoe Stringer * 4435d999994SJoe Stringer * **BPF_F_QUERY_EFFECTIVE** 4445d999994SJoe Stringer * Only return information regarding programs which are 4455d999994SJoe Stringer * currently effective at the specified *target_fd*. 4465d999994SJoe Stringer * 447f67c9cbfSJoe Stringer * Return 448f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 449f67c9cbfSJoe Stringer * is set appropriately. 450f67c9cbfSJoe Stringer * 451f67c9cbfSJoe Stringer * BPF_RAW_TRACEPOINT_OPEN 452f67c9cbfSJoe Stringer * Description 453f67c9cbfSJoe Stringer * Attach an eBPF program to a tracepoint *name* to access kernel 454f67c9cbfSJoe Stringer * internal arguments of the tracepoint in their raw form. 455f67c9cbfSJoe Stringer * 456f67c9cbfSJoe Stringer * The *prog_fd* must be a valid file descriptor associated with 457f67c9cbfSJoe Stringer * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 458f67c9cbfSJoe Stringer * 459f67c9cbfSJoe Stringer * No ABI guarantees are made about the content of tracepoint 460f67c9cbfSJoe Stringer * arguments exposed to the corresponding eBPF program. 461f67c9cbfSJoe Stringer * 462f67c9cbfSJoe Stringer * Applying **close**\ (2) to the file descriptor returned by 463f67c9cbfSJoe Stringer * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 464f67c9cbfSJoe Stringer * 465f67c9cbfSJoe Stringer * Return 466f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 467f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 468f67c9cbfSJoe Stringer * 469f67c9cbfSJoe Stringer * BPF_BTF_LOAD 470f67c9cbfSJoe Stringer * Description 471f67c9cbfSJoe Stringer * Verify and load BPF Type Format (BTF) metadata into the kernel, 472f67c9cbfSJoe Stringer * returning a new file descriptor associated with the metadata. 473f67c9cbfSJoe Stringer * BTF is described in more detail at 474f67c9cbfSJoe Stringer * https://www.kernel.org/doc/html/latest/bpf/btf.html. 475f67c9cbfSJoe Stringer * 476f67c9cbfSJoe Stringer * The *btf* parameter must point to valid memory providing 477f67c9cbfSJoe Stringer * *btf_size* bytes of BTF binary metadata. 478f67c9cbfSJoe Stringer * 479f67c9cbfSJoe Stringer * The returned file descriptor can be passed to other **bpf**\ () 480f67c9cbfSJoe Stringer * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 481f67c9cbfSJoe Stringer * associate the BTF with those objects. 482f67c9cbfSJoe Stringer * 483f67c9cbfSJoe Stringer * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 484f67c9cbfSJoe Stringer * parameters to specify a *btf_log_buf*, *btf_log_size* and 485f67c9cbfSJoe Stringer * *btf_log_level* which allow the kernel to return freeform log 486f67c9cbfSJoe Stringer * output regarding the BTF verification process. 487f67c9cbfSJoe Stringer * 488f67c9cbfSJoe Stringer * Return 489f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 490f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 491f67c9cbfSJoe Stringer * 492f67c9cbfSJoe Stringer * BPF_BTF_GET_FD_BY_ID 493f67c9cbfSJoe Stringer * Description 494f67c9cbfSJoe Stringer * Open a file descriptor for the BPF Type Format (BTF) 495f67c9cbfSJoe Stringer * corresponding to *btf_id*. 496f67c9cbfSJoe Stringer * 497f67c9cbfSJoe Stringer * Return 498f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 499f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 500f67c9cbfSJoe Stringer * 501f67c9cbfSJoe Stringer * BPF_TASK_FD_QUERY 502f67c9cbfSJoe Stringer * Description 503f67c9cbfSJoe Stringer * Obtain information about eBPF programs associated with the 504f67c9cbfSJoe Stringer * target process identified by *pid* and *fd*. 505f67c9cbfSJoe Stringer * 506f67c9cbfSJoe Stringer * If the *pid* and *fd* are associated with a tracepoint, kprobe 507f67c9cbfSJoe Stringer * or uprobe perf event, then the *prog_id* and *fd_type* will 508f67c9cbfSJoe Stringer * be populated with the eBPF program id and file descriptor type 509f67c9cbfSJoe Stringer * of type **bpf_task_fd_type**. If associated with a kprobe or 510f67c9cbfSJoe Stringer * uprobe, the *probe_offset* and *probe_addr* will also be 511f67c9cbfSJoe Stringer * populated. Optionally, if *buf* is provided, then up to 512f67c9cbfSJoe Stringer * *buf_len* bytes of *buf* will be populated with the name of 513f67c9cbfSJoe Stringer * the tracepoint, kprobe or uprobe. 514f67c9cbfSJoe Stringer * 515f67c9cbfSJoe Stringer * The resulting *prog_id* may be introspected in deeper detail 516f67c9cbfSJoe Stringer * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 517f67c9cbfSJoe Stringer * 518f67c9cbfSJoe Stringer * Return 519f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 520f67c9cbfSJoe Stringer * is set appropriately. 521f67c9cbfSJoe Stringer * 522f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_AND_DELETE_ELEM 523f67c9cbfSJoe Stringer * Description 524f67c9cbfSJoe Stringer * Look up an element with the given *key* in the map referred to 525f67c9cbfSJoe Stringer * by the file descriptor *fd*, and if found, delete the element. 526f67c9cbfSJoe Stringer * 5273e87f192SDenis Salopek * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 5283e87f192SDenis Salopek * types, the *flags* argument needs to be set to 0, but for other 5293e87f192SDenis Salopek * map types, it may be specified as: 5303e87f192SDenis Salopek * 5313e87f192SDenis Salopek * **BPF_F_LOCK** 5323e87f192SDenis Salopek * Look up and delete the value of a spin-locked map 5333e87f192SDenis Salopek * without returning the lock. This must be specified if 5343e87f192SDenis Salopek * the elements contain a spinlock. 5353e87f192SDenis Salopek * 536f67c9cbfSJoe Stringer * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 537f67c9cbfSJoe Stringer * implement this command as a "pop" operation, deleting the top 538f67c9cbfSJoe Stringer * element rather than one corresponding to *key*. 539f67c9cbfSJoe Stringer * The *key* and *key_len* parameters should be zeroed when 540f67c9cbfSJoe Stringer * issuing this operation for these map types. 541f67c9cbfSJoe Stringer * 542f67c9cbfSJoe Stringer * This command is only valid for the following map types: 543f67c9cbfSJoe Stringer * * **BPF_MAP_TYPE_QUEUE** 544f67c9cbfSJoe Stringer * * **BPF_MAP_TYPE_STACK** 5453e87f192SDenis Salopek * * **BPF_MAP_TYPE_HASH** 5463e87f192SDenis Salopek * * **BPF_MAP_TYPE_PERCPU_HASH** 5473e87f192SDenis Salopek * * **BPF_MAP_TYPE_LRU_HASH** 5483e87f192SDenis Salopek * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 549f67c9cbfSJoe Stringer * 550f67c9cbfSJoe Stringer * Return 551f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 552f67c9cbfSJoe Stringer * is set appropriately. 553f67c9cbfSJoe Stringer * 554f67c9cbfSJoe Stringer * BPF_MAP_FREEZE 555f67c9cbfSJoe Stringer * Description 556f67c9cbfSJoe Stringer * Freeze the permissions of the specified map. 557f67c9cbfSJoe Stringer * 558f67c9cbfSJoe Stringer * Write permissions may be frozen by passing zero *flags*. 559f67c9cbfSJoe Stringer * Upon success, no future syscall invocations may alter the 560f67c9cbfSJoe Stringer * map state of *map_fd*. Write operations from eBPF programs 561f67c9cbfSJoe Stringer * are still possible for a frozen map. 562f67c9cbfSJoe Stringer * 563f67c9cbfSJoe Stringer * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 564f67c9cbfSJoe Stringer * 565f67c9cbfSJoe Stringer * Return 566f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 567f67c9cbfSJoe Stringer * is set appropriately. 568f67c9cbfSJoe Stringer * 569f67c9cbfSJoe Stringer * BPF_BTF_GET_NEXT_ID 570f67c9cbfSJoe Stringer * Description 571f67c9cbfSJoe Stringer * Fetch the next BPF Type Format (BTF) object currently loaded 572f67c9cbfSJoe Stringer * into the kernel. 573f67c9cbfSJoe Stringer * 574f67c9cbfSJoe Stringer * Looks for the BTF object with an id greater than *start_id* 575f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other BTF objects 576f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 577f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 578f67c9cbfSJoe Stringer * 579f67c9cbfSJoe Stringer * Return 580f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 581f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 582f67c9cbfSJoe Stringer * 583f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_BATCH 584f67c9cbfSJoe Stringer * Description 585f67c9cbfSJoe Stringer * Iterate and fetch multiple elements in a map. 586f67c9cbfSJoe Stringer * 5870cb80454SJoe Stringer * Two opaque values are used to manage batch operations, 5880cb80454SJoe Stringer * *in_batch* and *out_batch*. Initially, *in_batch* must be set 5890cb80454SJoe Stringer * to NULL to begin the batched operation. After each subsequent 5900cb80454SJoe Stringer * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 5910cb80454SJoe Stringer * *out_batch* as the *in_batch* for the next operation to 5920cb80454SJoe Stringer * continue iteration from the current point. 5930cb80454SJoe Stringer * 5940cb80454SJoe Stringer * The *keys* and *values* are output parameters which must point 5950cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 5960cb80454SJoe Stringer * and value size of the map *map_fd*. The *keys* buffer must be 5970cb80454SJoe Stringer * of *key_size* * *count*. The *values* buffer must be of 5980cb80454SJoe Stringer * *value_size* * *count*. 5990cb80454SJoe Stringer * 6000cb80454SJoe Stringer * The *elem_flags* argument may be specified as one of the 6010cb80454SJoe Stringer * following: 6020cb80454SJoe Stringer * 6030cb80454SJoe Stringer * **BPF_F_LOCK** 6040cb80454SJoe Stringer * Look up the value of a spin-locked map without 6050cb80454SJoe Stringer * returning the lock. This must be specified if the 6060cb80454SJoe Stringer * elements contain a spinlock. 6070cb80454SJoe Stringer * 6080cb80454SJoe Stringer * On success, *count* elements from the map are copied into the 6090cb80454SJoe Stringer * user buffer, with the keys copied into *keys* and the values 6100cb80454SJoe Stringer * copied into the corresponding indices in *values*. 6110cb80454SJoe Stringer * 6120cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 6130cb80454SJoe Stringer * is set to the number of successfully processed elements. 6140cb80454SJoe Stringer * 615f67c9cbfSJoe Stringer * Return 616f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 617f67c9cbfSJoe Stringer * is set appropriately. 618f67c9cbfSJoe Stringer * 6190cb80454SJoe Stringer * May set *errno* to **ENOSPC** to indicate that *keys* or 6200cb80454SJoe Stringer * *values* is too small to dump an entire bucket during 6210cb80454SJoe Stringer * iteration of a hash-based map type. 6220cb80454SJoe Stringer * 623f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_AND_DELETE_BATCH 624f67c9cbfSJoe Stringer * Description 6250cb80454SJoe Stringer * Iterate and delete all elements in a map. 6260cb80454SJoe Stringer * 6270cb80454SJoe Stringer * This operation has the same behavior as 6280cb80454SJoe Stringer * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 6290cb80454SJoe Stringer * 6300cb80454SJoe Stringer * * Every element that is successfully returned is also deleted 6310cb80454SJoe Stringer * from the map. This is at least *count* elements. Note that 6320cb80454SJoe Stringer * *count* is both an input and an output parameter. 6330cb80454SJoe Stringer * * Upon returning with *errno* set to **EFAULT**, up to 6340cb80454SJoe Stringer * *count* elements may be deleted without returning the keys 6350cb80454SJoe Stringer * and values of the deleted elements. 636f67c9cbfSJoe Stringer * 637f67c9cbfSJoe Stringer * Return 638f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 639f67c9cbfSJoe Stringer * is set appropriately. 640f67c9cbfSJoe Stringer * 641f67c9cbfSJoe Stringer * BPF_MAP_UPDATE_BATCH 642f67c9cbfSJoe Stringer * Description 6430cb80454SJoe Stringer * Update multiple elements in a map by *key*. 6440cb80454SJoe Stringer * 6450cb80454SJoe Stringer * The *keys* and *values* are input parameters which must point 6460cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 6470cb80454SJoe Stringer * and value size of the map *map_fd*. The *keys* buffer must be 6480cb80454SJoe Stringer * of *key_size* * *count*. The *values* buffer must be of 6490cb80454SJoe Stringer * *value_size* * *count*. 6500cb80454SJoe Stringer * 6510cb80454SJoe Stringer * Each element specified in *keys* is sequentially updated to the 6520cb80454SJoe Stringer * value in the corresponding index in *values*. The *in_batch* 6530cb80454SJoe Stringer * and *out_batch* parameters are ignored and should be zeroed. 6540cb80454SJoe Stringer * 6550cb80454SJoe Stringer * The *elem_flags* argument should be specified as one of the 6560cb80454SJoe Stringer * following: 6570cb80454SJoe Stringer * 6580cb80454SJoe Stringer * **BPF_ANY** 6590cb80454SJoe Stringer * Create new elements or update a existing elements. 6600cb80454SJoe Stringer * **BPF_NOEXIST** 6610cb80454SJoe Stringer * Create new elements only if they do not exist. 6620cb80454SJoe Stringer * **BPF_EXIST** 6630cb80454SJoe Stringer * Update existing elements. 6640cb80454SJoe Stringer * **BPF_F_LOCK** 6650cb80454SJoe Stringer * Update spin_lock-ed map elements. This must be 6660cb80454SJoe Stringer * specified if the map value contains a spinlock. 6670cb80454SJoe Stringer * 6680cb80454SJoe Stringer * On success, *count* elements from the map are updated. 6690cb80454SJoe Stringer * 6700cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 6710cb80454SJoe Stringer * is set to the number of successfully processed elements. 672f67c9cbfSJoe Stringer * 673f67c9cbfSJoe Stringer * Return 674f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 675f67c9cbfSJoe Stringer * is set appropriately. 676f67c9cbfSJoe Stringer * 6770cb80454SJoe Stringer * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 6780cb80454SJoe Stringer * **E2BIG**. **E2BIG** indicates that the number of elements in 6790cb80454SJoe Stringer * the map reached the *max_entries* limit specified at map 6800cb80454SJoe Stringer * creation time. 6810cb80454SJoe Stringer * 6820cb80454SJoe Stringer * May set *errno* to one of the following error codes under 6830cb80454SJoe Stringer * specific circumstances: 6840cb80454SJoe Stringer * 6850cb80454SJoe Stringer * **EEXIST** 6860cb80454SJoe Stringer * If *flags* specifies **BPF_NOEXIST** and the element 6870cb80454SJoe Stringer * with *key* already exists in the map. 6880cb80454SJoe Stringer * **ENOENT** 6890cb80454SJoe Stringer * If *flags* specifies **BPF_EXIST** and the element with 6900cb80454SJoe Stringer * *key* does not exist in the map. 6910cb80454SJoe Stringer * 692f67c9cbfSJoe Stringer * BPF_MAP_DELETE_BATCH 693f67c9cbfSJoe Stringer * Description 6940cb80454SJoe Stringer * Delete multiple elements in a map by *key*. 6950cb80454SJoe Stringer * 6960cb80454SJoe Stringer * The *keys* parameter is an input parameter which must point 6970cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 6980cb80454SJoe Stringer * size of the map *map_fd*, that is, *key_size* * *count*. 6990cb80454SJoe Stringer * 7000cb80454SJoe Stringer * Each element specified in *keys* is sequentially deleted. The 7010cb80454SJoe Stringer * *in_batch*, *out_batch*, and *values* parameters are ignored 7020cb80454SJoe Stringer * and should be zeroed. 7030cb80454SJoe Stringer * 7040cb80454SJoe Stringer * The *elem_flags* argument may be specified as one of the 7050cb80454SJoe Stringer * following: 7060cb80454SJoe Stringer * 7070cb80454SJoe Stringer * **BPF_F_LOCK** 7080cb80454SJoe Stringer * Look up the value of a spin-locked map without 7090cb80454SJoe Stringer * returning the lock. This must be specified if the 7100cb80454SJoe Stringer * elements contain a spinlock. 7110cb80454SJoe Stringer * 7120cb80454SJoe Stringer * On success, *count* elements from the map are updated. 7130cb80454SJoe Stringer * 7140cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 7150cb80454SJoe Stringer * is set to the number of successfully processed elements. If 7160cb80454SJoe Stringer * *errno* is **EFAULT**, up to *count* elements may be been 7170cb80454SJoe Stringer * deleted. 718f67c9cbfSJoe Stringer * 719f67c9cbfSJoe Stringer * Return 720f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 721f67c9cbfSJoe Stringer * is set appropriately. 722f67c9cbfSJoe Stringer * 723f67c9cbfSJoe Stringer * BPF_LINK_CREATE 724f67c9cbfSJoe Stringer * Description 725f67c9cbfSJoe Stringer * Attach an eBPF program to a *target_fd* at the specified 726f67c9cbfSJoe Stringer * *attach_type* hook and return a file descriptor handle for 727f67c9cbfSJoe Stringer * managing the link. 728f67c9cbfSJoe Stringer * 729f67c9cbfSJoe Stringer * Return 730f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 731f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 732f67c9cbfSJoe Stringer * 733f67c9cbfSJoe Stringer * BPF_LINK_UPDATE 734f67c9cbfSJoe Stringer * Description 735f67c9cbfSJoe Stringer * Update the eBPF program in the specified *link_fd* to 736f67c9cbfSJoe Stringer * *new_prog_fd*. 737f67c9cbfSJoe Stringer * 738f67c9cbfSJoe Stringer * Return 739f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 740f67c9cbfSJoe Stringer * is set appropriately. 741f67c9cbfSJoe Stringer * 742f67c9cbfSJoe Stringer * BPF_LINK_GET_FD_BY_ID 743f67c9cbfSJoe Stringer * Description 744f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF Link corresponding to 745f67c9cbfSJoe Stringer * *link_id*. 746f67c9cbfSJoe Stringer * 747f67c9cbfSJoe Stringer * Return 748f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 749f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 750f67c9cbfSJoe Stringer * 751f67c9cbfSJoe Stringer * BPF_LINK_GET_NEXT_ID 752f67c9cbfSJoe Stringer * Description 753f67c9cbfSJoe Stringer * Fetch the next eBPF link currently loaded into the kernel. 754f67c9cbfSJoe Stringer * 755f67c9cbfSJoe Stringer * Looks for the eBPF link with an id greater than *start_id* 756f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF links 757f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 758f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 759f67c9cbfSJoe Stringer * 760f67c9cbfSJoe Stringer * Return 761f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 762f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 763f67c9cbfSJoe Stringer * 764f67c9cbfSJoe Stringer * BPF_ENABLE_STATS 765f67c9cbfSJoe Stringer * Description 766f67c9cbfSJoe Stringer * Enable eBPF runtime statistics gathering. 767f67c9cbfSJoe Stringer * 768f67c9cbfSJoe Stringer * Runtime statistics gathering for the eBPF runtime is disabled 769f67c9cbfSJoe Stringer * by default to minimize the corresponding performance overhead. 770f67c9cbfSJoe Stringer * This command enables statistics globally. 771f67c9cbfSJoe Stringer * 772f67c9cbfSJoe Stringer * Multiple programs may independently enable statistics. 773f67c9cbfSJoe Stringer * After gathering the desired statistics, eBPF runtime statistics 774f67c9cbfSJoe Stringer * may be disabled again by calling **close**\ (2) for the file 775f67c9cbfSJoe Stringer * descriptor returned by this function. Statistics will only be 776f67c9cbfSJoe Stringer * disabled system-wide when all outstanding file descriptors 777f67c9cbfSJoe Stringer * returned by prior calls for this subcommand are closed. 778f67c9cbfSJoe Stringer * 779f67c9cbfSJoe Stringer * Return 780f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 781f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 782f67c9cbfSJoe Stringer * 783f67c9cbfSJoe Stringer * BPF_ITER_CREATE 784f67c9cbfSJoe Stringer * Description 785f67c9cbfSJoe Stringer * Create an iterator on top of the specified *link_fd* (as 786f67c9cbfSJoe Stringer * previously created using **BPF_LINK_CREATE**) and return a 787f67c9cbfSJoe Stringer * file descriptor that can be used to trigger the iteration. 788f67c9cbfSJoe Stringer * 789f67c9cbfSJoe Stringer * If the resulting file descriptor is pinned to the filesystem 790f67c9cbfSJoe Stringer * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 791f67c9cbfSJoe Stringer * for that path will trigger the iterator to read kernel state 792f67c9cbfSJoe Stringer * using the eBPF program attached to *link_fd*. 793f67c9cbfSJoe Stringer * 794f67c9cbfSJoe Stringer * Return 795f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 796f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 797f67c9cbfSJoe Stringer * 798f67c9cbfSJoe Stringer * BPF_LINK_DETACH 799f67c9cbfSJoe Stringer * Description 800f67c9cbfSJoe Stringer * Forcefully detach the specified *link_fd* from its 801f67c9cbfSJoe Stringer * corresponding attachment point. 802f67c9cbfSJoe Stringer * 803f67c9cbfSJoe Stringer * Return 804f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 805f67c9cbfSJoe Stringer * is set appropriately. 806f67c9cbfSJoe Stringer * 807f67c9cbfSJoe Stringer * BPF_PROG_BIND_MAP 808f67c9cbfSJoe Stringer * Description 809f67c9cbfSJoe Stringer * Bind a map to the lifetime of an eBPF program. 810f67c9cbfSJoe Stringer * 811f67c9cbfSJoe Stringer * The map identified by *map_fd* is bound to the program 812f67c9cbfSJoe Stringer * identified by *prog_fd* and only released when *prog_fd* is 813f67c9cbfSJoe Stringer * released. This may be used in cases where metadata should be 814f67c9cbfSJoe Stringer * associated with a program which otherwise does not contain any 815f67c9cbfSJoe Stringer * references to the map (for example, embedded in the eBPF 816f67c9cbfSJoe Stringer * program instructions). 817f67c9cbfSJoe Stringer * 818f67c9cbfSJoe Stringer * Return 819f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 820f67c9cbfSJoe Stringer * is set appropriately. 821f67c9cbfSJoe Stringer * 8227799e4d9SJoe Stringer * NOTES 8237799e4d9SJoe Stringer * eBPF objects (maps and programs) can be shared between processes. 8248aacb3c8SJoe Stringer * 8258aacb3c8SJoe Stringer * * After **fork**\ (2), the child inherits file descriptors 8268aacb3c8SJoe Stringer * referring to the same eBPF objects. 8278aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be transferred over 8288aacb3c8SJoe Stringer * **unix**\ (7) domain sockets. 8298aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be duplicated in the 8308aacb3c8SJoe Stringer * usual way, using **dup**\ (2) and similar calls. 8318aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be pinned to the 8328aacb3c8SJoe Stringer * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 8338aacb3c8SJoe Stringer * 8348aacb3c8SJoe Stringer * An eBPF object is deallocated only after all file descriptors referring 8358aacb3c8SJoe Stringer * to the object have been closed and no references remain pinned to the 8368aacb3c8SJoe Stringer * filesystem or attached (for example, bound to a program or device). 8377799e4d9SJoe Stringer */ 83899c55f7dSAlexei Starovoitov enum bpf_cmd { 83999c55f7dSAlexei Starovoitov BPF_MAP_CREATE, 840db20fd2bSAlexei Starovoitov BPF_MAP_LOOKUP_ELEM, 841db20fd2bSAlexei Starovoitov BPF_MAP_UPDATE_ELEM, 842db20fd2bSAlexei Starovoitov BPF_MAP_DELETE_ELEM, 843db20fd2bSAlexei Starovoitov BPF_MAP_GET_NEXT_KEY, 84409756af4SAlexei Starovoitov BPF_PROG_LOAD, 845b2197755SDaniel Borkmann BPF_OBJ_PIN, 846b2197755SDaniel Borkmann BPF_OBJ_GET, 847f4324551SDaniel Mack BPF_PROG_ATTACH, 848f4324551SDaniel Mack BPF_PROG_DETACH, 8491cf1cae9SAlexei Starovoitov BPF_PROG_TEST_RUN, 8505d67f349SAlexei Starovoitov BPF_PROG_RUN = BPF_PROG_TEST_RUN, 85134ad5580SMartin KaFai Lau BPF_PROG_GET_NEXT_ID, 85234ad5580SMartin KaFai Lau BPF_MAP_GET_NEXT_ID, 853b16d9aa4SMartin KaFai Lau BPF_PROG_GET_FD_BY_ID, 854bd5f5f4eSMartin KaFai Lau BPF_MAP_GET_FD_BY_ID, 8551e270976SMartin KaFai Lau BPF_OBJ_GET_INFO_BY_FD, 856468e2f64SAlexei Starovoitov BPF_PROG_QUERY, 857c4f6699dSAlexei Starovoitov BPF_RAW_TRACEPOINT_OPEN, 858f56a653cSMartin KaFai Lau BPF_BTF_LOAD, 85978958fcaSMartin KaFai Lau BPF_BTF_GET_FD_BY_ID, 86041bdc4b4SYonghong Song BPF_TASK_FD_QUERY, 861bd513cd0SMauricio Vasquez B BPF_MAP_LOOKUP_AND_DELETE_ELEM, 86287df15deSDaniel Borkmann BPF_MAP_FREEZE, 8631b9ed84eSQuentin Monnet BPF_BTF_GET_NEXT_ID, 864cb4d03abSBrian Vazquez BPF_MAP_LOOKUP_BATCH, 86505799638SYonghong Song BPF_MAP_LOOKUP_AND_DELETE_BATCH, 866aa2e93b8SBrian Vazquez BPF_MAP_UPDATE_BATCH, 867aa2e93b8SBrian Vazquez BPF_MAP_DELETE_BATCH, 868af6eea57SAndrii Nakryiko BPF_LINK_CREATE, 8690c991ebcSAndrii Nakryiko BPF_LINK_UPDATE, 8702d602c8cSAndrii Nakryiko BPF_LINK_GET_FD_BY_ID, 8712d602c8cSAndrii Nakryiko BPF_LINK_GET_NEXT_ID, 872d46edd67SSong Liu BPF_ENABLE_STATS, 873ac51d99bSYonghong Song BPF_ITER_CREATE, 87473b11c2aSAndrii Nakryiko BPF_LINK_DETACH, 875ef15314aSYiFei Zhu BPF_PROG_BIND_MAP, 87699c55f7dSAlexei Starovoitov }; 87799c55f7dSAlexei Starovoitov 87899c55f7dSAlexei Starovoitov enum bpf_map_type { 87999c55f7dSAlexei Starovoitov BPF_MAP_TYPE_UNSPEC, 8800f8e4bd8SAlexei Starovoitov BPF_MAP_TYPE_HASH, 88128fbcfa0SAlexei Starovoitov BPF_MAP_TYPE_ARRAY, 88204fd61abSAlexei Starovoitov BPF_MAP_TYPE_PROG_ARRAY, 883ea317b26SKaixu Xia BPF_MAP_TYPE_PERF_EVENT_ARRAY, 884824bd0ceSAlexei Starovoitov BPF_MAP_TYPE_PERCPU_HASH, 885a10423b8SAlexei Starovoitov BPF_MAP_TYPE_PERCPU_ARRAY, 886d5a3b1f6SAlexei Starovoitov BPF_MAP_TYPE_STACK_TRACE, 8874ed8ec52SMartin KaFai Lau BPF_MAP_TYPE_CGROUP_ARRAY, 88829ba732aSMartin KaFai Lau BPF_MAP_TYPE_LRU_HASH, 8898f844938SMartin KaFai Lau BPF_MAP_TYPE_LRU_PERCPU_HASH, 890b95a5c4dSDaniel Mack BPF_MAP_TYPE_LPM_TRIE, 89156f668dfSMartin KaFai Lau BPF_MAP_TYPE_ARRAY_OF_MAPS, 892bcc6b1b7SMartin KaFai Lau BPF_MAP_TYPE_HASH_OF_MAPS, 893546ac1ffSJohn Fastabend BPF_MAP_TYPE_DEVMAP, 894174a79ffSJohn Fastabend BPF_MAP_TYPE_SOCKMAP, 8956710e112SJesper Dangaard Brouer BPF_MAP_TYPE_CPUMAP, 896fbfc504aSBjörn Töpel BPF_MAP_TYPE_XSKMAP, 89781110384SJohn Fastabend BPF_MAP_TYPE_SOCKHASH, 898de9cbbaaSRoman Gushchin BPF_MAP_TYPE_CGROUP_STORAGE, 8995dc4c4b7SMartin KaFai Lau BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 900b741f163SRoman Gushchin BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 901f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_QUEUE, 902f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_STACK, 9036ac99e8fSMartin KaFai Lau BPF_MAP_TYPE_SK_STORAGE, 9046f9d451aSToke Høiland-Jørgensen BPF_MAP_TYPE_DEVMAP_HASH, 90585d33df3SMartin KaFai Lau BPF_MAP_TYPE_STRUCT_OPS, 906457f4436SAndrii Nakryiko BPF_MAP_TYPE_RINGBUF, 9078ea63684SKP Singh BPF_MAP_TYPE_INODE_STORAGE, 9084cf1bc1fSKP Singh BPF_MAP_TYPE_TASK_STORAGE, 90999c55f7dSAlexei Starovoitov }; 91099c55f7dSAlexei Starovoitov 9116c4fc209SDaniel Borkmann /* Note that tracing related programs such as 9126c4fc209SDaniel Borkmann * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 9136c4fc209SDaniel Borkmann * are not subject to a stable API since kernel internal data 9146c4fc209SDaniel Borkmann * structures can change from release to release and may 9156c4fc209SDaniel Borkmann * therefore break existing tracing BPF programs. Tracing BPF 9166c4fc209SDaniel Borkmann * programs correspond to /a/ specific kernel which is to be 9176c4fc209SDaniel Borkmann * analyzed, and not /a/ specific kernel /and/ all future ones. 9186c4fc209SDaniel Borkmann */ 91909756af4SAlexei Starovoitov enum bpf_prog_type { 92009756af4SAlexei Starovoitov BPF_PROG_TYPE_UNSPEC, 921ddd872bcSAlexei Starovoitov BPF_PROG_TYPE_SOCKET_FILTER, 9222541517cSAlexei Starovoitov BPF_PROG_TYPE_KPROBE, 92396be4325SDaniel Borkmann BPF_PROG_TYPE_SCHED_CLS, 92494caee8cSDaniel Borkmann BPF_PROG_TYPE_SCHED_ACT, 92598b5c2c6SAlexei Starovoitov BPF_PROG_TYPE_TRACEPOINT, 9266a773a15SBrenden Blanco BPF_PROG_TYPE_XDP, 9270515e599SAlexei Starovoitov BPF_PROG_TYPE_PERF_EVENT, 9280e33661dSDaniel Mack BPF_PROG_TYPE_CGROUP_SKB, 92961023658SDavid Ahern BPF_PROG_TYPE_CGROUP_SOCK, 9303a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_IN, 9313a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_OUT, 9323a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_XMIT, 93340304b2aSLawrence Brakmo BPF_PROG_TYPE_SOCK_OPS, 934b005fd18SJohn Fastabend BPF_PROG_TYPE_SK_SKB, 935ebc614f6SRoman Gushchin BPF_PROG_TYPE_CGROUP_DEVICE, 9364f738adbSJohn Fastabend BPF_PROG_TYPE_SK_MSG, 937c4f6699dSAlexei Starovoitov BPF_PROG_TYPE_RAW_TRACEPOINT, 9384fbac77dSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 939004d4b27SMathieu Xhonneux BPF_PROG_TYPE_LWT_SEG6LOCAL, 940f4364dcfSSean Young BPF_PROG_TYPE_LIRC_MODE2, 9412dbb9b9eSMartin KaFai Lau BPF_PROG_TYPE_SK_REUSEPORT, 942d58e468bSPetar Penkov BPF_PROG_TYPE_FLOW_DISSECTOR, 9437b146cebSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SYSCTL, 9449df1c28bSMatt Mullins BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 9450d01da6aSStanislav Fomichev BPF_PROG_TYPE_CGROUP_SOCKOPT, 946f1b9509cSAlexei Starovoitov BPF_PROG_TYPE_TRACING, 94727ae7997SMartin KaFai Lau BPF_PROG_TYPE_STRUCT_OPS, 948be8704ffSAlexei Starovoitov BPF_PROG_TYPE_EXT, 949fc611f47SKP Singh BPF_PROG_TYPE_LSM, 950e9ddbb77SJakub Sitnicki BPF_PROG_TYPE_SK_LOOKUP, 95179a7f8bdSAlexei Starovoitov BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 95209756af4SAlexei Starovoitov }; 95309756af4SAlexei Starovoitov 9540e33661dSDaniel Mack enum bpf_attach_type { 9550e33661dSDaniel Mack BPF_CGROUP_INET_INGRESS, 9560e33661dSDaniel Mack BPF_CGROUP_INET_EGRESS, 95761023658SDavid Ahern BPF_CGROUP_INET_SOCK_CREATE, 95840304b2aSLawrence Brakmo BPF_CGROUP_SOCK_OPS, 959464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_PARSER, 960464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_VERDICT, 961ebc614f6SRoman Gushchin BPF_CGROUP_DEVICE, 9624f738adbSJohn Fastabend BPF_SK_MSG_VERDICT, 9634fbac77dSAndrey Ignatov BPF_CGROUP_INET4_BIND, 9644fbac77dSAndrey Ignatov BPF_CGROUP_INET6_BIND, 965d74bad4eSAndrey Ignatov BPF_CGROUP_INET4_CONNECT, 966d74bad4eSAndrey Ignatov BPF_CGROUP_INET6_CONNECT, 967aac3fc32SAndrey Ignatov BPF_CGROUP_INET4_POST_BIND, 968aac3fc32SAndrey Ignatov BPF_CGROUP_INET6_POST_BIND, 9691cedee13SAndrey Ignatov BPF_CGROUP_UDP4_SENDMSG, 9701cedee13SAndrey Ignatov BPF_CGROUP_UDP6_SENDMSG, 971f4364dcfSSean Young BPF_LIRC_MODE2, 972d58e468bSPetar Penkov BPF_FLOW_DISSECTOR, 9737b146cebSAndrey Ignatov BPF_CGROUP_SYSCTL, 974983695faSDaniel Borkmann BPF_CGROUP_UDP4_RECVMSG, 975983695faSDaniel Borkmann BPF_CGROUP_UDP6_RECVMSG, 9760d01da6aSStanislav Fomichev BPF_CGROUP_GETSOCKOPT, 9770d01da6aSStanislav Fomichev BPF_CGROUP_SETSOCKOPT, 978f1b9509cSAlexei Starovoitov BPF_TRACE_RAW_TP, 979fec56f58SAlexei Starovoitov BPF_TRACE_FENTRY, 980fec56f58SAlexei Starovoitov BPF_TRACE_FEXIT, 981ae240823SKP Singh BPF_MODIFY_RETURN, 982fc611f47SKP Singh BPF_LSM_MAC, 98315d83c4dSYonghong Song BPF_TRACE_ITER, 9841b66d253SDaniel Borkmann BPF_CGROUP_INET4_GETPEERNAME, 9851b66d253SDaniel Borkmann BPF_CGROUP_INET6_GETPEERNAME, 9861b66d253SDaniel Borkmann BPF_CGROUP_INET4_GETSOCKNAME, 9871b66d253SDaniel Borkmann BPF_CGROUP_INET6_GETSOCKNAME, 988fbee97feSDavid Ahern BPF_XDP_DEVMAP, 989f5836749SStanislav Fomichev BPF_CGROUP_INET_SOCK_RELEASE, 99092164774SLorenzo Bianconi BPF_XDP_CPUMAP, 991e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP, 992aa8d3a71SAndrii Nakryiko BPF_XDP, 993a7ba4558SCong Wang BPF_SK_SKB_VERDICT, 994d5e4ddaeSKuniyuki Iwashima BPF_SK_REUSEPORT_SELECT, 995d5e4ddaeSKuniyuki Iwashima BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 9960e33661dSDaniel Mack __MAX_BPF_ATTACH_TYPE 9970e33661dSDaniel Mack }; 9980e33661dSDaniel Mack 9990e33661dSDaniel Mack #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 10000e33661dSDaniel Mack 1001f2e10bffSAndrii Nakryiko enum bpf_link_type { 1002f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_UNSPEC = 0, 1003f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1004f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_TRACING = 2, 1005f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_CGROUP = 3, 1006de4e05caSYonghong Song BPF_LINK_TYPE_ITER = 4, 10077f045a49SJakub Sitnicki BPF_LINK_TYPE_NETNS = 5, 1008aa8d3a71SAndrii Nakryiko BPF_LINK_TYPE_XDP = 6, 1009f2e10bffSAndrii Nakryiko 1010f2e10bffSAndrii Nakryiko MAX_BPF_LINK_TYPE, 1011f2e10bffSAndrii Nakryiko }; 1012f2e10bffSAndrii Nakryiko 1013324bda9eSAlexei Starovoitov /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1014324bda9eSAlexei Starovoitov * 1015324bda9eSAlexei Starovoitov * NONE(default): No further bpf programs allowed in the subtree. 1016324bda9eSAlexei Starovoitov * 1017324bda9eSAlexei Starovoitov * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1018324bda9eSAlexei Starovoitov * the program in this cgroup yields to sub-cgroup program. 1019324bda9eSAlexei Starovoitov * 1020324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1021324bda9eSAlexei Starovoitov * that cgroup program gets run in addition to the program in this cgroup. 1022324bda9eSAlexei Starovoitov * 1023324bda9eSAlexei Starovoitov * Only one program is allowed to be attached to a cgroup with 1024324bda9eSAlexei Starovoitov * NONE or BPF_F_ALLOW_OVERRIDE flag. 1025324bda9eSAlexei Starovoitov * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1026324bda9eSAlexei Starovoitov * release old program and attach the new one. Attach flags has to match. 1027324bda9eSAlexei Starovoitov * 1028324bda9eSAlexei Starovoitov * Multiple programs are allowed to be attached to a cgroup with 1029324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1030324bda9eSAlexei Starovoitov * (those that were attached first, run first) 1031324bda9eSAlexei Starovoitov * The programs of sub-cgroup are executed first, then programs of 1032324bda9eSAlexei Starovoitov * this cgroup and then programs of parent cgroup. 1033324bda9eSAlexei Starovoitov * When children program makes decision (like picking TCP CA or sock bind) 1034324bda9eSAlexei Starovoitov * parent program has a chance to override it. 1035324bda9eSAlexei Starovoitov * 10367dd68b32SAndrey Ignatov * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 10377dd68b32SAndrey Ignatov * programs for a cgroup. Though it's possible to replace an old program at 10387dd68b32SAndrey Ignatov * any position by also specifying BPF_F_REPLACE flag and position itself in 10397dd68b32SAndrey Ignatov * replace_bpf_fd attribute. Old program at this position will be released. 10407dd68b32SAndrey Ignatov * 1041324bda9eSAlexei Starovoitov * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1042324bda9eSAlexei Starovoitov * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1043324bda9eSAlexei Starovoitov * Ex1: 1044324bda9eSAlexei Starovoitov * cgrp1 (MULTI progs A, B) -> 1045324bda9eSAlexei Starovoitov * cgrp2 (OVERRIDE prog C) -> 1046324bda9eSAlexei Starovoitov * cgrp3 (MULTI prog D) -> 1047324bda9eSAlexei Starovoitov * cgrp4 (OVERRIDE prog E) -> 1048324bda9eSAlexei Starovoitov * cgrp5 (NONE prog F) 1049324bda9eSAlexei Starovoitov * the event in cgrp5 triggers execution of F,D,A,B in that order. 1050324bda9eSAlexei Starovoitov * if prog F is detached, the execution is E,D,A,B 1051324bda9eSAlexei Starovoitov * if prog F and D are detached, the execution is E,A,B 1052324bda9eSAlexei Starovoitov * if prog F, E and D are detached, the execution is C,A,B 1053324bda9eSAlexei Starovoitov * 1054324bda9eSAlexei Starovoitov * All eligible programs are executed regardless of return code from 1055324bda9eSAlexei Starovoitov * earlier programs. 10567f677633SAlexei Starovoitov */ 10577f677633SAlexei Starovoitov #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1058324bda9eSAlexei Starovoitov #define BPF_F_ALLOW_MULTI (1U << 1) 10597dd68b32SAndrey Ignatov #define BPF_F_REPLACE (1U << 2) 10607f677633SAlexei Starovoitov 1061e07b98d9SDavid S. Miller /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1062e07b98d9SDavid S. Miller * verifier will perform strict alignment checking as if the kernel 1063e07b98d9SDavid S. Miller * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1064e07b98d9SDavid S. Miller * and NET_IP_ALIGN defined to 2. 1065e07b98d9SDavid S. Miller */ 1066e07b98d9SDavid S. Miller #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1067e07b98d9SDavid S. Miller 1068e9ee9efcSDavid Miller /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 1069e9ee9efcSDavid Miller * verifier will allow any alignment whatsoever. On platforms 1070e9ee9efcSDavid Miller * with strict alignment requirements for loads ands stores (such 1071e9ee9efcSDavid Miller * as sparc and mips) the verifier validates that all loads and 1072e9ee9efcSDavid Miller * stores provably follow this requirement. This flag turns that 1073e9ee9efcSDavid Miller * checking and enforcement off. 1074e9ee9efcSDavid Miller * 1075e9ee9efcSDavid Miller * It is mostly used for testing when we want to validate the 1076e9ee9efcSDavid Miller * context and memory access aspects of the verifier, but because 1077e9ee9efcSDavid Miller * of an unaligned access the alignment check would trigger before 1078e9ee9efcSDavid Miller * the one we are interested in. 1079e9ee9efcSDavid Miller */ 1080e9ee9efcSDavid Miller #define BPF_F_ANY_ALIGNMENT (1U << 1) 1081e9ee9efcSDavid Miller 1082c240eff6SJiong Wang /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1083c240eff6SJiong Wang * Verifier does sub-register def/use analysis and identifies instructions whose 1084c240eff6SJiong Wang * def only matters for low 32-bit, high 32-bit is never referenced later 1085c240eff6SJiong Wang * through implicit zero extension. Therefore verifier notifies JIT back-ends 1086c240eff6SJiong Wang * that it is safe to ignore clearing high 32-bit for these instructions. This 1087c240eff6SJiong Wang * saves some back-ends a lot of code-gen. However such optimization is not 1088c240eff6SJiong Wang * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1089c240eff6SJiong Wang * hence hasn't used verifier's analysis result. But, we really want to have a 1090c240eff6SJiong Wang * way to be able to verify the correctness of the described optimization on 1091c240eff6SJiong Wang * x86_64 on which testsuites are frequently exercised. 1092c240eff6SJiong Wang * 1093c240eff6SJiong Wang * So, this flag is introduced. Once it is set, verifier will randomize high 1094c240eff6SJiong Wang * 32-bit for those instructions who has been identified as safe to ignore them. 1095c240eff6SJiong Wang * Then, if verifier is not doing correct analysis, such randomization will 1096c240eff6SJiong Wang * regress tests to expose bugs. 1097c240eff6SJiong Wang */ 1098c240eff6SJiong Wang #define BPF_F_TEST_RND_HI32 (1U << 2) 1099c240eff6SJiong Wang 110010d274e8SAlexei Starovoitov /* The verifier internal test flag. Behavior is undefined */ 110110d274e8SAlexei Starovoitov #define BPF_F_TEST_STATE_FREQ (1U << 3) 110210d274e8SAlexei Starovoitov 11031e6c62a8SAlexei Starovoitov /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 11041e6c62a8SAlexei Starovoitov * restrict map and helper usage for such programs. Sleepable BPF programs can 11051e6c62a8SAlexei Starovoitov * only be attached to hooks where kernel execution context allows sleeping. 11061e6c62a8SAlexei Starovoitov * Such programs are allowed to use helpers that may sleep like 11071e6c62a8SAlexei Starovoitov * bpf_copy_from_user(). 11081e6c62a8SAlexei Starovoitov */ 11091e6c62a8SAlexei Starovoitov #define BPF_F_SLEEPABLE (1U << 4) 11101e6c62a8SAlexei Starovoitov 1111d8eca5bbSDaniel Borkmann /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 11124976b718SHao Luo * the following extensions: 1113d8eca5bbSDaniel Borkmann * 1114387544bfSAlexei Starovoitov * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1115387544bfSAlexei Starovoitov * insn[0].imm: map fd or fd_idx 11164976b718SHao Luo * insn[1].imm: 0 11174976b718SHao Luo * insn[0].off: 0 11184976b718SHao Luo * insn[1].off: 0 11194976b718SHao Luo * ldimm64 rewrite: address of map 11204976b718SHao Luo * verifier type: CONST_PTR_TO_MAP 1121d8eca5bbSDaniel Borkmann */ 1122f1a66f85SDaniel Borkmann #define BPF_PSEUDO_MAP_FD 1 1123387544bfSAlexei Starovoitov #define BPF_PSEUDO_MAP_IDX 5 1124387544bfSAlexei Starovoitov 1125387544bfSAlexei Starovoitov /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1126387544bfSAlexei Starovoitov * insn[0].imm: map fd or fd_idx 11274976b718SHao Luo * insn[1].imm: offset into value 11284976b718SHao Luo * insn[0].off: 0 11294976b718SHao Luo * insn[1].off: 0 11304976b718SHao Luo * ldimm64 rewrite: address of map[0]+offset 11314976b718SHao Luo * verifier type: PTR_TO_MAP_VALUE 11324976b718SHao Luo */ 1133d8eca5bbSDaniel Borkmann #define BPF_PSEUDO_MAP_VALUE 2 1134387544bfSAlexei Starovoitov #define BPF_PSEUDO_MAP_IDX_VALUE 6 1135387544bfSAlexei Starovoitov 11364976b718SHao Luo /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 11374976b718SHao Luo * insn[0].imm: kernel btd id of VAR 11384976b718SHao Luo * insn[1].imm: 0 11394976b718SHao Luo * insn[0].off: 0 11404976b718SHao Luo * insn[1].off: 0 11414976b718SHao Luo * ldimm64 rewrite: address of the kernel variable 11424976b718SHao Luo * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 11434976b718SHao Luo * is struct/union. 11444976b718SHao Luo */ 11454976b718SHao Luo #define BPF_PSEUDO_BTF_ID 3 114669c087baSYonghong Song /* insn[0].src_reg: BPF_PSEUDO_FUNC 114769c087baSYonghong Song * insn[0].imm: insn offset to the func 114869c087baSYonghong Song * insn[1].imm: 0 114969c087baSYonghong Song * insn[0].off: 0 115069c087baSYonghong Song * insn[1].off: 0 115169c087baSYonghong Song * ldimm64 rewrite: address of the function 115269c087baSYonghong Song * verifier type: PTR_TO_FUNC. 115369c087baSYonghong Song */ 115469c087baSYonghong Song #define BPF_PSEUDO_FUNC 4 1155f1a66f85SDaniel Borkmann 1156cc8b0b92SAlexei Starovoitov /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1157cc8b0b92SAlexei Starovoitov * offset to another bpf function 1158cc8b0b92SAlexei Starovoitov */ 1159cc8b0b92SAlexei Starovoitov #define BPF_PSEUDO_CALL 1 1160e6ac2450SMartin KaFai Lau /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1161e6ac2450SMartin KaFai Lau * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1162e6ac2450SMartin KaFai Lau */ 1163e6ac2450SMartin KaFai Lau #define BPF_PSEUDO_KFUNC_CALL 2 1164cc8b0b92SAlexei Starovoitov 11653274f520SAlexei Starovoitov /* flags for BPF_MAP_UPDATE_ELEM command */ 11661aae4bddSAndrii Nakryiko enum { 11671aae4bddSAndrii Nakryiko BPF_ANY = 0, /* create new element or update existing */ 11681aae4bddSAndrii Nakryiko BPF_NOEXIST = 1, /* create new element if it didn't exist */ 11691aae4bddSAndrii Nakryiko BPF_EXIST = 2, /* update existing element */ 11701aae4bddSAndrii Nakryiko BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 11711aae4bddSAndrii Nakryiko }; 11723274f520SAlexei Starovoitov 117396eabe7aSMartin KaFai Lau /* flags for BPF_MAP_CREATE command */ 11741aae4bddSAndrii Nakryiko enum { 11751aae4bddSAndrii Nakryiko BPF_F_NO_PREALLOC = (1U << 0), 117629ba732aSMartin KaFai Lau /* Instead of having one common LRU list in the 11778f844938SMartin KaFai Lau * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 117829ba732aSMartin KaFai Lau * which can scale and perform better. 117929ba732aSMartin KaFai Lau * Note, the LRU nodes (including free nodes) cannot be moved 118029ba732aSMartin KaFai Lau * across different LRU lists. 118129ba732aSMartin KaFai Lau */ 11821aae4bddSAndrii Nakryiko BPF_F_NO_COMMON_LRU = (1U << 1), 118396eabe7aSMartin KaFai Lau /* Specify numa node during map creation */ 11841aae4bddSAndrii Nakryiko BPF_F_NUMA_NODE = (1U << 2), 1185cb4d2b3fSMartin KaFai Lau 1186591fe988SDaniel Borkmann /* Flags for accessing BPF object from syscall side. */ 11871aae4bddSAndrii Nakryiko BPF_F_RDONLY = (1U << 3), 11881aae4bddSAndrii Nakryiko BPF_F_WRONLY = (1U << 4), 11896e71b04aSChenbo Feng 1190615755a7SSong Liu /* Flag for stack_map, store build_id+offset instead of pointer */ 11911aae4bddSAndrii Nakryiko BPF_F_STACK_BUILD_ID = (1U << 5), 1192615755a7SSong Liu 119396b3b6c9SLorenz Bauer /* Zero-initialize hash function seed. This should only be used for testing. */ 11941aae4bddSAndrii Nakryiko BPF_F_ZERO_SEED = (1U << 6), 119596b3b6c9SLorenz Bauer 1196591fe988SDaniel Borkmann /* Flags for accessing BPF object from program side. */ 11971aae4bddSAndrii Nakryiko BPF_F_RDONLY_PROG = (1U << 7), 11981aae4bddSAndrii Nakryiko BPF_F_WRONLY_PROG = (1U << 8), 1199591fe988SDaniel Borkmann 12008f51dfc7SStanislav Fomichev /* Clone map from listener for newly accepted socket */ 12011aae4bddSAndrii Nakryiko BPF_F_CLONE = (1U << 9), 12028f51dfc7SStanislav Fomichev 1203fc970227SAndrii Nakryiko /* Enable memory-mapping BPF map */ 12041aae4bddSAndrii Nakryiko BPF_F_MMAPABLE = (1U << 10), 1205792cacccSSong Liu 1206792cacccSSong Liu /* Share perf_event among processes */ 1207792cacccSSong Liu BPF_F_PRESERVE_ELEMS = (1U << 11), 12084a8f87e6SDaniel Borkmann 12094a8f87e6SDaniel Borkmann /* Create a map that is suitable to be an inner map with dynamic max entries */ 12104a8f87e6SDaniel Borkmann BPF_F_INNER_MAP = (1U << 12), 12111aae4bddSAndrii Nakryiko }; 1212fc970227SAndrii Nakryiko 1213f5bfcd95SAndrey Ignatov /* Flags for BPF_PROG_QUERY. */ 1214f5bfcd95SAndrey Ignatov 1215f5bfcd95SAndrey Ignatov /* Query effective (directly attached + inherited from ancestor cgroups) 1216f5bfcd95SAndrey Ignatov * programs that will be executed for events within a cgroup. 1217f5bfcd95SAndrey Ignatov * attach_flags with this flag are returned only for directly attached programs. 1218f5bfcd95SAndrey Ignatov */ 12192f183360SLorenz Bauer #define BPF_F_QUERY_EFFECTIVE (1U << 0) 12202f183360SLorenz Bauer 12211b4d60ecSSong Liu /* Flags for BPF_PROG_TEST_RUN */ 12221b4d60ecSSong Liu 12231b4d60ecSSong Liu /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 12241b4d60ecSSong Liu #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 12251b4d60ecSSong Liu 1226d46edd67SSong Liu /* type for BPF_ENABLE_STATS */ 1227d46edd67SSong Liu enum bpf_stats_type { 1228d46edd67SSong Liu /* enabled run_time_ns and run_cnt */ 1229d46edd67SSong Liu BPF_STATS_RUN_TIME = 0, 1230d46edd67SSong Liu }; 1231d46edd67SSong Liu 1232615755a7SSong Liu enum bpf_stack_build_id_status { 1233615755a7SSong Liu /* user space need an empty entry to identify end of a trace */ 1234615755a7SSong Liu BPF_STACK_BUILD_ID_EMPTY = 0, 1235615755a7SSong Liu /* with valid build_id and offset */ 1236615755a7SSong Liu BPF_STACK_BUILD_ID_VALID = 1, 1237615755a7SSong Liu /* couldn't get build_id, fallback to ip */ 1238615755a7SSong Liu BPF_STACK_BUILD_ID_IP = 2, 1239615755a7SSong Liu }; 1240615755a7SSong Liu 1241615755a7SSong Liu #define BPF_BUILD_ID_SIZE 20 1242615755a7SSong Liu struct bpf_stack_build_id { 1243615755a7SSong Liu __s32 status; 1244615755a7SSong Liu unsigned char build_id[BPF_BUILD_ID_SIZE]; 1245615755a7SSong Liu union { 1246615755a7SSong Liu __u64 offset; 1247615755a7SSong Liu __u64 ip; 1248615755a7SSong Liu }; 1249615755a7SSong Liu }; 1250615755a7SSong Liu 12511aae4bddSAndrii Nakryiko #define BPF_OBJ_NAME_LEN 16U 12521aae4bddSAndrii Nakryiko 125399c55f7dSAlexei Starovoitov union bpf_attr { 125499c55f7dSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_CREATE command */ 125599c55f7dSAlexei Starovoitov __u32 map_type; /* one of enum bpf_map_type */ 125699c55f7dSAlexei Starovoitov __u32 key_size; /* size of key in bytes */ 125799c55f7dSAlexei Starovoitov __u32 value_size; /* size of value in bytes */ 125899c55f7dSAlexei Starovoitov __u32 max_entries; /* max number of entries in a map */ 125996eabe7aSMartin KaFai Lau __u32 map_flags; /* BPF_MAP_CREATE related 126096eabe7aSMartin KaFai Lau * flags defined above. 126196eabe7aSMartin KaFai Lau */ 126256f668dfSMartin KaFai Lau __u32 inner_map_fd; /* fd pointing to the inner map */ 126396eabe7aSMartin KaFai Lau __u32 numa_node; /* numa node (effective only if 126496eabe7aSMartin KaFai Lau * BPF_F_NUMA_NODE is set). 126596eabe7aSMartin KaFai Lau */ 1266067cae47SMartin KaFai Lau char map_name[BPF_OBJ_NAME_LEN]; 1267a3884572SJakub Kicinski __u32 map_ifindex; /* ifindex of netdev to create on */ 1268a26ca7c9SMartin KaFai Lau __u32 btf_fd; /* fd pointing to a BTF type data */ 12699b2cf328SMartin KaFai Lau __u32 btf_key_type_id; /* BTF type_id of the key */ 12709b2cf328SMartin KaFai Lau __u32 btf_value_type_id; /* BTF type_id of the value */ 127185d33df3SMartin KaFai Lau __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 127285d33df3SMartin KaFai Lau * struct stored as the 127385d33df3SMartin KaFai Lau * map value 127485d33df3SMartin KaFai Lau */ 127599c55f7dSAlexei Starovoitov }; 1276db20fd2bSAlexei Starovoitov 1277db20fd2bSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1278db20fd2bSAlexei Starovoitov __u32 map_fd; 1279db20fd2bSAlexei Starovoitov __aligned_u64 key; 1280db20fd2bSAlexei Starovoitov union { 1281db20fd2bSAlexei Starovoitov __aligned_u64 value; 1282db20fd2bSAlexei Starovoitov __aligned_u64 next_key; 1283db20fd2bSAlexei Starovoitov }; 12843274f520SAlexei Starovoitov __u64 flags; 1285db20fd2bSAlexei Starovoitov }; 128609756af4SAlexei Starovoitov 1287cb4d03abSBrian Vazquez struct { /* struct used by BPF_MAP_*_BATCH commands */ 1288cb4d03abSBrian Vazquez __aligned_u64 in_batch; /* start batch, 1289cb4d03abSBrian Vazquez * NULL to start from beginning 1290cb4d03abSBrian Vazquez */ 1291cb4d03abSBrian Vazquez __aligned_u64 out_batch; /* output: next start batch */ 1292cb4d03abSBrian Vazquez __aligned_u64 keys; 1293cb4d03abSBrian Vazquez __aligned_u64 values; 1294cb4d03abSBrian Vazquez __u32 count; /* input/output: 1295cb4d03abSBrian Vazquez * input: # of key/value 1296cb4d03abSBrian Vazquez * elements 1297cb4d03abSBrian Vazquez * output: # of filled elements 1298cb4d03abSBrian Vazquez */ 1299cb4d03abSBrian Vazquez __u32 map_fd; 1300cb4d03abSBrian Vazquez __u64 elem_flags; 1301cb4d03abSBrian Vazquez __u64 flags; 1302cb4d03abSBrian Vazquez } batch; 1303cb4d03abSBrian Vazquez 130409756af4SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_LOAD command */ 130509756af4SAlexei Starovoitov __u32 prog_type; /* one of enum bpf_prog_type */ 130609756af4SAlexei Starovoitov __u32 insn_cnt; 130709756af4SAlexei Starovoitov __aligned_u64 insns; 130809756af4SAlexei Starovoitov __aligned_u64 license; 1309cbd35700SAlexei Starovoitov __u32 log_level; /* verbosity level of verifier */ 1310cbd35700SAlexei Starovoitov __u32 log_size; /* size of user buffer */ 1311cbd35700SAlexei Starovoitov __aligned_u64 log_buf; /* user supplied buffer */ 13126c4fc209SDaniel Borkmann __u32 kern_version; /* not used */ 1313e07b98d9SDavid S. Miller __u32 prog_flags; 1314067cae47SMartin KaFai Lau char prog_name[BPF_OBJ_NAME_LEN]; 13151f6f4cb7SJakub Kicinski __u32 prog_ifindex; /* ifindex of netdev to prep for */ 13165e43f899SAndrey Ignatov /* For some prog types expected attach type must be known at 13175e43f899SAndrey Ignatov * load time to verify attach type specific parts of prog 13185e43f899SAndrey Ignatov * (context accesses, allowed helpers, etc). 13195e43f899SAndrey Ignatov */ 13205e43f899SAndrey Ignatov __u32 expected_attach_type; 1321838e9690SYonghong Song __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1322838e9690SYonghong Song __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1323838e9690SYonghong Song __aligned_u64 func_info; /* func info */ 1324838e9690SYonghong Song __u32 func_info_cnt; /* number of bpf_func_info records */ 1325c454a46bSMartin KaFai Lau __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1326c454a46bSMartin KaFai Lau __aligned_u64 line_info; /* line info */ 1327c454a46bSMartin KaFai Lau __u32 line_info_cnt; /* number of bpf_line_info records */ 1328ccfe29ebSAlexei Starovoitov __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1329290248a5SAndrii Nakryiko union { 1330290248a5SAndrii Nakryiko /* valid prog_fd to attach to bpf prog */ 1331290248a5SAndrii Nakryiko __u32 attach_prog_fd; 1332290248a5SAndrii Nakryiko /* or valid module BTF object fd or 0 to attach to vmlinux */ 1333290248a5SAndrii Nakryiko __u32 attach_btf_obj_fd; 1334290248a5SAndrii Nakryiko }; 1335387544bfSAlexei Starovoitov __u32 :32; /* pad */ 1336387544bfSAlexei Starovoitov __aligned_u64 fd_array; /* array of FDs */ 133709756af4SAlexei Starovoitov }; 1338b2197755SDaniel Borkmann 1339b2197755SDaniel Borkmann struct { /* anonymous struct used by BPF_OBJ_* commands */ 1340b2197755SDaniel Borkmann __aligned_u64 pathname; 1341b2197755SDaniel Borkmann __u32 bpf_fd; 13426e71b04aSChenbo Feng __u32 file_flags; 1343b2197755SDaniel Borkmann }; 1344f4324551SDaniel Mack 1345f4324551SDaniel Mack struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1346f4324551SDaniel Mack __u32 target_fd; /* container object to attach to */ 1347f4324551SDaniel Mack __u32 attach_bpf_fd; /* eBPF program to attach */ 1348f4324551SDaniel Mack __u32 attach_type; 13497f677633SAlexei Starovoitov __u32 attach_flags; 13507dd68b32SAndrey Ignatov __u32 replace_bpf_fd; /* previously attached eBPF 13517dd68b32SAndrey Ignatov * program to replace if 13527dd68b32SAndrey Ignatov * BPF_F_REPLACE is used 13537dd68b32SAndrey Ignatov */ 1354f4324551SDaniel Mack }; 13551cf1cae9SAlexei Starovoitov 13561cf1cae9SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 13571cf1cae9SAlexei Starovoitov __u32 prog_fd; 13581cf1cae9SAlexei Starovoitov __u32 retval; 1359b5a36b1eSLorenz Bauer __u32 data_size_in; /* input: len of data_in */ 1360b5a36b1eSLorenz Bauer __u32 data_size_out; /* input/output: len of data_out 1361b5a36b1eSLorenz Bauer * returns ENOSPC if data_out 1362b5a36b1eSLorenz Bauer * is too small. 1363b5a36b1eSLorenz Bauer */ 13641cf1cae9SAlexei Starovoitov __aligned_u64 data_in; 13651cf1cae9SAlexei Starovoitov __aligned_u64 data_out; 13661cf1cae9SAlexei Starovoitov __u32 repeat; 13671cf1cae9SAlexei Starovoitov __u32 duration; 1368b0b9395dSStanislav Fomichev __u32 ctx_size_in; /* input: len of ctx_in */ 1369b0b9395dSStanislav Fomichev __u32 ctx_size_out; /* input/output: len of ctx_out 1370b0b9395dSStanislav Fomichev * returns ENOSPC if ctx_out 1371b0b9395dSStanislav Fomichev * is too small. 1372b0b9395dSStanislav Fomichev */ 1373b0b9395dSStanislav Fomichev __aligned_u64 ctx_in; 1374b0b9395dSStanislav Fomichev __aligned_u64 ctx_out; 13751b4d60ecSSong Liu __u32 flags; 13761b4d60ecSSong Liu __u32 cpu; 13771cf1cae9SAlexei Starovoitov } test; 137834ad5580SMartin KaFai Lau 1379b16d9aa4SMartin KaFai Lau struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1380b16d9aa4SMartin KaFai Lau union { 138134ad5580SMartin KaFai Lau __u32 start_id; 1382b16d9aa4SMartin KaFai Lau __u32 prog_id; 1383bd5f5f4eSMartin KaFai Lau __u32 map_id; 138478958fcaSMartin KaFai Lau __u32 btf_id; 1385a3b80e10SAndrii Nakryiko __u32 link_id; 1386b16d9aa4SMartin KaFai Lau }; 138734ad5580SMartin KaFai Lau __u32 next_id; 13886e71b04aSChenbo Feng __u32 open_flags; 138934ad5580SMartin KaFai Lau }; 13901e270976SMartin KaFai Lau 13911e270976SMartin KaFai Lau struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 13921e270976SMartin KaFai Lau __u32 bpf_fd; 13931e270976SMartin KaFai Lau __u32 info_len; 13941e270976SMartin KaFai Lau __aligned_u64 info; 13951e270976SMartin KaFai Lau } info; 1396468e2f64SAlexei Starovoitov 1397468e2f64SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1398468e2f64SAlexei Starovoitov __u32 target_fd; /* container object to query */ 1399468e2f64SAlexei Starovoitov __u32 attach_type; 1400468e2f64SAlexei Starovoitov __u32 query_flags; 1401468e2f64SAlexei Starovoitov __u32 attach_flags; 1402468e2f64SAlexei Starovoitov __aligned_u64 prog_ids; 1403468e2f64SAlexei Starovoitov __u32 prog_cnt; 1404468e2f64SAlexei Starovoitov } query; 1405c4f6699dSAlexei Starovoitov 1406af6eea57SAndrii Nakryiko struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1407c4f6699dSAlexei Starovoitov __u64 name; 1408c4f6699dSAlexei Starovoitov __u32 prog_fd; 1409c4f6699dSAlexei Starovoitov } raw_tracepoint; 1410f56a653cSMartin KaFai Lau 1411f56a653cSMartin KaFai Lau struct { /* anonymous struct for BPF_BTF_LOAD */ 1412f56a653cSMartin KaFai Lau __aligned_u64 btf; 1413f56a653cSMartin KaFai Lau __aligned_u64 btf_log_buf; 1414f56a653cSMartin KaFai Lau __u32 btf_size; 1415f56a653cSMartin KaFai Lau __u32 btf_log_size; 1416f56a653cSMartin KaFai Lau __u32 btf_log_level; 1417f56a653cSMartin KaFai Lau }; 141841bdc4b4SYonghong Song 141941bdc4b4SYonghong Song struct { 142041bdc4b4SYonghong Song __u32 pid; /* input: pid */ 142141bdc4b4SYonghong Song __u32 fd; /* input: fd */ 142241bdc4b4SYonghong Song __u32 flags; /* input: flags */ 142341bdc4b4SYonghong Song __u32 buf_len; /* input/output: buf len */ 142441bdc4b4SYonghong Song __aligned_u64 buf; /* input/output: 142541bdc4b4SYonghong Song * tp_name for tracepoint 142641bdc4b4SYonghong Song * symbol for kprobe 142741bdc4b4SYonghong Song * filename for uprobe 142841bdc4b4SYonghong Song */ 142941bdc4b4SYonghong Song __u32 prog_id; /* output: prod_id */ 143041bdc4b4SYonghong Song __u32 fd_type; /* output: BPF_FD_TYPE_* */ 143141bdc4b4SYonghong Song __u64 probe_offset; /* output: probe_offset */ 143241bdc4b4SYonghong Song __u64 probe_addr; /* output: probe_addr */ 143341bdc4b4SYonghong Song } task_fd_query; 1434af6eea57SAndrii Nakryiko 1435af6eea57SAndrii Nakryiko struct { /* struct used by BPF_LINK_CREATE command */ 1436af6eea57SAndrii Nakryiko __u32 prog_fd; /* eBPF program to attach */ 1437aa8d3a71SAndrii Nakryiko union { 1438af6eea57SAndrii Nakryiko __u32 target_fd; /* object to attach to */ 1439aa8d3a71SAndrii Nakryiko __u32 target_ifindex; /* target ifindex */ 1440aa8d3a71SAndrii Nakryiko }; 1441af6eea57SAndrii Nakryiko __u32 attach_type; /* attach type */ 1442af6eea57SAndrii Nakryiko __u32 flags; /* extra flags */ 14434a1e7c0cSToke Høiland-Jørgensen union { 14444a1e7c0cSToke Høiland-Jørgensen __u32 target_btf_id; /* btf_id of target to attach to */ 14454a1e7c0cSToke Høiland-Jørgensen struct { 14465e7b3020SYonghong Song __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 14475e7b3020SYonghong Song __u32 iter_info_len; /* iter_info length */ 14484a1e7c0cSToke Høiland-Jørgensen }; 14494a1e7c0cSToke Høiland-Jørgensen }; 1450af6eea57SAndrii Nakryiko } link_create; 14510c991ebcSAndrii Nakryiko 14520c991ebcSAndrii Nakryiko struct { /* struct used by BPF_LINK_UPDATE command */ 14530c991ebcSAndrii Nakryiko __u32 link_fd; /* link fd */ 14540c991ebcSAndrii Nakryiko /* new program fd to update link with */ 14550c991ebcSAndrii Nakryiko __u32 new_prog_fd; 14560c991ebcSAndrii Nakryiko __u32 flags; /* extra flags */ 14570c991ebcSAndrii Nakryiko /* expected link's program fd; is specified only if 14580c991ebcSAndrii Nakryiko * BPF_F_REPLACE flag is set in flags */ 14590c991ebcSAndrii Nakryiko __u32 old_prog_fd; 14600c991ebcSAndrii Nakryiko } link_update; 14610c991ebcSAndrii Nakryiko 146273b11c2aSAndrii Nakryiko struct { 146373b11c2aSAndrii Nakryiko __u32 link_fd; 146473b11c2aSAndrii Nakryiko } link_detach; 146573b11c2aSAndrii Nakryiko 1466d46edd67SSong Liu struct { /* struct used by BPF_ENABLE_STATS command */ 1467d46edd67SSong Liu __u32 type; 1468d46edd67SSong Liu } enable_stats; 1469d46edd67SSong Liu 1470ac51d99bSYonghong Song struct { /* struct used by BPF_ITER_CREATE command */ 1471ac51d99bSYonghong Song __u32 link_fd; 1472ac51d99bSYonghong Song __u32 flags; 1473ac51d99bSYonghong Song } iter_create; 1474ac51d99bSYonghong Song 1475ef15314aSYiFei Zhu struct { /* struct used by BPF_PROG_BIND_MAP command */ 1476ef15314aSYiFei Zhu __u32 prog_fd; 1477ef15314aSYiFei Zhu __u32 map_fd; 1478ef15314aSYiFei Zhu __u32 flags; /* extra flags */ 1479ef15314aSYiFei Zhu } prog_bind_map; 1480ef15314aSYiFei Zhu 148199c55f7dSAlexei Starovoitov } __attribute__((aligned(8))); 148299c55f7dSAlexei Starovoitov 148356a092c8SQuentin Monnet /* The description below is an attempt at providing documentation to eBPF 148456a092c8SQuentin Monnet * developers about the multiple available eBPF helper functions. It can be 148556a092c8SQuentin Monnet * parsed and used to produce a manual page. The workflow is the following, 148656a092c8SQuentin Monnet * and requires the rst2man utility: 1487ebb676daSThomas Graf * 1488923a932cSJoe Stringer * $ ./scripts/bpf_doc.py \ 148956a092c8SQuentin Monnet * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 149056a092c8SQuentin Monnet * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 149156a092c8SQuentin Monnet * $ man /tmp/bpf-helpers.7 1492ebb676daSThomas Graf * 149356a092c8SQuentin Monnet * Note that in order to produce this external documentation, some RST 149456a092c8SQuentin Monnet * formatting is used in the descriptions to get "bold" and "italics" in 149556a092c8SQuentin Monnet * manual pages. Also note that the few trailing white spaces are 149656a092c8SQuentin Monnet * intentional, removing them would break paragraphs for rst2man. 1497ebb676daSThomas Graf * 149856a092c8SQuentin Monnet * Start of BPF helper function descriptions: 1499ad4a5223SQuentin Monnet * 1500ad4a5223SQuentin Monnet * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1501ad4a5223SQuentin Monnet * Description 1502ad4a5223SQuentin Monnet * Perform a lookup in *map* for an entry associated to *key*. 1503ad4a5223SQuentin Monnet * Return 1504ad4a5223SQuentin Monnet * Map value associated to *key*, or **NULL** if no entry was 1505ad4a5223SQuentin Monnet * found. 1506ad4a5223SQuentin Monnet * 1507bdb7b79bSAndrii Nakryiko * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1508ad4a5223SQuentin Monnet * Description 1509ad4a5223SQuentin Monnet * Add or update the value of the entry associated to *key* in 1510ad4a5223SQuentin Monnet * *map* with *value*. *flags* is one of: 1511ad4a5223SQuentin Monnet * 1512ad4a5223SQuentin Monnet * **BPF_NOEXIST** 1513ad4a5223SQuentin Monnet * The entry for *key* must not exist in the map. 1514ad4a5223SQuentin Monnet * **BPF_EXIST** 1515ad4a5223SQuentin Monnet * The entry for *key* must already exist in the map. 1516ad4a5223SQuentin Monnet * **BPF_ANY** 1517ad4a5223SQuentin Monnet * No condition on the existence of the entry for *key*. 1518ad4a5223SQuentin Monnet * 1519ad4a5223SQuentin Monnet * Flag value **BPF_NOEXIST** cannot be used for maps of types 1520ad4a5223SQuentin Monnet * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1521ad4a5223SQuentin Monnet * elements always exist), the helper would return an error. 1522ad4a5223SQuentin Monnet * Return 1523ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1524ad4a5223SQuentin Monnet * 1525bdb7b79bSAndrii Nakryiko * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1526ad4a5223SQuentin Monnet * Description 1527ad4a5223SQuentin Monnet * Delete entry with *key* from *map*. 1528ad4a5223SQuentin Monnet * Return 1529ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1530ad4a5223SQuentin Monnet * 1531bdb7b79bSAndrii Nakryiko * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1532ad4a5223SQuentin Monnet * Description 1533ad4a5223SQuentin Monnet * For tracing programs, safely attempt to read *size* bytes from 15346ae08ae3SDaniel Borkmann * kernel space address *unsafe_ptr* and store the data in *dst*. 15356ae08ae3SDaniel Borkmann * 1536ab8d7809SQuentin Monnet * Generally, use **bpf_probe_read_user**\ () or 1537ab8d7809SQuentin Monnet * **bpf_probe_read_kernel**\ () instead. 1538ad4a5223SQuentin Monnet * Return 1539ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1540ad4a5223SQuentin Monnet * 1541ad4a5223SQuentin Monnet * u64 bpf_ktime_get_ns(void) 1542ad4a5223SQuentin Monnet * Description 1543ad4a5223SQuentin Monnet * Return the time elapsed since system boot, in nanoseconds. 154471d19214SMaciej Żenczykowski * Does not include time the system was suspended. 1545ab8d7809SQuentin Monnet * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1546ad4a5223SQuentin Monnet * Return 1547ad4a5223SQuentin Monnet * Current *ktime*. 1548ad4a5223SQuentin Monnet * 1549bdb7b79bSAndrii Nakryiko * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1550ad4a5223SQuentin Monnet * Description 1551ad4a5223SQuentin Monnet * This helper is a "printk()-like" facility for debugging. It 1552ad4a5223SQuentin Monnet * prints a message defined by format *fmt* (of size *fmt_size*) 1553ad4a5223SQuentin Monnet * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 1554ad4a5223SQuentin Monnet * available. It can take up to three additional **u64** 1555ad4a5223SQuentin Monnet * arguments (as an eBPF helpers, the total number of arguments is 1556ad4a5223SQuentin Monnet * limited to five). 1557ad4a5223SQuentin Monnet * 1558ad4a5223SQuentin Monnet * Each time the helper is called, it appends a line to the trace. 155955c33dfbSPeter Wu * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 156055c33dfbSPeter Wu * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 1561ad4a5223SQuentin Monnet * The format of the trace is customizable, and the exact output 1562ad4a5223SQuentin Monnet * one will get depends on the options set in 1563ad4a5223SQuentin Monnet * *\/sys/kernel/debug/tracing/trace_options* (see also the 1564ad4a5223SQuentin Monnet * *README* file under the same directory). However, it usually 1565ad4a5223SQuentin Monnet * defaults to something like: 1566ad4a5223SQuentin Monnet * 1567ad4a5223SQuentin Monnet * :: 1568ad4a5223SQuentin Monnet * 1569ad4a5223SQuentin Monnet * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1570ad4a5223SQuentin Monnet * 1571ad4a5223SQuentin Monnet * In the above: 1572ad4a5223SQuentin Monnet * 1573ad4a5223SQuentin Monnet * * ``telnet`` is the name of the current task. 1574ad4a5223SQuentin Monnet * * ``470`` is the PID of the current task. 1575ad4a5223SQuentin Monnet * * ``001`` is the CPU number on which the task is 1576ad4a5223SQuentin Monnet * running. 1577ad4a5223SQuentin Monnet * * In ``.N..``, each character refers to a set of 1578ad4a5223SQuentin Monnet * options (whether irqs are enabled, scheduling 1579ad4a5223SQuentin Monnet * options, whether hard/softirqs are running, level of 1580ad4a5223SQuentin Monnet * preempt_disabled respectively). **N** means that 1581ad4a5223SQuentin Monnet * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1582ad4a5223SQuentin Monnet * are set. 1583ad4a5223SQuentin Monnet * * ``419421.045894`` is a timestamp. 1584ad4a5223SQuentin Monnet * * ``0x00000001`` is a fake value used by BPF for the 1585ad4a5223SQuentin Monnet * instruction pointer register. 1586ad4a5223SQuentin Monnet * * ``<formatted msg>`` is the message formatted with 1587ad4a5223SQuentin Monnet * *fmt*. 1588ad4a5223SQuentin Monnet * 1589ad4a5223SQuentin Monnet * The conversion specifiers supported by *fmt* are similar, but 1590ad4a5223SQuentin Monnet * more limited than for printk(). They are **%d**, **%i**, 1591ad4a5223SQuentin Monnet * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1592ad4a5223SQuentin Monnet * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1593ad4a5223SQuentin Monnet * of field, padding with zeroes, etc.) is available, and the 1594ad4a5223SQuentin Monnet * helper will return **-EINVAL** (but print nothing) if it 1595ad4a5223SQuentin Monnet * encounters an unknown specifier. 1596ad4a5223SQuentin Monnet * 1597ad4a5223SQuentin Monnet * Also, note that **bpf_trace_printk**\ () is slow, and should 1598ad4a5223SQuentin Monnet * only be used for debugging purposes. For this reason, a notice 1599b16fc097STobias Klauser * block (spanning several lines) is printed to kernel logs and 1600ad4a5223SQuentin Monnet * states that the helper should not be used "for production use" 1601ad4a5223SQuentin Monnet * the first time this helper is used (or more precisely, when 1602ad4a5223SQuentin Monnet * **trace_printk**\ () buffers are allocated). For passing values 1603ad4a5223SQuentin Monnet * to user space, perf events should be preferred. 1604ad4a5223SQuentin Monnet * Return 1605ad4a5223SQuentin Monnet * The number of bytes written to the buffer, or a negative error 1606ad4a5223SQuentin Monnet * in case of failure. 1607ad4a5223SQuentin Monnet * 16081fdd08beSQuentin Monnet * u32 bpf_get_prandom_u32(void) 16091fdd08beSQuentin Monnet * Description 16101fdd08beSQuentin Monnet * Get a pseudo-random number. 16111fdd08beSQuentin Monnet * 16121fdd08beSQuentin Monnet * From a security point of view, this helper uses its own 16131fdd08beSQuentin Monnet * pseudo-random internal state, and cannot be used to infer the 16141fdd08beSQuentin Monnet * seed of other random functions in the kernel. However, it is 16151fdd08beSQuentin Monnet * essential to note that the generator used by the helper is not 16161fdd08beSQuentin Monnet * cryptographically secure. 16171fdd08beSQuentin Monnet * Return 16181fdd08beSQuentin Monnet * A random 32-bit unsigned value. 16191fdd08beSQuentin Monnet * 16201fdd08beSQuentin Monnet * u32 bpf_get_smp_processor_id(void) 16211fdd08beSQuentin Monnet * Description 16221fdd08beSQuentin Monnet * Get the SMP (symmetric multiprocessing) processor id. Note that 16231fdd08beSQuentin Monnet * all programs run with preemption disabled, which means that the 16241fdd08beSQuentin Monnet * SMP processor id is stable during all the execution of the 16251fdd08beSQuentin Monnet * program. 16261fdd08beSQuentin Monnet * Return 16271fdd08beSQuentin Monnet * The SMP id of the processor running the program. 16281fdd08beSQuentin Monnet * 1629bdb7b79bSAndrii Nakryiko * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1630ad4a5223SQuentin Monnet * Description 1631ad4a5223SQuentin Monnet * Store *len* bytes from address *from* into the packet 1632ad4a5223SQuentin Monnet * associated to *skb*, at *offset*. *flags* are a combination of 1633ad4a5223SQuentin Monnet * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1634ad4a5223SQuentin Monnet * checksum for the packet after storing the bytes) and 1635ad4a5223SQuentin Monnet * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1636ad4a5223SQuentin Monnet * **->swhash** and *skb*\ **->l4hash** to 0). 1637ad4a5223SQuentin Monnet * 163832e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1639ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1640ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1641ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1642ad4a5223SQuentin Monnet * direct packet access. 1643ad4a5223SQuentin Monnet * Return 1644ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1645ad4a5223SQuentin Monnet * 1646bdb7b79bSAndrii Nakryiko * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1647ad4a5223SQuentin Monnet * Description 1648ad4a5223SQuentin Monnet * Recompute the layer 3 (e.g. IP) checksum for the packet 1649ad4a5223SQuentin Monnet * associated to *skb*. Computation is incremental, so the helper 1650ad4a5223SQuentin Monnet * must know the former value of the header field that was 1651ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 1652ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored in *size*. 1653ad4a5223SQuentin Monnet * Alternatively, it is possible to store the difference between 1654ad4a5223SQuentin Monnet * the previous and the new values of the header field in *to*, by 1655ad4a5223SQuentin Monnet * setting *from* and *size* to 0. For both methods, *offset* 1656ad4a5223SQuentin Monnet * indicates the location of the IP checksum within the packet. 1657ad4a5223SQuentin Monnet * 1658ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 1659ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 1660ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 1661ad4a5223SQuentin Monnet * checksum to update. 1662ad4a5223SQuentin Monnet * 166332e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1664ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1665ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1666ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1667ad4a5223SQuentin Monnet * direct packet access. 1668ad4a5223SQuentin Monnet * Return 1669ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1670ad4a5223SQuentin Monnet * 1671bdb7b79bSAndrii Nakryiko * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1672ad4a5223SQuentin Monnet * Description 1673ad4a5223SQuentin Monnet * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1674ad4a5223SQuentin Monnet * packet associated to *skb*. Computation is incremental, so the 1675ad4a5223SQuentin Monnet * helper must know the former value of the header field that was 1676ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 1677ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored on the lowest 1678ad4a5223SQuentin Monnet * four bits of *flags*. Alternatively, it is possible to store 1679ad4a5223SQuentin Monnet * the difference between the previous and the new values of the 1680ad4a5223SQuentin Monnet * header field in *to*, by setting *from* and the four lowest 1681ad4a5223SQuentin Monnet * bits of *flags* to 0. For both methods, *offset* indicates the 1682ad4a5223SQuentin Monnet * location of the IP checksum within the packet. In addition to 1683ad4a5223SQuentin Monnet * the size of the field, *flags* can be added (bitwise OR) actual 1684ad4a5223SQuentin Monnet * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1685ad4a5223SQuentin Monnet * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1686ad4a5223SQuentin Monnet * for updates resulting in a null checksum the value is set to 1687ad4a5223SQuentin Monnet * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1688ad4a5223SQuentin Monnet * the checksum is to be computed against a pseudo-header. 1689ad4a5223SQuentin Monnet * 1690ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 1691ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 1692ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 1693ad4a5223SQuentin Monnet * checksum to update. 1694ad4a5223SQuentin Monnet * 169532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1696ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1697ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1698ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1699ad4a5223SQuentin Monnet * direct packet access. 1700ad4a5223SQuentin Monnet * Return 1701ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1702ad4a5223SQuentin Monnet * 1703bdb7b79bSAndrii Nakryiko * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 1704ad4a5223SQuentin Monnet * Description 1705ad4a5223SQuentin Monnet * This special helper is used to trigger a "tail call", or in 1706ad4a5223SQuentin Monnet * other words, to jump into another eBPF program. The same stack 1707ad4a5223SQuentin Monnet * frame is used (but values on stack and in registers for the 1708ad4a5223SQuentin Monnet * caller are not accessible to the callee). This mechanism allows 1709ad4a5223SQuentin Monnet * for program chaining, either for raising the maximum number of 1710ad4a5223SQuentin Monnet * available eBPF instructions, or to execute given programs in 1711ad4a5223SQuentin Monnet * conditional blocks. For security reasons, there is an upper 1712ad4a5223SQuentin Monnet * limit to the number of successive tail calls that can be 1713ad4a5223SQuentin Monnet * performed. 1714ad4a5223SQuentin Monnet * 1715ad4a5223SQuentin Monnet * Upon call of this helper, the program attempts to jump into a 1716ad4a5223SQuentin Monnet * program referenced at index *index* in *prog_array_map*, a 1717ad4a5223SQuentin Monnet * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 1718ad4a5223SQuentin Monnet * *ctx*, a pointer to the context. 1719ad4a5223SQuentin Monnet * 1720ad4a5223SQuentin Monnet * If the call succeeds, the kernel immediately runs the first 1721ad4a5223SQuentin Monnet * instruction of the new program. This is not a function call, 1722ad4a5223SQuentin Monnet * and it never returns to the previous program. If the call 1723ad4a5223SQuentin Monnet * fails, then the helper has no effect, and the caller continues 1724ad4a5223SQuentin Monnet * to run its subsequent instructions. A call can fail if the 1725ad4a5223SQuentin Monnet * destination program for the jump does not exist (i.e. *index* 1726ad4a5223SQuentin Monnet * is superior to the number of entries in *prog_array_map*), or 1727ad4a5223SQuentin Monnet * if the maximum number of tail calls has been reached for this 1728ad4a5223SQuentin Monnet * chain of programs. This limit is defined in the kernel by the 1729ad4a5223SQuentin Monnet * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 1730ad4a5223SQuentin Monnet * which is currently set to 32. 1731ad4a5223SQuentin Monnet * Return 1732ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1733ad4a5223SQuentin Monnet * 1734bdb7b79bSAndrii Nakryiko * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 1735ad4a5223SQuentin Monnet * Description 1736ad4a5223SQuentin Monnet * Clone and redirect the packet associated to *skb* to another 1737ad4a5223SQuentin Monnet * net device of index *ifindex*. Both ingress and egress 1738ad4a5223SQuentin Monnet * interfaces can be used for redirection. The **BPF_F_INGRESS** 1739ad4a5223SQuentin Monnet * value in *flags* is used to make the distinction (ingress path 1740ad4a5223SQuentin Monnet * is selected if the flag is present, egress path otherwise). 1741ad4a5223SQuentin Monnet * This is the only flag supported for now. 1742ad4a5223SQuentin Monnet * 1743ad4a5223SQuentin Monnet * In comparison with **bpf_redirect**\ () helper, 1744ad4a5223SQuentin Monnet * **bpf_clone_redirect**\ () has the associated cost of 1745ad4a5223SQuentin Monnet * duplicating the packet buffer, but this can be executed out of 1746ad4a5223SQuentin Monnet * the eBPF program. Conversely, **bpf_redirect**\ () is more 1747ad4a5223SQuentin Monnet * efficient, but it is handled through an action code where the 1748ad4a5223SQuentin Monnet * redirection happens only after the eBPF program has returned. 1749ad4a5223SQuentin Monnet * 175032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1751ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1752ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1753ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1754ad4a5223SQuentin Monnet * direct packet access. 1755ad4a5223SQuentin Monnet * Return 1756ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1757c456dec4SQuentin Monnet * 1758c456dec4SQuentin Monnet * u64 bpf_get_current_pid_tgid(void) 1759c456dec4SQuentin Monnet * Return 1760c456dec4SQuentin Monnet * A 64-bit integer containing the current tgid and pid, and 1761c456dec4SQuentin Monnet * created as such: 1762c456dec4SQuentin Monnet * *current_task*\ **->tgid << 32 \|** 1763c456dec4SQuentin Monnet * *current_task*\ **->pid**. 1764c456dec4SQuentin Monnet * 1765c456dec4SQuentin Monnet * u64 bpf_get_current_uid_gid(void) 1766c456dec4SQuentin Monnet * Return 1767c456dec4SQuentin Monnet * A 64-bit integer containing the current GID and UID, and 1768c456dec4SQuentin Monnet * created as such: *current_gid* **<< 32 \|** *current_uid*. 1769c456dec4SQuentin Monnet * 1770bdb7b79bSAndrii Nakryiko * long bpf_get_current_comm(void *buf, u32 size_of_buf) 1771c456dec4SQuentin Monnet * Description 1772c456dec4SQuentin Monnet * Copy the **comm** attribute of the current task into *buf* of 1773c456dec4SQuentin Monnet * *size_of_buf*. The **comm** attribute contains the name of 1774c456dec4SQuentin Monnet * the executable (excluding the path) for the current task. The 1775c456dec4SQuentin Monnet * *size_of_buf* must be strictly positive. On success, the 1776c456dec4SQuentin Monnet * helper makes sure that the *buf* is NUL-terminated. On failure, 1777c456dec4SQuentin Monnet * it is filled with zeroes. 1778c456dec4SQuentin Monnet * Return 1779c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1780c456dec4SQuentin Monnet * 17811fdd08beSQuentin Monnet * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 17821fdd08beSQuentin Monnet * Description 17831fdd08beSQuentin Monnet * Retrieve the classid for the current task, i.e. for the net_cls 17841fdd08beSQuentin Monnet * cgroup to which *skb* belongs. 17851fdd08beSQuentin Monnet * 17861fdd08beSQuentin Monnet * This helper can be used on TC egress path, but not on ingress. 17871fdd08beSQuentin Monnet * 17881fdd08beSQuentin Monnet * The net_cls cgroup provides an interface to tag network packets 17891fdd08beSQuentin Monnet * based on a user-provided identifier for all traffic coming from 17901fdd08beSQuentin Monnet * the tasks belonging to the related cgroup. See also the related 17911fdd08beSQuentin Monnet * kernel documentation, available from the Linux sources in file 1792da82c92fSMauro Carvalho Chehab * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 17931fdd08beSQuentin Monnet * 17941fdd08beSQuentin Monnet * The Linux kernel has two versions for cgroups: there are 17951fdd08beSQuentin Monnet * cgroups v1 and cgroups v2. Both are available to users, who can 17961fdd08beSQuentin Monnet * use a mixture of them, but note that the net_cls cgroup is for 17971fdd08beSQuentin Monnet * cgroup v1 only. This makes it incompatible with BPF programs 17981fdd08beSQuentin Monnet * run on cgroups, which is a cgroup-v2-only feature (a socket can 17991fdd08beSQuentin Monnet * only hold data for one version of cgroups at a time). 18001fdd08beSQuentin Monnet * 18011fdd08beSQuentin Monnet * This helper is only available is the kernel was compiled with 18021fdd08beSQuentin Monnet * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 18031fdd08beSQuentin Monnet * "**y**" or to "**m**". 18041fdd08beSQuentin Monnet * Return 18051fdd08beSQuentin Monnet * The classid, or 0 for the default unconfigured classid. 18061fdd08beSQuentin Monnet * 1807bdb7b79bSAndrii Nakryiko * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1808c456dec4SQuentin Monnet * Description 1809c456dec4SQuentin Monnet * Push a *vlan_tci* (VLAN tag control information) of protocol 1810c456dec4SQuentin Monnet * *vlan_proto* to the packet associated to *skb*, then update 1811c456dec4SQuentin Monnet * the checksum. Note that if *vlan_proto* is different from 1812c456dec4SQuentin Monnet * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1813c456dec4SQuentin Monnet * be **ETH_P_8021Q**. 1814c456dec4SQuentin Monnet * 181532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1816c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1817c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 1818c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 1819c456dec4SQuentin Monnet * direct packet access. 1820c456dec4SQuentin Monnet * Return 1821c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1822c456dec4SQuentin Monnet * 1823bdb7b79bSAndrii Nakryiko * long bpf_skb_vlan_pop(struct sk_buff *skb) 1824c456dec4SQuentin Monnet * Description 1825c456dec4SQuentin Monnet * Pop a VLAN header from the packet associated to *skb*. 1826c456dec4SQuentin Monnet * 182732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1828c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1829c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 1830c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 1831c456dec4SQuentin Monnet * direct packet access. 1832c456dec4SQuentin Monnet * Return 1833c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1834c456dec4SQuentin Monnet * 1835bdb7b79bSAndrii Nakryiko * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1836c456dec4SQuentin Monnet * Description 1837c456dec4SQuentin Monnet * Get tunnel metadata. This helper takes a pointer *key* to an 1838c456dec4SQuentin Monnet * empty **struct bpf_tunnel_key** of **size**, that will be 1839c456dec4SQuentin Monnet * filled with tunnel metadata for the packet associated to *skb*. 1840c456dec4SQuentin Monnet * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1841c456dec4SQuentin Monnet * indicates that the tunnel is based on IPv6 protocol instead of 1842c456dec4SQuentin Monnet * IPv4. 1843c456dec4SQuentin Monnet * 1844c456dec4SQuentin Monnet * The **struct bpf_tunnel_key** is an object that generalizes the 1845c456dec4SQuentin Monnet * principal parameters used by various tunneling protocols into a 1846c456dec4SQuentin Monnet * single struct. This way, it can be used to easily make a 1847c456dec4SQuentin Monnet * decision based on the contents of the encapsulation header, 1848c456dec4SQuentin Monnet * "summarized" in this struct. In particular, it holds the IP 1849c456dec4SQuentin Monnet * address of the remote end (IPv4 or IPv6, depending on the case) 1850c456dec4SQuentin Monnet * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1851c456dec4SQuentin Monnet * this struct exposes the *key*\ **->tunnel_id**, which is 1852c456dec4SQuentin Monnet * generally mapped to a VNI (Virtual Network Identifier), making 1853c456dec4SQuentin Monnet * it programmable together with the **bpf_skb_set_tunnel_key**\ 1854c456dec4SQuentin Monnet * () helper. 1855c456dec4SQuentin Monnet * 1856c456dec4SQuentin Monnet * Let's imagine that the following code is part of a program 1857c456dec4SQuentin Monnet * attached to the TC ingress interface, on one end of a GRE 1858c456dec4SQuentin Monnet * tunnel, and is supposed to filter out all messages coming from 1859c456dec4SQuentin Monnet * remote ends with IPv4 address other than 10.0.0.1: 1860c456dec4SQuentin Monnet * 1861c456dec4SQuentin Monnet * :: 1862c456dec4SQuentin Monnet * 1863c456dec4SQuentin Monnet * int ret; 1864c456dec4SQuentin Monnet * struct bpf_tunnel_key key = {}; 1865c456dec4SQuentin Monnet * 1866c456dec4SQuentin Monnet * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1867c456dec4SQuentin Monnet * if (ret < 0) 1868c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 1869c456dec4SQuentin Monnet * 1870c456dec4SQuentin Monnet * if (key.remote_ipv4 != 0x0a000001) 1871c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 1872c456dec4SQuentin Monnet * 1873c456dec4SQuentin Monnet * return TC_ACT_OK; // accept packet 1874c456dec4SQuentin Monnet * 1875c456dec4SQuentin Monnet * This interface can also be used with all encapsulation devices 1876c456dec4SQuentin Monnet * that can operate in "collect metadata" mode: instead of having 1877c456dec4SQuentin Monnet * one network device per specific configuration, the "collect 1878c456dec4SQuentin Monnet * metadata" mode only requires a single device where the 1879c456dec4SQuentin Monnet * configuration can be extracted from this helper. 1880c456dec4SQuentin Monnet * 1881c456dec4SQuentin Monnet * This can be used together with various tunnels such as VXLan, 1882c456dec4SQuentin Monnet * Geneve, GRE or IP in IP (IPIP). 1883c456dec4SQuentin Monnet * Return 1884c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1885c456dec4SQuentin Monnet * 1886bdb7b79bSAndrii Nakryiko * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1887c456dec4SQuentin Monnet * Description 1888c456dec4SQuentin Monnet * Populate tunnel metadata for packet associated to *skb.* The 1889c456dec4SQuentin Monnet * tunnel metadata is set to the contents of *key*, of *size*. The 1890c456dec4SQuentin Monnet * *flags* can be set to a combination of the following values: 1891c456dec4SQuentin Monnet * 1892c456dec4SQuentin Monnet * **BPF_F_TUNINFO_IPV6** 1893c456dec4SQuentin Monnet * Indicate that the tunnel is based on IPv6 protocol 1894c456dec4SQuentin Monnet * instead of IPv4. 1895c456dec4SQuentin Monnet * **BPF_F_ZERO_CSUM_TX** 1896c456dec4SQuentin Monnet * For IPv4 packets, add a flag to tunnel metadata 1897c456dec4SQuentin Monnet * indicating that checksum computation should be skipped 1898c456dec4SQuentin Monnet * and checksum set to zeroes. 1899c456dec4SQuentin Monnet * **BPF_F_DONT_FRAGMENT** 1900c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that the 1901c456dec4SQuentin Monnet * packet should not be fragmented. 1902c456dec4SQuentin Monnet * **BPF_F_SEQ_NUMBER** 1903c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that a 1904c456dec4SQuentin Monnet * sequence number should be added to tunnel header before 1905c456dec4SQuentin Monnet * sending the packet. This flag was added for GRE 1906c456dec4SQuentin Monnet * encapsulation, but might be used with other protocols 1907c456dec4SQuentin Monnet * as well in the future. 1908c456dec4SQuentin Monnet * 1909c456dec4SQuentin Monnet * Here is a typical usage on the transmit path: 1910c456dec4SQuentin Monnet * 1911c456dec4SQuentin Monnet * :: 1912c456dec4SQuentin Monnet * 1913c456dec4SQuentin Monnet * struct bpf_tunnel_key key; 1914c456dec4SQuentin Monnet * populate key ... 1915c456dec4SQuentin Monnet * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1916c456dec4SQuentin Monnet * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1917c456dec4SQuentin Monnet * 1918c456dec4SQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_key**\ () 1919c456dec4SQuentin Monnet * helper for additional information. 1920c456dec4SQuentin Monnet * Return 1921c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1922c456dec4SQuentin Monnet * 1923c6b5fb86SQuentin Monnet * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1924c6b5fb86SQuentin Monnet * Description 1925c6b5fb86SQuentin Monnet * Read the value of a perf event counter. This helper relies on a 1926c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1927c6b5fb86SQuentin Monnet * the perf event counter is selected when *map* is updated with 1928c6b5fb86SQuentin Monnet * perf event file descriptors. The *map* is an array whose size 1929c6b5fb86SQuentin Monnet * is the number of available CPUs, and each cell contains a value 1930c6b5fb86SQuentin Monnet * relative to one CPU. The value to retrieve is indicated by 1931c6b5fb86SQuentin Monnet * *flags*, that contains the index of the CPU to look up, masked 1932c6b5fb86SQuentin Monnet * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1933c6b5fb86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 1934c6b5fb86SQuentin Monnet * current CPU should be retrieved. 1935c6b5fb86SQuentin Monnet * 1936c6b5fb86SQuentin Monnet * Note that before Linux 4.13, only hardware perf event can be 1937c6b5fb86SQuentin Monnet * retrieved. 1938c6b5fb86SQuentin Monnet * 1939c6b5fb86SQuentin Monnet * Also, be aware that the newer helper 1940c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () is recommended over 19413bd5a09bSQuentin Monnet * **bpf_perf_event_read**\ () in general. The latter has some ABI 1942c6b5fb86SQuentin Monnet * quirks where error and counter value are used as a return code 1943c6b5fb86SQuentin Monnet * (which is wrong to do since ranges may overlap). This issue is 19443bd5a09bSQuentin Monnet * fixed with **bpf_perf_event_read_value**\ (), which at the same 19453bd5a09bSQuentin Monnet * time provides more features over the **bpf_perf_event_read**\ 19463bd5a09bSQuentin Monnet * () interface. Please refer to the description of 1947c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () for details. 1948c6b5fb86SQuentin Monnet * Return 1949c6b5fb86SQuentin Monnet * The value of the perf event counter read from the map, or a 1950c6b5fb86SQuentin Monnet * negative error code in case of failure. 1951c6b5fb86SQuentin Monnet * 1952bdb7b79bSAndrii Nakryiko * long bpf_redirect(u32 ifindex, u64 flags) 1953c456dec4SQuentin Monnet * Description 1954c456dec4SQuentin Monnet * Redirect the packet to another net device of index *ifindex*. 1955c456dec4SQuentin Monnet * This helper is somewhat similar to **bpf_clone_redirect**\ 1956c456dec4SQuentin Monnet * (), except that the packet is not cloned, which provides 1957c456dec4SQuentin Monnet * increased performance. 1958c456dec4SQuentin Monnet * 1959c456dec4SQuentin Monnet * Except for XDP, both ingress and egress interfaces can be used 1960c456dec4SQuentin Monnet * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1961c456dec4SQuentin Monnet * to make the distinction (ingress path is selected if the flag 1962c456dec4SQuentin Monnet * is present, egress path otherwise). Currently, XDP only 1963c456dec4SQuentin Monnet * supports redirection to the egress interface, and accepts no 1964c456dec4SQuentin Monnet * flag at all. 1965c456dec4SQuentin Monnet * 1966f25975f4SToke Høiland-Jørgensen * The same effect can also be attained with the more generic 1967f25975f4SToke Høiland-Jørgensen * **bpf_redirect_map**\ (), which uses a BPF map to store the 1968f25975f4SToke Høiland-Jørgensen * redirect target instead of providing it directly to the helper. 1969c456dec4SQuentin Monnet * Return 1970c456dec4SQuentin Monnet * For XDP, the helper returns **XDP_REDIRECT** on success or 1971c456dec4SQuentin Monnet * **XDP_ABORTED** on error. For other program types, the values 1972c456dec4SQuentin Monnet * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1973c456dec4SQuentin Monnet * error. 1974c456dec4SQuentin Monnet * 19751fdd08beSQuentin Monnet * u32 bpf_get_route_realm(struct sk_buff *skb) 19761fdd08beSQuentin Monnet * Description 19771fdd08beSQuentin Monnet * Retrieve the realm or the route, that is to say the 19781fdd08beSQuentin Monnet * **tclassid** field of the destination for the *skb*. The 1979b16fc097STobias Klauser * identifier retrieved is a user-provided tag, similar to the 19801fdd08beSQuentin Monnet * one used with the net_cls cgroup (see description for 19811fdd08beSQuentin Monnet * **bpf_get_cgroup_classid**\ () helper), but here this tag is 19821fdd08beSQuentin Monnet * held by a route (a destination entry), not by a task. 19831fdd08beSQuentin Monnet * 19841fdd08beSQuentin Monnet * Retrieving this identifier works with the clsact TC egress hook 19851fdd08beSQuentin Monnet * (see also **tc-bpf(8)**), or alternatively on conventional 19861fdd08beSQuentin Monnet * classful egress qdiscs, but not on TC ingress path. In case of 19871fdd08beSQuentin Monnet * clsact TC egress hook, this has the advantage that, internally, 19881fdd08beSQuentin Monnet * the destination entry has not been dropped yet in the transmit 19891fdd08beSQuentin Monnet * path. Therefore, the destination entry does not need to be 19901fdd08beSQuentin Monnet * artificially held via **netif_keep_dst**\ () for a classful 19911fdd08beSQuentin Monnet * qdisc until the *skb* is freed. 19921fdd08beSQuentin Monnet * 19931fdd08beSQuentin Monnet * This helper is available only if the kernel was compiled with 19941fdd08beSQuentin Monnet * **CONFIG_IP_ROUTE_CLASSID** configuration option. 19951fdd08beSQuentin Monnet * Return 19961fdd08beSQuentin Monnet * The realm of the route for the packet associated to *skb*, or 0 19971fdd08beSQuentin Monnet * if none was found. 19981fdd08beSQuentin Monnet * 1999bdb7b79bSAndrii Nakryiko * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2000c456dec4SQuentin Monnet * Description 2001c456dec4SQuentin Monnet * Write raw *data* blob into a special BPF perf event held by 2002c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2003c456dec4SQuentin Monnet * event must have the following attributes: **PERF_SAMPLE_RAW** 2004c456dec4SQuentin Monnet * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2005c456dec4SQuentin Monnet * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2006c456dec4SQuentin Monnet * 2007c456dec4SQuentin Monnet * The *flags* are used to indicate the index in *map* for which 2008c456dec4SQuentin Monnet * the value must be put, masked with **BPF_F_INDEX_MASK**. 2009c456dec4SQuentin Monnet * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2010c456dec4SQuentin Monnet * to indicate that the index of the current CPU core should be 2011c456dec4SQuentin Monnet * used. 2012c456dec4SQuentin Monnet * 2013c456dec4SQuentin Monnet * The value to write, of *size*, is passed through eBPF stack and 2014c456dec4SQuentin Monnet * pointed by *data*. 2015c456dec4SQuentin Monnet * 2016c456dec4SQuentin Monnet * The context of the program *ctx* needs also be passed to the 2017c456dec4SQuentin Monnet * helper. 2018c456dec4SQuentin Monnet * 2019c456dec4SQuentin Monnet * On user space, a program willing to read the values needs to 2020c456dec4SQuentin Monnet * call **perf_event_open**\ () on the perf event (either for 2021c456dec4SQuentin Monnet * one or for all CPUs) and to store the file descriptor into the 2022c456dec4SQuentin Monnet * *map*. This must be done before the eBPF program can send data 2023c456dec4SQuentin Monnet * into it. An example is available in file 2024c456dec4SQuentin Monnet * *samples/bpf/trace_output_user.c* in the Linux kernel source 2025c456dec4SQuentin Monnet * tree (the eBPF program counterpart is in 2026c456dec4SQuentin Monnet * *samples/bpf/trace_output_kern.c*). 2027c456dec4SQuentin Monnet * 2028c456dec4SQuentin Monnet * **bpf_perf_event_output**\ () achieves better performance 2029c456dec4SQuentin Monnet * than **bpf_trace_printk**\ () for sharing data with user 2030c456dec4SQuentin Monnet * space, and is much better suitable for streaming data from eBPF 2031c456dec4SQuentin Monnet * programs. 2032c456dec4SQuentin Monnet * 2033c456dec4SQuentin Monnet * Note that this helper is not restricted to tracing use cases 2034c456dec4SQuentin Monnet * and can be used with programs attached to TC or XDP as well, 2035c456dec4SQuentin Monnet * where it allows for passing data to user space listeners. Data 2036c456dec4SQuentin Monnet * can be: 2037c456dec4SQuentin Monnet * 2038c456dec4SQuentin Monnet * * Only custom structs, 2039c456dec4SQuentin Monnet * * Only the packet payload, or 2040c456dec4SQuentin Monnet * * A combination of both. 2041c456dec4SQuentin Monnet * Return 2042c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 2043c456dec4SQuentin Monnet * 2044bdb7b79bSAndrii Nakryiko * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 20451fdd08beSQuentin Monnet * Description 20461fdd08beSQuentin Monnet * This helper was provided as an easy way to load data from a 20471fdd08beSQuentin Monnet * packet. It can be used to load *len* bytes from *offset* from 20481fdd08beSQuentin Monnet * the packet associated to *skb*, into the buffer pointed by 20491fdd08beSQuentin Monnet * *to*. 20501fdd08beSQuentin Monnet * 20511fdd08beSQuentin Monnet * Since Linux 4.7, usage of this helper has mostly been replaced 20521fdd08beSQuentin Monnet * by "direct packet access", enabling packet data to be 20531fdd08beSQuentin Monnet * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 20541fdd08beSQuentin Monnet * pointing respectively to the first byte of packet data and to 20551fdd08beSQuentin Monnet * the byte after the last byte of packet data. However, it 20561fdd08beSQuentin Monnet * remains useful if one wishes to read large quantities of data 20571fdd08beSQuentin Monnet * at once from a packet into the eBPF stack. 20581fdd08beSQuentin Monnet * Return 20591fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 20601fdd08beSQuentin Monnet * 2061bdb7b79bSAndrii Nakryiko * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2062c456dec4SQuentin Monnet * Description 2063c456dec4SQuentin Monnet * Walk a user or a kernel stack and return its id. To achieve 2064c456dec4SQuentin Monnet * this, the helper needs *ctx*, which is a pointer to the context 2065c456dec4SQuentin Monnet * on which the tracing program is executed, and a pointer to a 2066c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2067c456dec4SQuentin Monnet * 2068c456dec4SQuentin Monnet * The last argument, *flags*, holds the number of stack frames to 2069c456dec4SQuentin Monnet * skip (from 0 to 255), masked with 2070c456dec4SQuentin Monnet * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2071c456dec4SQuentin Monnet * a combination of the following flags: 2072c456dec4SQuentin Monnet * 2073c456dec4SQuentin Monnet * **BPF_F_USER_STACK** 2074c456dec4SQuentin Monnet * Collect a user space stack instead of a kernel stack. 2075c456dec4SQuentin Monnet * **BPF_F_FAST_STACK_CMP** 2076c456dec4SQuentin Monnet * Compare stacks by hash only. 2077c456dec4SQuentin Monnet * **BPF_F_REUSE_STACKID** 2078c456dec4SQuentin Monnet * If two different stacks hash into the same *stackid*, 2079c456dec4SQuentin Monnet * discard the old one. 2080c456dec4SQuentin Monnet * 2081c456dec4SQuentin Monnet * The stack id retrieved is a 32 bit long integer handle which 2082c456dec4SQuentin Monnet * can be further combined with other data (including other stack 2083c456dec4SQuentin Monnet * ids) and used as a key into maps. This can be useful for 2084c456dec4SQuentin Monnet * generating a variety of graphs (such as flame graphs or off-cpu 2085c456dec4SQuentin Monnet * graphs). 2086c456dec4SQuentin Monnet * 2087c456dec4SQuentin Monnet * For walking a stack, this helper is an improvement over 2088c456dec4SQuentin Monnet * **bpf_probe_read**\ (), which can be used with unrolled loops 2089c456dec4SQuentin Monnet * but is not efficient and consumes a lot of eBPF instructions. 2090c456dec4SQuentin Monnet * Instead, **bpf_get_stackid**\ () can collect up to 2091c456dec4SQuentin Monnet * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2092c456dec4SQuentin Monnet * this limit can be controlled with the **sysctl** program, and 2093c456dec4SQuentin Monnet * that it should be manually increased in order to profile long 2094c456dec4SQuentin Monnet * user stacks (such as stacks for Java programs). To do so, use: 2095c456dec4SQuentin Monnet * 2096c456dec4SQuentin Monnet * :: 2097c456dec4SQuentin Monnet * 2098c456dec4SQuentin Monnet * # sysctl kernel.perf_event_max_stack=<new value> 2099c456dec4SQuentin Monnet * Return 2100c456dec4SQuentin Monnet * The positive or null stack id on success, or a negative error 2101c456dec4SQuentin Monnet * in case of failure. 2102c456dec4SQuentin Monnet * 21031fdd08beSQuentin Monnet * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 21041fdd08beSQuentin Monnet * Description 21051fdd08beSQuentin Monnet * Compute a checksum difference, from the raw buffer pointed by 21061fdd08beSQuentin Monnet * *from*, of length *from_size* (that must be a multiple of 4), 21071fdd08beSQuentin Monnet * towards the raw buffer pointed by *to*, of size *to_size* 21081fdd08beSQuentin Monnet * (same remark). An optional *seed* can be added to the value 21091fdd08beSQuentin Monnet * (this can be cascaded, the seed may come from a previous call 21101fdd08beSQuentin Monnet * to the helper). 21111fdd08beSQuentin Monnet * 21121fdd08beSQuentin Monnet * This is flexible enough to be used in several ways: 21131fdd08beSQuentin Monnet * 21141fdd08beSQuentin Monnet * * With *from_size* == 0, *to_size* > 0 and *seed* set to 21151fdd08beSQuentin Monnet * checksum, it can be used when pushing new data. 21161fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* == 0 and *seed* set to 21171fdd08beSQuentin Monnet * checksum, it can be used when removing data from a packet. 21181fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 21191fdd08beSQuentin Monnet * can be used to compute a diff. Note that *from_size* and 21201fdd08beSQuentin Monnet * *to_size* do not need to be equal. 21211fdd08beSQuentin Monnet * 21221fdd08beSQuentin Monnet * This helper can be used in combination with 21231fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 21241fdd08beSQuentin Monnet * which one can feed in the difference computed with 21251fdd08beSQuentin Monnet * **bpf_csum_diff**\ (). 21261fdd08beSQuentin Monnet * Return 21271fdd08beSQuentin Monnet * The checksum result, or a negative error code in case of 21281fdd08beSQuentin Monnet * failure. 21291fdd08beSQuentin Monnet * 2130bdb7b79bSAndrii Nakryiko * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 21311fdd08beSQuentin Monnet * Description 21321fdd08beSQuentin Monnet * Retrieve tunnel options metadata for the packet associated to 21331fdd08beSQuentin Monnet * *skb*, and store the raw tunnel option data to the buffer *opt* 21341fdd08beSQuentin Monnet * of *size*. 21351fdd08beSQuentin Monnet * 21361fdd08beSQuentin Monnet * This helper can be used with encapsulation devices that can 21371fdd08beSQuentin Monnet * operate in "collect metadata" mode (please refer to the related 21381fdd08beSQuentin Monnet * note in the description of **bpf_skb_get_tunnel_key**\ () for 21391fdd08beSQuentin Monnet * more details). A particular example where this can be used is 21401fdd08beSQuentin Monnet * in combination with the Geneve encapsulation protocol, where it 21411fdd08beSQuentin Monnet * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 21421fdd08beSQuentin Monnet * and retrieving arbitrary TLVs (Type-Length-Value headers) from 21431fdd08beSQuentin Monnet * the eBPF program. This allows for full customization of these 21441fdd08beSQuentin Monnet * headers. 21451fdd08beSQuentin Monnet * Return 21461fdd08beSQuentin Monnet * The size of the option data retrieved. 21471fdd08beSQuentin Monnet * 2148bdb7b79bSAndrii Nakryiko * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 21491fdd08beSQuentin Monnet * Description 21501fdd08beSQuentin Monnet * Set tunnel options metadata for the packet associated to *skb* 21511fdd08beSQuentin Monnet * to the option data contained in the raw buffer *opt* of *size*. 21521fdd08beSQuentin Monnet * 21531fdd08beSQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_opt**\ () 21541fdd08beSQuentin Monnet * helper for additional information. 21551fdd08beSQuentin Monnet * Return 21561fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 21571fdd08beSQuentin Monnet * 2158bdb7b79bSAndrii Nakryiko * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 21591fdd08beSQuentin Monnet * Description 21601fdd08beSQuentin Monnet * Change the protocol of the *skb* to *proto*. Currently 21611fdd08beSQuentin Monnet * supported are transition from IPv4 to IPv6, and from IPv6 to 21621fdd08beSQuentin Monnet * IPv4. The helper takes care of the groundwork for the 21631fdd08beSQuentin Monnet * transition, including resizing the socket buffer. The eBPF 21641fdd08beSQuentin Monnet * program is expected to fill the new headers, if any, via 21651fdd08beSQuentin Monnet * **skb_store_bytes**\ () and to recompute the checksums with 21661fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 21671fdd08beSQuentin Monnet * (). The main case for this helper is to perform NAT64 21681fdd08beSQuentin Monnet * operations out of an eBPF program. 21691fdd08beSQuentin Monnet * 21701fdd08beSQuentin Monnet * Internally, the GSO type is marked as dodgy so that headers are 21711fdd08beSQuentin Monnet * checked and segments are recalculated by the GSO/GRO engine. 21721fdd08beSQuentin Monnet * The size for GSO target is adapted as well. 21731fdd08beSQuentin Monnet * 21741fdd08beSQuentin Monnet * All values for *flags* are reserved for future usage, and must 21751fdd08beSQuentin Monnet * be left at zero. 21761fdd08beSQuentin Monnet * 217732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 21781fdd08beSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 21791fdd08beSQuentin Monnet * previously done by the verifier are invalidated and must be 21801fdd08beSQuentin Monnet * performed again, if the helper is used in combination with 21811fdd08beSQuentin Monnet * direct packet access. 21821fdd08beSQuentin Monnet * Return 21831fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 21841fdd08beSQuentin Monnet * 2185bdb7b79bSAndrii Nakryiko * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 21861fdd08beSQuentin Monnet * Description 21871fdd08beSQuentin Monnet * Change the packet type for the packet associated to *skb*. This 21881fdd08beSQuentin Monnet * comes down to setting *skb*\ **->pkt_type** to *type*, except 21891fdd08beSQuentin Monnet * the eBPF program does not have a write access to *skb*\ 21901fdd08beSQuentin Monnet * **->pkt_type** beside this helper. Using a helper here allows 21911fdd08beSQuentin Monnet * for graceful handling of errors. 21921fdd08beSQuentin Monnet * 21931fdd08beSQuentin Monnet * The major use case is to change incoming *skb*s to 21941fdd08beSQuentin Monnet * **PACKET_HOST** in a programmatic way instead of having to 21951fdd08beSQuentin Monnet * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 21961fdd08beSQuentin Monnet * example. 21971fdd08beSQuentin Monnet * 21981fdd08beSQuentin Monnet * Note that *type* only allows certain values. At this time, they 21991fdd08beSQuentin Monnet * are: 22001fdd08beSQuentin Monnet * 22011fdd08beSQuentin Monnet * **PACKET_HOST** 22021fdd08beSQuentin Monnet * Packet is for us. 22031fdd08beSQuentin Monnet * **PACKET_BROADCAST** 22041fdd08beSQuentin Monnet * Send packet to all. 22051fdd08beSQuentin Monnet * **PACKET_MULTICAST** 22061fdd08beSQuentin Monnet * Send packet to group. 22071fdd08beSQuentin Monnet * **PACKET_OTHERHOST** 22081fdd08beSQuentin Monnet * Send packet to someone else. 22091fdd08beSQuentin Monnet * Return 22101fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 22111fdd08beSQuentin Monnet * 2212bdb7b79bSAndrii Nakryiko * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2213c6b5fb86SQuentin Monnet * Description 2214c6b5fb86SQuentin Monnet * Check whether *skb* is a descendant of the cgroup2 held by 2215c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2216c6b5fb86SQuentin Monnet * Return 2217c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 2218c6b5fb86SQuentin Monnet * 2219c6b5fb86SQuentin Monnet * * 0, if the *skb* failed the cgroup2 descendant test. 2220c6b5fb86SQuentin Monnet * * 1, if the *skb* succeeded the cgroup2 descendant test. 2221c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 2222c6b5fb86SQuentin Monnet * 2223fa15601aSQuentin Monnet * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2224fa15601aSQuentin Monnet * Description 2225fa15601aSQuentin Monnet * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2226fa15601aSQuentin Monnet * not set, in particular if the hash was cleared due to mangling, 2227fa15601aSQuentin Monnet * recompute this hash. Later accesses to the hash can be done 2228fa15601aSQuentin Monnet * directly with *skb*\ **->hash**. 2229fa15601aSQuentin Monnet * 2230fa15601aSQuentin Monnet * Calling **bpf_set_hash_invalid**\ (), changing a packet 2231fa15601aSQuentin Monnet * prototype with **bpf_skb_change_proto**\ (), or calling 2232fa15601aSQuentin Monnet * **bpf_skb_store_bytes**\ () with the 2233fa15601aSQuentin Monnet * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2234fa15601aSQuentin Monnet * the hash and to trigger a new computation for the next call to 2235fa15601aSQuentin Monnet * **bpf_get_hash_recalc**\ (). 2236fa15601aSQuentin Monnet * Return 2237fa15601aSQuentin Monnet * The 32-bit hash. 2238fa15601aSQuentin Monnet * 2239c456dec4SQuentin Monnet * u64 bpf_get_current_task(void) 2240c456dec4SQuentin Monnet * Return 2241c456dec4SQuentin Monnet * A pointer to the current task struct. 2242fa15601aSQuentin Monnet * 2243bdb7b79bSAndrii Nakryiko * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2244c6b5fb86SQuentin Monnet * Description 2245c6b5fb86SQuentin Monnet * Attempt in a safe way to write *len* bytes from the buffer 2246c6b5fb86SQuentin Monnet * *src* to *dst* in memory. It only works for threads that are in 2247c6b5fb86SQuentin Monnet * user context, and *dst* must be a valid user space address. 2248c6b5fb86SQuentin Monnet * 2249c6b5fb86SQuentin Monnet * This helper should not be used to implement any kind of 2250c6b5fb86SQuentin Monnet * security mechanism because of TOC-TOU attacks, but rather to 2251c6b5fb86SQuentin Monnet * debug, divert, and manipulate execution of semi-cooperative 2252c6b5fb86SQuentin Monnet * processes. 2253c6b5fb86SQuentin Monnet * 2254c6b5fb86SQuentin Monnet * Keep in mind that this feature is meant for experiments, and it 2255c6b5fb86SQuentin Monnet * has a risk of crashing the system and running programs. 2256c6b5fb86SQuentin Monnet * Therefore, when an eBPF program using this helper is attached, 2257c6b5fb86SQuentin Monnet * a warning including PID and process name is printed to kernel 2258c6b5fb86SQuentin Monnet * logs. 2259c6b5fb86SQuentin Monnet * Return 2260c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2261c6b5fb86SQuentin Monnet * 2262bdb7b79bSAndrii Nakryiko * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2263c6b5fb86SQuentin Monnet * Description 2264c6b5fb86SQuentin Monnet * Check whether the probe is being run is the context of a given 2265c6b5fb86SQuentin Monnet * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2266c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2267c6b5fb86SQuentin Monnet * Return 2268c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 2269c6b5fb86SQuentin Monnet * 22701aef5b43SSong Liu * * 0, if current task belongs to the cgroup2. 22711aef5b43SSong Liu * * 1, if current task does not belong to the cgroup2. 2272c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 2273c6b5fb86SQuentin Monnet * 2274bdb7b79bSAndrii Nakryiko * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2275fa15601aSQuentin Monnet * Description 2276fa15601aSQuentin Monnet * Resize (trim or grow) the packet associated to *skb* to the 2277fa15601aSQuentin Monnet * new *len*. The *flags* are reserved for future usage, and must 2278fa15601aSQuentin Monnet * be left at zero. 2279fa15601aSQuentin Monnet * 2280fa15601aSQuentin Monnet * The basic idea is that the helper performs the needed work to 2281fa15601aSQuentin Monnet * change the size of the packet, then the eBPF program rewrites 2282fa15601aSQuentin Monnet * the rest via helpers like **bpf_skb_store_bytes**\ (), 2283fa15601aSQuentin Monnet * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2284fa15601aSQuentin Monnet * and others. This helper is a slow path utility intended for 2285fa15601aSQuentin Monnet * replies with control messages. And because it is targeted for 2286fa15601aSQuentin Monnet * slow path, the helper itself can afford to be slow: it 2287fa15601aSQuentin Monnet * implicitly linearizes, unclones and drops offloads from the 2288fa15601aSQuentin Monnet * *skb*. 2289fa15601aSQuentin Monnet * 229032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2291fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2292fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2293fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2294fa15601aSQuentin Monnet * direct packet access. 2295fa15601aSQuentin Monnet * Return 2296fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2297fa15601aSQuentin Monnet * 2298bdb7b79bSAndrii Nakryiko * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2299fa15601aSQuentin Monnet * Description 2300fa15601aSQuentin Monnet * Pull in non-linear data in case the *skb* is non-linear and not 2301fa15601aSQuentin Monnet * all of *len* are part of the linear section. Make *len* bytes 2302fa15601aSQuentin Monnet * from *skb* readable and writable. If a zero value is passed for 2303fa15601aSQuentin Monnet * *len*, then the whole length of the *skb* is pulled. 2304fa15601aSQuentin Monnet * 2305fa15601aSQuentin Monnet * This helper is only needed for reading and writing with direct 2306fa15601aSQuentin Monnet * packet access. 2307fa15601aSQuentin Monnet * 2308fa15601aSQuentin Monnet * For direct packet access, testing that offsets to access 2309fa15601aSQuentin Monnet * are within packet boundaries (test on *skb*\ **->data_end**) is 2310fa15601aSQuentin Monnet * susceptible to fail if offsets are invalid, or if the requested 2311fa15601aSQuentin Monnet * data is in non-linear parts of the *skb*. On failure the 2312fa15601aSQuentin Monnet * program can just bail out, or in the case of a non-linear 2313fa15601aSQuentin Monnet * buffer, use a helper to make the data available. The 2314fa15601aSQuentin Monnet * **bpf_skb_load_bytes**\ () helper is a first solution to access 2315fa15601aSQuentin Monnet * the data. Another one consists in using **bpf_skb_pull_data** 2316fa15601aSQuentin Monnet * to pull in once the non-linear parts, then retesting and 2317fa15601aSQuentin Monnet * eventually access the data. 2318fa15601aSQuentin Monnet * 2319fa15601aSQuentin Monnet * At the same time, this also makes sure the *skb* is uncloned, 2320fa15601aSQuentin Monnet * which is a necessary condition for direct write. As this needs 2321fa15601aSQuentin Monnet * to be an invariant for the write part only, the verifier 2322fa15601aSQuentin Monnet * detects writes and adds a prologue that is calling 2323fa15601aSQuentin Monnet * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2324fa15601aSQuentin Monnet * the very beginning in case it is indeed cloned. 2325fa15601aSQuentin Monnet * 232632e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2327fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2328fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2329fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2330fa15601aSQuentin Monnet * direct packet access. 2331fa15601aSQuentin Monnet * Return 2332fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2333fa15601aSQuentin Monnet * 2334fa15601aSQuentin Monnet * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2335fa15601aSQuentin Monnet * Description 2336fa15601aSQuentin Monnet * Add the checksum *csum* into *skb*\ **->csum** in case the 2337fa15601aSQuentin Monnet * driver has supplied a checksum for the entire packet into that 2338fa15601aSQuentin Monnet * field. Return an error otherwise. This helper is intended to be 2339fa15601aSQuentin Monnet * used in combination with **bpf_csum_diff**\ (), in particular 2340fa15601aSQuentin Monnet * when the checksum needs to be updated after data has been 2341fa15601aSQuentin Monnet * written into the packet through direct packet access. 2342fa15601aSQuentin Monnet * Return 2343fa15601aSQuentin Monnet * The checksum on success, or a negative error code in case of 2344fa15601aSQuentin Monnet * failure. 2345fa15601aSQuentin Monnet * 2346fa15601aSQuentin Monnet * void bpf_set_hash_invalid(struct sk_buff *skb) 2347fa15601aSQuentin Monnet * Description 2348fa15601aSQuentin Monnet * Invalidate the current *skb*\ **->hash**. It can be used after 2349fa15601aSQuentin Monnet * mangling on headers through direct packet access, in order to 2350fa15601aSQuentin Monnet * indicate that the hash is outdated and to trigger a 2351fa15601aSQuentin Monnet * recalculation the next time the kernel tries to access this 2352fa15601aSQuentin Monnet * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2353fa15601aSQuentin Monnet * 2354bdb7b79bSAndrii Nakryiko * long bpf_get_numa_node_id(void) 2355fa15601aSQuentin Monnet * Description 2356fa15601aSQuentin Monnet * Return the id of the current NUMA node. The primary use case 2357fa15601aSQuentin Monnet * for this helper is the selection of sockets for the local NUMA 2358fa15601aSQuentin Monnet * node, when the program is attached to sockets using the 2359fa15601aSQuentin Monnet * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2360fa15601aSQuentin Monnet * but the helper is also available to other eBPF program types, 2361fa15601aSQuentin Monnet * similarly to **bpf_get_smp_processor_id**\ (). 2362fa15601aSQuentin Monnet * Return 2363fa15601aSQuentin Monnet * The id of current NUMA node. 2364fa15601aSQuentin Monnet * 2365bdb7b79bSAndrii Nakryiko * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2366c6b5fb86SQuentin Monnet * Description 2367c6b5fb86SQuentin Monnet * Grows headroom of packet associated to *skb* and adjusts the 2368c6b5fb86SQuentin Monnet * offset of the MAC header accordingly, adding *len* bytes of 2369c6b5fb86SQuentin Monnet * space. It automatically extends and reallocates memory as 2370c6b5fb86SQuentin Monnet * required. 2371c6b5fb86SQuentin Monnet * 2372c6b5fb86SQuentin Monnet * This helper can be used on a layer 3 *skb* to push a MAC header 2373c6b5fb86SQuentin Monnet * for redirection into a layer 2 device. 2374c6b5fb86SQuentin Monnet * 2375c6b5fb86SQuentin Monnet * All values for *flags* are reserved for future usage, and must 2376c6b5fb86SQuentin Monnet * be left at zero. 2377c6b5fb86SQuentin Monnet * 237832e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2379c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2380c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 2381c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 2382c6b5fb86SQuentin Monnet * direct packet access. 2383c6b5fb86SQuentin Monnet * Return 2384c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2385c6b5fb86SQuentin Monnet * 2386bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2387c6b5fb86SQuentin Monnet * Description 2388c6b5fb86SQuentin Monnet * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2389c6b5fb86SQuentin Monnet * it is possible to use a negative value for *delta*. This helper 2390c6b5fb86SQuentin Monnet * can be used to prepare the packet for pushing or popping 2391c6b5fb86SQuentin Monnet * headers. 2392c6b5fb86SQuentin Monnet * 239332e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2394c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2395c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 2396c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 2397c6b5fb86SQuentin Monnet * direct packet access. 2398c6b5fb86SQuentin Monnet * Return 2399c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2400c6b5fb86SQuentin Monnet * 2401bdb7b79bSAndrii Nakryiko * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2402c6b5fb86SQuentin Monnet * Description 24036ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe kernel address 2404ab8d7809SQuentin Monnet * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 24056ae08ae3SDaniel Borkmann * more details. 2406c6b5fb86SQuentin Monnet * 2407ab8d7809SQuentin Monnet * Generally, use **bpf_probe_read_user_str**\ () or 2408ab8d7809SQuentin Monnet * **bpf_probe_read_kernel_str**\ () instead. 2409c6b5fb86SQuentin Monnet * Return 2410c6b5fb86SQuentin Monnet * On success, the strictly positive length of the string, 2411c6b5fb86SQuentin Monnet * including the trailing NUL character. On error, a negative 2412c6b5fb86SQuentin Monnet * value. 2413c6b5fb86SQuentin Monnet * 2414c6b5fb86SQuentin Monnet * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2415c6b5fb86SQuentin Monnet * Description 2416c6b5fb86SQuentin Monnet * If the **struct sk_buff** pointed by *skb* has a known socket, 2417c6b5fb86SQuentin Monnet * retrieve the cookie (generated by the kernel) of this socket. 2418c6b5fb86SQuentin Monnet * If no cookie has been set yet, generate a new cookie. Once 2419c6b5fb86SQuentin Monnet * generated, the socket cookie remains stable for the life of the 2420c6b5fb86SQuentin Monnet * socket. This helper can be useful for monitoring per socket 2421cd48bddaSDaniel Borkmann * networking traffic statistics as it provides a global socket 2422cd48bddaSDaniel Borkmann * identifier that can be assumed unique. 2423c6b5fb86SQuentin Monnet * Return 242407881ccbSFlorent Revest * A 8-byte long unique number on success, or 0 if the socket 242507881ccbSFlorent Revest * field is missing inside *skb*. 2426c6b5fb86SQuentin Monnet * 2427d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2428d692f113SAndrey Ignatov * Description 2429d692f113SAndrey Ignatov * Equivalent to bpf_get_socket_cookie() helper that accepts 243062369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_addr** context. 2431d692f113SAndrey Ignatov * Return 243207881ccbSFlorent Revest * A 8-byte long unique number. 2433d692f113SAndrey Ignatov * 2434d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2435d692f113SAndrey Ignatov * Description 2436ab8d7809SQuentin Monnet * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 243762369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_ops** context. 2438d692f113SAndrey Ignatov * Return 243907881ccbSFlorent Revest * A 8-byte long unique number. 2440d692f113SAndrey Ignatov * 2441c5dbb89fSFlorent Revest * u64 bpf_get_socket_cookie(struct sock *sk) 2442c5dbb89fSFlorent Revest * Description 2443c5dbb89fSFlorent Revest * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2444c5dbb89fSFlorent Revest * *sk*, but gets socket from a BTF **struct sock**. This helper 2445c5dbb89fSFlorent Revest * also works for sleepable programs. 2446c5dbb89fSFlorent Revest * Return 2447c5dbb89fSFlorent Revest * A 8-byte long unique number or 0 if *sk* is NULL. 2448c5dbb89fSFlorent Revest * 2449c6b5fb86SQuentin Monnet * u32 bpf_get_socket_uid(struct sk_buff *skb) 2450c6b5fb86SQuentin Monnet * Return 2451c6b5fb86SQuentin Monnet * The owner UID of the socket associated to *skb*. If the socket 2452c6b5fb86SQuentin Monnet * is **NULL**, or if it is not a full socket (i.e. if it is a 2453c6b5fb86SQuentin Monnet * time-wait or a request socket instead), **overflowuid** value 2454c6b5fb86SQuentin Monnet * is returned (note that **overflowuid** might also be the actual 2455c6b5fb86SQuentin Monnet * UID value for the socket). 2456c6b5fb86SQuentin Monnet * 2457bdb7b79bSAndrii Nakryiko * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2458fa15601aSQuentin Monnet * Description 2459fa15601aSQuentin Monnet * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2460fa15601aSQuentin Monnet * to value *hash*. 2461fa15601aSQuentin Monnet * Return 2462fa15601aSQuentin Monnet * 0 2463fa15601aSQuentin Monnet * 2464bdb7b79bSAndrii Nakryiko * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 24657aa79a86SQuentin Monnet * Description 24667aa79a86SQuentin Monnet * Emulate a call to **setsockopt()** on the socket associated to 24677aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 24687aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 24697aa79a86SQuentin Monnet * must be specified, see **setsockopt(2)** for more information. 24707aa79a86SQuentin Monnet * The option value of length *optlen* is pointed by *optval*. 24717aa79a86SQuentin Monnet * 2472beecf11bSStanislav Fomichev * *bpf_socket* should be one of the following: 2473ab8d7809SQuentin Monnet * 2474beecf11bSStanislav Fomichev * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2475beecf11bSStanislav Fomichev * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2476beecf11bSStanislav Fomichev * and **BPF_CGROUP_INET6_CONNECT**. 2477beecf11bSStanislav Fomichev * 24787aa79a86SQuentin Monnet * This helper actually implements a subset of **setsockopt()**. 24797aa79a86SQuentin Monnet * It supports the following *level*\ s: 24807aa79a86SQuentin Monnet * 24817aa79a86SQuentin Monnet * * **SOL_SOCKET**, which supports the following *optname*\ s: 24827aa79a86SQuentin Monnet * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2483f9bcf968SDmitry Yakunin * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2484f9bcf968SDmitry Yakunin * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 24857aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports the following *optname*\ s: 24867aa79a86SQuentin Monnet * **TCP_CONGESTION**, **TCP_BPF_IW**, 2487f9bcf968SDmitry Yakunin * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2488f9bcf968SDmitry Yakunin * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2489eca43ee6SNikita V. Shirokov * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**. 24907aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 24917aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 24927aa79a86SQuentin Monnet * Return 24937aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 24947aa79a86SQuentin Monnet * 2495bdb7b79bSAndrii Nakryiko * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2496fa15601aSQuentin Monnet * Description 2497fa15601aSQuentin Monnet * Grow or shrink the room for data in the packet associated to 2498fa15601aSQuentin Monnet * *skb* by *len_diff*, and according to the selected *mode*. 2499fa15601aSQuentin Monnet * 2500836e66c2SDaniel Borkmann * By default, the helper will reset any offloaded checksum 2501836e66c2SDaniel Borkmann * indicator of the skb to CHECKSUM_NONE. This can be avoided 2502836e66c2SDaniel Borkmann * by the following flag: 2503836e66c2SDaniel Borkmann * 2504836e66c2SDaniel Borkmann * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2505836e66c2SDaniel Borkmann * checksum data of the skb to CHECKSUM_NONE. 2506836e66c2SDaniel Borkmann * 250714aa3192SWillem de Bruijn * There are two supported modes at this time: 250814aa3192SWillem de Bruijn * 250914aa3192SWillem de Bruijn * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 251014aa3192SWillem de Bruijn * (room space is added or removed below the layer 2 header). 2511fa15601aSQuentin Monnet * 2512fa15601aSQuentin Monnet * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2513fa15601aSQuentin Monnet * (room space is added or removed below the layer 3 header). 2514fa15601aSQuentin Monnet * 2515868d5235SWillem de Bruijn * The following flags are supported at this time: 25162278f6ccSWillem de Bruijn * 25172278f6ccSWillem de Bruijn * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 25182278f6ccSWillem de Bruijn * Adjusting mss in this way is not allowed for datagrams. 2519fa15601aSQuentin Monnet * 252080867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 252180867c5eSQuentin Monnet * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2522868d5235SWillem de Bruijn * Any new space is reserved to hold a tunnel header. 2523868d5235SWillem de Bruijn * Configure skb offsets and other fields accordingly. 2524868d5235SWillem de Bruijn * 252580867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 252680867c5eSQuentin Monnet * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2527868d5235SWillem de Bruijn * Use with ENCAP_L3 flags to further specify the tunnel type. 2528868d5235SWillem de Bruijn * 252980867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 253058dfc900SAlan Maguire * Use with ENCAP_L3/L4 flags to further specify the tunnel 253180867c5eSQuentin Monnet * type; *len* is the length of the inner MAC header. 253258dfc900SAlan Maguire * 2533d01b59c9SXuesen Huang * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2534d01b59c9SXuesen Huang * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2535d01b59c9SXuesen Huang * L2 type as Ethernet. 2536d01b59c9SXuesen Huang * 253732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2538fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2539fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2540fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2541fa15601aSQuentin Monnet * direct packet access. 2542fa15601aSQuentin Monnet * Return 2543fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2544fa15601aSQuentin Monnet * 2545bdb7b79bSAndrii Nakryiko * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 2546ab127040SQuentin Monnet * Description 2547ab127040SQuentin Monnet * Redirect the packet to the endpoint referenced by *map* at 2548ab127040SQuentin Monnet * index *key*. Depending on its type, this *map* can contain 2549ab127040SQuentin Monnet * references to net devices (for forwarding packets through other 2550ab127040SQuentin Monnet * ports), or to CPUs (for redirecting XDP frames to another CPU; 2551ab127040SQuentin Monnet * but this is only implemented for native XDP (with driver 2552ab127040SQuentin Monnet * support) as of this writing). 2553ab127040SQuentin Monnet * 255443e74c02SToke Høiland-Jørgensen * The lower two bits of *flags* are used as the return code if 255543e74c02SToke Høiland-Jørgensen * the map lookup fails. This is so that the return value can be 2556ab8d7809SQuentin Monnet * one of the XDP program return codes up to **XDP_TX**, as chosen 2557e624d4edSHangbin Liu * by the caller. The higher bits of *flags* can be set to 2558e624d4edSHangbin Liu * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2559e624d4edSHangbin Liu * 2560e624d4edSHangbin Liu * With BPF_F_BROADCAST the packet will be broadcasted to all the 2561e624d4edSHangbin Liu * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2562e624d4edSHangbin Liu * interface will be excluded when do broadcasting. 2563ab127040SQuentin Monnet * 2564ab8d7809SQuentin Monnet * See also **bpf_redirect**\ (), which only supports redirecting 2565ab8d7809SQuentin Monnet * to an ifindex, but doesn't require a map to do so. 2566ab127040SQuentin Monnet * Return 2567f25975f4SToke Høiland-Jørgensen * **XDP_REDIRECT** on success, or the value of the two lower bits 2568a33d3147SJakub Wilk * of the *flags* argument on error. 2569ab127040SQuentin Monnet * 2570bdb7b79bSAndrii Nakryiko * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2571ab127040SQuentin Monnet * Description 2572ab127040SQuentin Monnet * Redirect the packet to the socket referenced by *map* (of type 2573ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2574ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 2575ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 2576ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 2577ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 2578ab127040SQuentin Monnet * Return 2579ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 2580ab127040SQuentin Monnet * 2581bdb7b79bSAndrii Nakryiko * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2582ab127040SQuentin Monnet * Description 2583ab127040SQuentin Monnet * Add an entry to, or update a *map* referencing sockets. The 2584ab127040SQuentin Monnet * *skops* is used as a new value for the entry associated to 2585ab127040SQuentin Monnet * *key*. *flags* is one of: 2586ab127040SQuentin Monnet * 2587ab127040SQuentin Monnet * **BPF_NOEXIST** 2588ab127040SQuentin Monnet * The entry for *key* must not exist in the map. 2589ab127040SQuentin Monnet * **BPF_EXIST** 2590ab127040SQuentin Monnet * The entry for *key* must already exist in the map. 2591ab127040SQuentin Monnet * **BPF_ANY** 2592ab127040SQuentin Monnet * No condition on the existence of the entry for *key*. 2593ab127040SQuentin Monnet * 2594ab127040SQuentin Monnet * If the *map* has eBPF programs (parser and verdict), those will 2595ab127040SQuentin Monnet * be inherited by the socket being added. If the socket is 2596ab127040SQuentin Monnet * already attached to eBPF programs, this results in an error. 2597ab127040SQuentin Monnet * Return 2598ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 2599ab127040SQuentin Monnet * 2600bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2601fa15601aSQuentin Monnet * Description 2602fa15601aSQuentin Monnet * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2603fa15601aSQuentin Monnet * *delta* (which can be positive or negative). Note that this 2604fa15601aSQuentin Monnet * operation modifies the address stored in *xdp_md*\ **->data**, 2605fa15601aSQuentin Monnet * so the latter must be loaded only after the helper has been 2606fa15601aSQuentin Monnet * called. 2607fa15601aSQuentin Monnet * 2608fa15601aSQuentin Monnet * The use of *xdp_md*\ **->data_meta** is optional and programs 2609fa15601aSQuentin Monnet * are not required to use it. The rationale is that when the 2610fa15601aSQuentin Monnet * packet is processed with XDP (e.g. as DoS filter), it is 2611fa15601aSQuentin Monnet * possible to push further meta data along with it before passing 2612fa15601aSQuentin Monnet * to the stack, and to give the guarantee that an ingress eBPF 2613fa15601aSQuentin Monnet * program attached as a TC classifier on the same device can pick 2614fa15601aSQuentin Monnet * this up for further post-processing. Since TC works with socket 2615fa15601aSQuentin Monnet * buffers, it remains possible to set from XDP the **mark** or 2616fa15601aSQuentin Monnet * **priority** pointers, or other pointers for the socket buffer. 2617fa15601aSQuentin Monnet * Having this scratch space generic and programmable allows for 2618fa15601aSQuentin Monnet * more flexibility as the user is free to store whatever meta 2619fa15601aSQuentin Monnet * data they need. 2620fa15601aSQuentin Monnet * 262132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2622fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2623fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2624fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2625fa15601aSQuentin Monnet * direct packet access. 2626fa15601aSQuentin Monnet * Return 2627fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 26287aa79a86SQuentin Monnet * 2629bdb7b79bSAndrii Nakryiko * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 26307aa79a86SQuentin Monnet * Description 26317aa79a86SQuentin Monnet * Read the value of a perf event counter, and store it into *buf* 26327aa79a86SQuentin Monnet * of size *buf_size*. This helper relies on a *map* of type 26337aa79a86SQuentin Monnet * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 26347aa79a86SQuentin Monnet * counter is selected when *map* is updated with perf event file 26357aa79a86SQuentin Monnet * descriptors. The *map* is an array whose size is the number of 26367aa79a86SQuentin Monnet * available CPUs, and each cell contains a value relative to one 26377aa79a86SQuentin Monnet * CPU. The value to retrieve is indicated by *flags*, that 26387aa79a86SQuentin Monnet * contains the index of the CPU to look up, masked with 26397aa79a86SQuentin Monnet * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 26407aa79a86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 26417aa79a86SQuentin Monnet * current CPU should be retrieved. 26427aa79a86SQuentin Monnet * 26437aa79a86SQuentin Monnet * This helper behaves in a way close to 26447aa79a86SQuentin Monnet * **bpf_perf_event_read**\ () helper, save that instead of 26457aa79a86SQuentin Monnet * just returning the value observed, it fills the *buf* 26467aa79a86SQuentin Monnet * structure. This allows for additional data to be retrieved: in 26477aa79a86SQuentin Monnet * particular, the enabled and running times (in *buf*\ 26487aa79a86SQuentin Monnet * **->enabled** and *buf*\ **->running**, respectively) are 26497aa79a86SQuentin Monnet * copied. In general, **bpf_perf_event_read_value**\ () is 26507aa79a86SQuentin Monnet * recommended over **bpf_perf_event_read**\ (), which has some 26517aa79a86SQuentin Monnet * ABI issues and provides fewer functionalities. 26527aa79a86SQuentin Monnet * 26537aa79a86SQuentin Monnet * These values are interesting, because hardware PMU (Performance 26547aa79a86SQuentin Monnet * Monitoring Unit) counters are limited resources. When there are 26557aa79a86SQuentin Monnet * more PMU based perf events opened than available counters, 26567aa79a86SQuentin Monnet * kernel will multiplex these events so each event gets certain 26577aa79a86SQuentin Monnet * percentage (but not all) of the PMU time. In case that 26587aa79a86SQuentin Monnet * multiplexing happens, the number of samples or counter value 26597aa79a86SQuentin Monnet * will not reflect the case compared to when no multiplexing 26607aa79a86SQuentin Monnet * occurs. This makes comparison between different runs difficult. 26617aa79a86SQuentin Monnet * Typically, the counter value should be normalized before 26627aa79a86SQuentin Monnet * comparing to other experiments. The usual normalization is done 26637aa79a86SQuentin Monnet * as follows. 26647aa79a86SQuentin Monnet * 26657aa79a86SQuentin Monnet * :: 26667aa79a86SQuentin Monnet * 26677aa79a86SQuentin Monnet * normalized_counter = counter * t_enabled / t_running 26687aa79a86SQuentin Monnet * 26697aa79a86SQuentin Monnet * Where t_enabled is the time enabled for event and t_running is 26707aa79a86SQuentin Monnet * the time running for event since last normalization. The 26717aa79a86SQuentin Monnet * enabled and running times are accumulated since the perf event 26727aa79a86SQuentin Monnet * open. To achieve scaling factor between two invocations of an 2673ab8d7809SQuentin Monnet * eBPF program, users can use CPU id as the key (which is 26747aa79a86SQuentin Monnet * typical for perf array usage model) to remember the previous 26757aa79a86SQuentin Monnet * value and do the calculation inside the eBPF program. 26767aa79a86SQuentin Monnet * Return 26777aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 26787aa79a86SQuentin Monnet * 2679bdb7b79bSAndrii Nakryiko * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 26807aa79a86SQuentin Monnet * Description 26817aa79a86SQuentin Monnet * For en eBPF program attached to a perf event, retrieve the 26827aa79a86SQuentin Monnet * value of the event counter associated to *ctx* and store it in 26837aa79a86SQuentin Monnet * the structure pointed by *buf* and of size *buf_size*. Enabled 26847aa79a86SQuentin Monnet * and running times are also stored in the structure (see 26857aa79a86SQuentin Monnet * description of helper **bpf_perf_event_read_value**\ () for 26867aa79a86SQuentin Monnet * more details). 26877aa79a86SQuentin Monnet * Return 26887aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 26897aa79a86SQuentin Monnet * 2690bdb7b79bSAndrii Nakryiko * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 26917aa79a86SQuentin Monnet * Description 26927aa79a86SQuentin Monnet * Emulate a call to **getsockopt()** on the socket associated to 26937aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 26947aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 26957aa79a86SQuentin Monnet * must be specified, see **getsockopt(2)** for more information. 26967aa79a86SQuentin Monnet * The retrieved value is stored in the structure pointed by 26977aa79a86SQuentin Monnet * *opval* and of length *optlen*. 26987aa79a86SQuentin Monnet * 2699beecf11bSStanislav Fomichev * *bpf_socket* should be one of the following: 2700ab8d7809SQuentin Monnet * 2701beecf11bSStanislav Fomichev * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2702beecf11bSStanislav Fomichev * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2703beecf11bSStanislav Fomichev * and **BPF_CGROUP_INET6_CONNECT**. 2704beecf11bSStanislav Fomichev * 27057aa79a86SQuentin Monnet * This helper actually implements a subset of **getsockopt()**. 27067aa79a86SQuentin Monnet * It supports the following *level*\ s: 27077aa79a86SQuentin Monnet * 27087aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports *optname* 27097aa79a86SQuentin Monnet * **TCP_CONGESTION**. 27107aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 27117aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 27127aa79a86SQuentin Monnet * Return 27137aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 27147aa79a86SQuentin Monnet * 2715bdb7b79bSAndrii Nakryiko * long bpf_override_return(struct pt_regs *regs, u64 rc) 27167aa79a86SQuentin Monnet * Description 27177aa79a86SQuentin Monnet * Used for error injection, this helper uses kprobes to override 27187aa79a86SQuentin Monnet * the return value of the probed function, and to set it to *rc*. 27197aa79a86SQuentin Monnet * The first argument is the context *regs* on which the kprobe 27207aa79a86SQuentin Monnet * works. 27217aa79a86SQuentin Monnet * 2722ab8d7809SQuentin Monnet * This helper works by setting the PC (program counter) 27237aa79a86SQuentin Monnet * to an override function which is run in place of the original 27247aa79a86SQuentin Monnet * probed function. This means the probed function is not run at 27257aa79a86SQuentin Monnet * all. The replacement function just returns with the required 27267aa79a86SQuentin Monnet * value. 27277aa79a86SQuentin Monnet * 27287aa79a86SQuentin Monnet * This helper has security implications, and thus is subject to 27297aa79a86SQuentin Monnet * restrictions. It is only available if the kernel was compiled 27307aa79a86SQuentin Monnet * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 27317aa79a86SQuentin Monnet * option, and in this case it only works on functions tagged with 27327aa79a86SQuentin Monnet * **ALLOW_ERROR_INJECTION** in the kernel code. 27337aa79a86SQuentin Monnet * 27347aa79a86SQuentin Monnet * Also, the helper is only available for the architectures having 27357aa79a86SQuentin Monnet * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 27367aa79a86SQuentin Monnet * x86 architecture is the only one to support this feature. 27377aa79a86SQuentin Monnet * Return 27387aa79a86SQuentin Monnet * 0 27397aa79a86SQuentin Monnet * 2740bdb7b79bSAndrii Nakryiko * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 27417aa79a86SQuentin Monnet * Description 27427aa79a86SQuentin Monnet * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 27437aa79a86SQuentin Monnet * for the full TCP socket associated to *bpf_sock_ops* to 27447aa79a86SQuentin Monnet * *argval*. 27457aa79a86SQuentin Monnet * 27467aa79a86SQuentin Monnet * The primary use of this field is to determine if there should 27477aa79a86SQuentin Monnet * be calls to eBPF programs of type 27487aa79a86SQuentin Monnet * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 27497aa79a86SQuentin Monnet * code. A program of the same type can change its value, per 27507aa79a86SQuentin Monnet * connection and as necessary, when the connection is 27517aa79a86SQuentin Monnet * established. This field is directly accessible for reading, but 27527aa79a86SQuentin Monnet * this helper must be used for updates in order to return an 27537aa79a86SQuentin Monnet * error if an eBPF program tries to set a callback that is not 27547aa79a86SQuentin Monnet * supported in the current kernel. 27557aa79a86SQuentin Monnet * 2756725721a6SViet Hoang Tran * *argval* is a flag array which can combine these flags: 27577aa79a86SQuentin Monnet * 27587aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 27597aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 27607aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 276123729ff2SStanislav Fomichev * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 27627aa79a86SQuentin Monnet * 2763725721a6SViet Hoang Tran * Therefore, this function can be used to clear a callback flag by 2764725721a6SViet Hoang Tran * setting the appropriate bit to zero. e.g. to disable the RTO 2765725721a6SViet Hoang Tran * callback: 2766725721a6SViet Hoang Tran * 2767725721a6SViet Hoang Tran * **bpf_sock_ops_cb_flags_set(bpf_sock,** 2768725721a6SViet Hoang Tran * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 2769725721a6SViet Hoang Tran * 27707aa79a86SQuentin Monnet * Here are some examples of where one could call such eBPF 27717aa79a86SQuentin Monnet * program: 27727aa79a86SQuentin Monnet * 27737aa79a86SQuentin Monnet * * When RTO fires. 27747aa79a86SQuentin Monnet * * When a packet is retransmitted. 27757aa79a86SQuentin Monnet * * When the connection terminates. 27767aa79a86SQuentin Monnet * * When a packet is sent. 27777aa79a86SQuentin Monnet * * When a packet is received. 27787aa79a86SQuentin Monnet * Return 27797aa79a86SQuentin Monnet * Code **-EINVAL** if the socket is not a full TCP socket; 27807aa79a86SQuentin Monnet * otherwise, a positive number containing the bits that could not 27817aa79a86SQuentin Monnet * be set is returned (which comes down to 0 if all bits were set 27827aa79a86SQuentin Monnet * as required). 27837aa79a86SQuentin Monnet * 2784bdb7b79bSAndrii Nakryiko * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 2785ab127040SQuentin Monnet * Description 2786ab127040SQuentin Monnet * This helper is used in programs implementing policies at the 2787ab127040SQuentin Monnet * socket level. If the message *msg* is allowed to pass (i.e. if 2788ab127040SQuentin Monnet * the verdict eBPF program returns **SK_PASS**), redirect it to 2789ab127040SQuentin Monnet * the socket referenced by *map* (of type 2790ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2791ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 2792ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 2793ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 2794ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 2795ab127040SQuentin Monnet * Return 2796ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 2797ab127040SQuentin Monnet * 2798bdb7b79bSAndrii Nakryiko * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 2799ab127040SQuentin Monnet * Description 2800ab127040SQuentin Monnet * For socket policies, apply the verdict of the eBPF program to 2801ab127040SQuentin Monnet * the next *bytes* (number of bytes) of message *msg*. 2802ab127040SQuentin Monnet * 2803ab127040SQuentin Monnet * For example, this helper can be used in the following cases: 2804ab127040SQuentin Monnet * 2805ab127040SQuentin Monnet * * A single **sendmsg**\ () or **sendfile**\ () system call 2806ab127040SQuentin Monnet * contains multiple logical messages that the eBPF program is 2807ab127040SQuentin Monnet * supposed to read and for which it should apply a verdict. 2808ab127040SQuentin Monnet * * An eBPF program only cares to read the first *bytes* of a 2809ab127040SQuentin Monnet * *msg*. If the message has a large payload, then setting up 2810ab127040SQuentin Monnet * and calling the eBPF program repeatedly for all bytes, even 2811ab127040SQuentin Monnet * though the verdict is already known, would create unnecessary 2812ab127040SQuentin Monnet * overhead. 2813ab127040SQuentin Monnet * 2814ab127040SQuentin Monnet * When called from within an eBPF program, the helper sets a 2815ab127040SQuentin Monnet * counter internal to the BPF infrastructure, that is used to 2816ab127040SQuentin Monnet * apply the last verdict to the next *bytes*. If *bytes* is 2817ab127040SQuentin Monnet * smaller than the current data being processed from a 2818ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () system call, the first 2819ab127040SQuentin Monnet * *bytes* will be sent and the eBPF program will be re-run with 2820ab127040SQuentin Monnet * the pointer for start of data pointing to byte number *bytes* 2821ab127040SQuentin Monnet * **+ 1**. If *bytes* is larger than the current data being 2822ab127040SQuentin Monnet * processed, then the eBPF verdict will be applied to multiple 2823ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2824ab127040SQuentin Monnet * consumed. 2825ab127040SQuentin Monnet * 2826ab127040SQuentin Monnet * Note that if a socket closes with the internal counter holding 2827ab127040SQuentin Monnet * a non-zero value, this is not a problem because data is not 2828ab127040SQuentin Monnet * being buffered for *bytes* and is sent as it is received. 2829ab127040SQuentin Monnet * Return 2830ab127040SQuentin Monnet * 0 2831ab127040SQuentin Monnet * 2832bdb7b79bSAndrii Nakryiko * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 2833ab127040SQuentin Monnet * Description 2834ab127040SQuentin Monnet * For socket policies, prevent the execution of the verdict eBPF 2835ab127040SQuentin Monnet * program for message *msg* until *bytes* (byte number) have been 2836ab127040SQuentin Monnet * accumulated. 2837ab127040SQuentin Monnet * 2838ab127040SQuentin Monnet * This can be used when one needs a specific number of bytes 2839ab127040SQuentin Monnet * before a verdict can be assigned, even if the data spans 2840ab127040SQuentin Monnet * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 2841ab127040SQuentin Monnet * case would be a user calling **sendmsg**\ () repeatedly with 2842ab127040SQuentin Monnet * 1-byte long message segments. Obviously, this is bad for 2843ab127040SQuentin Monnet * performance, but it is still valid. If the eBPF program needs 2844ab127040SQuentin Monnet * *bytes* bytes to validate a header, this helper can be used to 2845ab127040SQuentin Monnet * prevent the eBPF program to be called again until *bytes* have 2846ab127040SQuentin Monnet * been accumulated. 2847ab127040SQuentin Monnet * Return 2848ab127040SQuentin Monnet * 0 2849ab127040SQuentin Monnet * 2850bdb7b79bSAndrii Nakryiko * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2851ab127040SQuentin Monnet * Description 2852ab127040SQuentin Monnet * For socket policies, pull in non-linear data from user space 2853ab127040SQuentin Monnet * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2854ab127040SQuentin Monnet * **->data_end** to *start* and *end* bytes offsets into *msg*, 2855ab127040SQuentin Monnet * respectively. 2856ab127040SQuentin Monnet * 2857ab127040SQuentin Monnet * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2858ab127040SQuentin Monnet * *msg* it can only parse data that the (**data**, **data_end**) 2859ab127040SQuentin Monnet * pointers have already consumed. For **sendmsg**\ () hooks this 2860ab127040SQuentin Monnet * is likely the first scatterlist element. But for calls relying 2861ab127040SQuentin Monnet * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2862ab127040SQuentin Monnet * be the range (**0**, **0**) because the data is shared with 2863ab127040SQuentin Monnet * user space and by default the objective is to avoid allowing 2864ab127040SQuentin Monnet * user space to modify data while (or after) eBPF verdict is 2865ab127040SQuentin Monnet * being decided. This helper can be used to pull in data and to 2866ab127040SQuentin Monnet * set the start and end pointer to given values. Data will be 2867ab127040SQuentin Monnet * copied if necessary (i.e. if data was not linear and if start 2868ab127040SQuentin Monnet * and end pointers do not point to the same chunk). 2869ab127040SQuentin Monnet * 287032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2871ab127040SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2872ab127040SQuentin Monnet * previously done by the verifier are invalidated and must be 2873ab127040SQuentin Monnet * performed again, if the helper is used in combination with 2874ab127040SQuentin Monnet * direct packet access. 2875ab127040SQuentin Monnet * 2876ab127040SQuentin Monnet * All values for *flags* are reserved for future usage, and must 2877ab127040SQuentin Monnet * be left at zero. 2878ab127040SQuentin Monnet * Return 2879ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 2880ab127040SQuentin Monnet * 2881bdb7b79bSAndrii Nakryiko * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 28827aa79a86SQuentin Monnet * Description 28837aa79a86SQuentin Monnet * Bind the socket associated to *ctx* to the address pointed by 28847aa79a86SQuentin Monnet * *addr*, of length *addr_len*. This allows for making outgoing 28857aa79a86SQuentin Monnet * connection from the desired IP address, which can be useful for 28867aa79a86SQuentin Monnet * example when all processes inside a cgroup should use one 28877aa79a86SQuentin Monnet * single IP address on a host that has multiple IP configured. 28887aa79a86SQuentin Monnet * 28897aa79a86SQuentin Monnet * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 28907aa79a86SQuentin Monnet * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 28918086fbafSStanislav Fomichev * **AF_INET6**). It's advised to pass zero port (**sin_port** 28928086fbafSStanislav Fomichev * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 28938086fbafSStanislav Fomichev * behavior and lets the kernel efficiently pick up an unused 28948086fbafSStanislav Fomichev * port as long as 4-tuple is unique. Passing non-zero port might 28958086fbafSStanislav Fomichev * lead to degraded performance. 28967aa79a86SQuentin Monnet * Return 28977aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 28982d020dd7SQuentin Monnet * 2899bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 29002d020dd7SQuentin Monnet * Description 29012d020dd7SQuentin Monnet * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2902c8741e2bSJesper Dangaard Brouer * possible to both shrink and grow the packet tail. 2903c8741e2bSJesper Dangaard Brouer * Shrink done via *delta* being a negative integer. 29042d020dd7SQuentin Monnet * 290532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 29062d020dd7SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 29072d020dd7SQuentin Monnet * previously done by the verifier are invalidated and must be 29082d020dd7SQuentin Monnet * performed again, if the helper is used in combination with 29092d020dd7SQuentin Monnet * direct packet access. 29102d020dd7SQuentin Monnet * Return 29112d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 29122d020dd7SQuentin Monnet * 2913bdb7b79bSAndrii Nakryiko * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 29142d020dd7SQuentin Monnet * Description 29152d020dd7SQuentin Monnet * Retrieve the XFRM state (IP transform framework, see also 29162d020dd7SQuentin Monnet * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 29172d020dd7SQuentin Monnet * 29182d020dd7SQuentin Monnet * The retrieved value is stored in the **struct bpf_xfrm_state** 29192d020dd7SQuentin Monnet * pointed by *xfrm_state* and of length *size*. 29202d020dd7SQuentin Monnet * 29212d020dd7SQuentin Monnet * All values for *flags* are reserved for future usage, and must 29222d020dd7SQuentin Monnet * be left at zero. 29232d020dd7SQuentin Monnet * 29242d020dd7SQuentin Monnet * This helper is available only if the kernel was compiled with 29252d020dd7SQuentin Monnet * **CONFIG_XFRM** configuration option. 29262d020dd7SQuentin Monnet * Return 29272d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 2928c195651eSYonghong Song * 2929bdb7b79bSAndrii Nakryiko * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2930c195651eSYonghong Song * Description 2931c195651eSYonghong Song * Return a user or a kernel stack in bpf program provided buffer. 2932c195651eSYonghong Song * To achieve this, the helper needs *ctx*, which is a pointer 2933c195651eSYonghong Song * to the context on which the tracing program is executed. 2934c195651eSYonghong Song * To store the stacktrace, the bpf program provides *buf* with 2935c195651eSYonghong Song * a nonnegative *size*. 2936c195651eSYonghong Song * 2937c195651eSYonghong Song * The last argument, *flags*, holds the number of stack frames to 2938c195651eSYonghong Song * skip (from 0 to 255), masked with 2939c195651eSYonghong Song * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2940c195651eSYonghong Song * the following flags: 2941c195651eSYonghong Song * 2942c195651eSYonghong Song * **BPF_F_USER_STACK** 2943c195651eSYonghong Song * Collect a user space stack instead of a kernel stack. 2944c195651eSYonghong Song * **BPF_F_USER_BUILD_ID** 2945c195651eSYonghong Song * Collect buildid+offset instead of ips for user stack, 2946c195651eSYonghong Song * only valid if **BPF_F_USER_STACK** is also specified. 2947c195651eSYonghong Song * 2948c195651eSYonghong Song * **bpf_get_stack**\ () can collect up to 2949c195651eSYonghong Song * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 2950c195651eSYonghong Song * to sufficient large buffer size. Note that 2951c195651eSYonghong Song * this limit can be controlled with the **sysctl** program, and 2952c195651eSYonghong Song * that it should be manually increased in order to profile long 2953c195651eSYonghong Song * user stacks (such as stacks for Java programs). To do so, use: 2954c195651eSYonghong Song * 2955c195651eSYonghong Song * :: 2956c195651eSYonghong Song * 2957c195651eSYonghong Song * # sysctl kernel.perf_event_max_stack=<new value> 2958c195651eSYonghong Song * Return 29597a279e93SQuentin Monnet * A non-negative value equal to or less than *size* on success, 29607a279e93SQuentin Monnet * or a negative error in case of failure. 29614e1ec56cSDaniel Borkmann * 2962bdb7b79bSAndrii Nakryiko * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 29634e1ec56cSDaniel Borkmann * Description 29644e1ec56cSDaniel Borkmann * This helper is similar to **bpf_skb_load_bytes**\ () in that 29654e1ec56cSDaniel Borkmann * it provides an easy way to load *len* bytes from *offset* 29664e1ec56cSDaniel Borkmann * from the packet associated to *skb*, into the buffer pointed 29674e1ec56cSDaniel Borkmann * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 29684e1ec56cSDaniel Borkmann * a fifth argument *start_header* exists in order to select a 29694e1ec56cSDaniel Borkmann * base offset to start from. *start_header* can be one of: 29704e1ec56cSDaniel Borkmann * 29714e1ec56cSDaniel Borkmann * **BPF_HDR_START_MAC** 29724e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s mac header. 29734e1ec56cSDaniel Borkmann * **BPF_HDR_START_NET** 29744e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s network header. 29754e1ec56cSDaniel Borkmann * 29764e1ec56cSDaniel Borkmann * In general, "direct packet access" is the preferred method to 29774e1ec56cSDaniel Borkmann * access packet data, however, this helper is in particular useful 29784e1ec56cSDaniel Borkmann * in socket filters where *skb*\ **->data** does not always point 29794e1ec56cSDaniel Borkmann * to the start of the mac header and where "direct packet access" 29804e1ec56cSDaniel Borkmann * is not available. 29814e1ec56cSDaniel Borkmann * Return 29824e1ec56cSDaniel Borkmann * 0 on success, or a negative error in case of failure. 29834e1ec56cSDaniel Borkmann * 2984bdb7b79bSAndrii Nakryiko * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 298587f5fc7eSDavid Ahern * Description 298687f5fc7eSDavid Ahern * Do FIB lookup in kernel tables using parameters in *params*. 298787f5fc7eSDavid Ahern * If lookup is successful and result shows packet is to be 298887f5fc7eSDavid Ahern * forwarded, the neighbor tables are searched for the nexthop. 298987f5fc7eSDavid Ahern * If successful (ie., FIB lookup shows forwarding and nexthop 2990fa898d76SDavid Ahern * is resolved), the nexthop address is returned in ipv4_dst 2991fa898d76SDavid Ahern * or ipv6_dst based on family, smac is set to mac address of 2992fa898d76SDavid Ahern * egress device, dmac is set to nexthop mac address, rt_metric 29934c79579bSDavid Ahern * is set to metric from route (IPv4/IPv6 only), and ifindex 29944c79579bSDavid Ahern * is set to the device index of the nexthop from the FIB lookup. 299587f5fc7eSDavid Ahern * 299687f5fc7eSDavid Ahern * *plen* argument is the size of the passed in struct. 29977a279e93SQuentin Monnet * *flags* argument can be a combination of one or more of the 29987a279e93SQuentin Monnet * following values: 299987f5fc7eSDavid Ahern * 30007a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_DIRECT** 30017a279e93SQuentin Monnet * Do a direct table lookup vs full lookup using FIB 30027a279e93SQuentin Monnet * rules. 30037a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_OUTPUT** 30047a279e93SQuentin Monnet * Perform lookup from an egress perspective (default is 30057a279e93SQuentin Monnet * ingress). 300687f5fc7eSDavid Ahern * 300787f5fc7eSDavid Ahern * *ctx* is either **struct xdp_md** for XDP programs or 300887f5fc7eSDavid Ahern * **struct sk_buff** tc cls_act programs. 300987f5fc7eSDavid Ahern * Return 30104c79579bSDavid Ahern * * < 0 if any input argument is invalid 30114c79579bSDavid Ahern * * 0 on success (packet is forwarded, nexthop neighbor exists) 30124c79579bSDavid Ahern * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 30132bae79d2SQuentin Monnet * packet is not forwarded or needs assist from full stack 301481110384SJohn Fastabend * 3015e1850ea9SJesper Dangaard Brouer * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3016e1850ea9SJesper Dangaard Brouer * was exceeded and output params->mtu_result contains the MTU. 3017e1850ea9SJesper Dangaard Brouer * 3018bdb7b79bSAndrii Nakryiko * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 301981110384SJohn Fastabend * Description 302081110384SJohn Fastabend * Add an entry to, or update a sockhash *map* referencing sockets. 302181110384SJohn Fastabend * The *skops* is used as a new value for the entry associated to 302281110384SJohn Fastabend * *key*. *flags* is one of: 302381110384SJohn Fastabend * 302481110384SJohn Fastabend * **BPF_NOEXIST** 302581110384SJohn Fastabend * The entry for *key* must not exist in the map. 302681110384SJohn Fastabend * **BPF_EXIST** 302781110384SJohn Fastabend * The entry for *key* must already exist in the map. 302881110384SJohn Fastabend * **BPF_ANY** 302981110384SJohn Fastabend * No condition on the existence of the entry for *key*. 303081110384SJohn Fastabend * 303181110384SJohn Fastabend * If the *map* has eBPF programs (parser and verdict), those will 303281110384SJohn Fastabend * be inherited by the socket being added. If the socket is 303381110384SJohn Fastabend * already attached to eBPF programs, this results in an error. 303481110384SJohn Fastabend * Return 303581110384SJohn Fastabend * 0 on success, or a negative error in case of failure. 303681110384SJohn Fastabend * 3037bdb7b79bSAndrii Nakryiko * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 303881110384SJohn Fastabend * Description 303981110384SJohn Fastabend * This helper is used in programs implementing policies at the 304081110384SJohn Fastabend * socket level. If the message *msg* is allowed to pass (i.e. if 304181110384SJohn Fastabend * the verdict eBPF program returns **SK_PASS**), redirect it to 304281110384SJohn Fastabend * the socket referenced by *map* (of type 304381110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 304481110384SJohn Fastabend * egress interfaces can be used for redirection. The 304581110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 304681110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 304781110384SJohn Fastabend * egress path otherwise). This is the only flag supported for now. 304881110384SJohn Fastabend * Return 304981110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 305081110384SJohn Fastabend * 3051bdb7b79bSAndrii Nakryiko * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 305281110384SJohn Fastabend * Description 305381110384SJohn Fastabend * This helper is used in programs implementing policies at the 305481110384SJohn Fastabend * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 305549f3d12bSJakub Wilk * if the verdict eBPF program returns **SK_PASS**), redirect it 305681110384SJohn Fastabend * to the socket referenced by *map* (of type 305781110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 305881110384SJohn Fastabend * egress interfaces can be used for redirection. The 305981110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 306081110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 306181110384SJohn Fastabend * egress otherwise). This is the only flag supported for now. 306281110384SJohn Fastabend * Return 306381110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 3064fe94cc29SMathieu Xhonneux * 3065bdb7b79bSAndrii Nakryiko * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3066fe94cc29SMathieu Xhonneux * Description 3067fe94cc29SMathieu Xhonneux * Encapsulate the packet associated to *skb* within a Layer 3 3068fe94cc29SMathieu Xhonneux * protocol header. This header is provided in the buffer at 3069fe94cc29SMathieu Xhonneux * address *hdr*, with *len* its size in bytes. *type* indicates 3070fe94cc29SMathieu Xhonneux * the protocol of the header and can be one of: 3071fe94cc29SMathieu Xhonneux * 3072fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6** 3073fe94cc29SMathieu Xhonneux * IPv6 encapsulation with Segment Routing Header 3074fe94cc29SMathieu Xhonneux * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3075fe94cc29SMathieu Xhonneux * the IPv6 header is computed by the kernel. 3076fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6_INLINE** 3077fe94cc29SMathieu Xhonneux * Only works if *skb* contains an IPv6 packet. Insert a 3078fe94cc29SMathieu Xhonneux * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3079fe94cc29SMathieu Xhonneux * the IPv6 header. 30803e0bd37cSPeter Oskolkov * **BPF_LWT_ENCAP_IP** 30813e0bd37cSPeter Oskolkov * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 30823e0bd37cSPeter Oskolkov * must be IPv4 or IPv6, followed by zero or more 308380867c5eSQuentin Monnet * additional headers, up to **LWT_BPF_MAX_HEADROOM** 308480867c5eSQuentin Monnet * total bytes in all prepended headers. Please note that 308580867c5eSQuentin Monnet * if **skb_is_gso**\ (*skb*) is true, no more than two 308680867c5eSQuentin Monnet * headers can be prepended, and the inner header, if 308780867c5eSQuentin Monnet * present, should be either GRE or UDP/GUE. 30883e0bd37cSPeter Oskolkov * 308980867c5eSQuentin Monnet * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 309080867c5eSQuentin Monnet * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 309180867c5eSQuentin Monnet * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 309280867c5eSQuentin Monnet * **BPF_PROG_TYPE_LWT_XMIT**. 3093fe94cc29SMathieu Xhonneux * 309432e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3095fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3096fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3097fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3098fe94cc29SMathieu Xhonneux * direct packet access. 3099fe94cc29SMathieu Xhonneux * Return 3100fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3101fe94cc29SMathieu Xhonneux * 3102bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3103fe94cc29SMathieu Xhonneux * Description 3104fe94cc29SMathieu Xhonneux * Store *len* bytes from address *from* into the packet 3105fe94cc29SMathieu Xhonneux * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3106fe94cc29SMathieu Xhonneux * inside the outermost IPv6 Segment Routing Header can be 3107fe94cc29SMathieu Xhonneux * modified through this helper. 3108fe94cc29SMathieu Xhonneux * 310932e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3110fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3111fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3112fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3113fe94cc29SMathieu Xhonneux * direct packet access. 3114fe94cc29SMathieu Xhonneux * Return 3115fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3116fe94cc29SMathieu Xhonneux * 3117bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3118fe94cc29SMathieu Xhonneux * Description 3119fe94cc29SMathieu Xhonneux * Adjust the size allocated to TLVs in the outermost IPv6 3120fe94cc29SMathieu Xhonneux * Segment Routing Header contained in the packet associated to 3121fe94cc29SMathieu Xhonneux * *skb*, at position *offset* by *delta* bytes. Only offsets 3122fe94cc29SMathieu Xhonneux * after the segments are accepted. *delta* can be as well 3123fe94cc29SMathieu Xhonneux * positive (growing) as negative (shrinking). 3124fe94cc29SMathieu Xhonneux * 312532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3126fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3127fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3128fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3129fe94cc29SMathieu Xhonneux * direct packet access. 3130fe94cc29SMathieu Xhonneux * Return 3131fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3132fe94cc29SMathieu Xhonneux * 3133bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3134fe94cc29SMathieu Xhonneux * Description 3135fe94cc29SMathieu Xhonneux * Apply an IPv6 Segment Routing action of type *action* to the 3136fe94cc29SMathieu Xhonneux * packet associated to *skb*. Each action takes a parameter 3137fe94cc29SMathieu Xhonneux * contained at address *param*, and of length *param_len* bytes. 3138fe94cc29SMathieu Xhonneux * *action* can be one of: 3139fe94cc29SMathieu Xhonneux * 3140fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_X** 3141fe94cc29SMathieu Xhonneux * End.X action: Endpoint with Layer-3 cross-connect. 3142fe94cc29SMathieu Xhonneux * Type of *param*: **struct in6_addr**. 3143fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_T** 3144fe94cc29SMathieu Xhonneux * End.T action: Endpoint with specific IPv6 table lookup. 3145fe94cc29SMathieu Xhonneux * Type of *param*: **int**. 3146fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6** 3147fe94cc29SMathieu Xhonneux * End.B6 action: Endpoint bound to an SRv6 policy. 314880867c5eSQuentin Monnet * Type of *param*: **struct ipv6_sr_hdr**. 3149fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3150fe94cc29SMathieu Xhonneux * End.B6.Encap action: Endpoint bound to an SRv6 3151fe94cc29SMathieu Xhonneux * encapsulation policy. 315280867c5eSQuentin Monnet * Type of *param*: **struct ipv6_sr_hdr**. 3153fe94cc29SMathieu Xhonneux * 315432e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3155fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3156fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3157fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3158fe94cc29SMathieu Xhonneux * direct packet access. 3159fe94cc29SMathieu Xhonneux * Return 3160fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3161f4364dcfSSean Young * 3162bdb7b79bSAndrii Nakryiko * long bpf_rc_repeat(void *ctx) 316362369db2SQuentin Monnet * Description 316462369db2SQuentin Monnet * This helper is used in programs implementing IR decoding, to 316562369db2SQuentin Monnet * report a successfully decoded repeat key message. This delays 316662369db2SQuentin Monnet * the generation of a key up event for previously generated 316762369db2SQuentin Monnet * key down event. 316862369db2SQuentin Monnet * 316962369db2SQuentin Monnet * Some IR protocols like NEC have a special IR message for 317062369db2SQuentin Monnet * repeating last button, for when a button is held down. 317162369db2SQuentin Monnet * 317262369db2SQuentin Monnet * The *ctx* should point to the lirc sample as passed into 317362369db2SQuentin Monnet * the program. 317462369db2SQuentin Monnet * 317562369db2SQuentin Monnet * This helper is only available is the kernel was compiled with 317662369db2SQuentin Monnet * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 317762369db2SQuentin Monnet * "**y**". 317862369db2SQuentin Monnet * Return 317962369db2SQuentin Monnet * 0 318062369db2SQuentin Monnet * 3181bdb7b79bSAndrii Nakryiko * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3182f4364dcfSSean Young * Description 3183f4364dcfSSean Young * This helper is used in programs implementing IR decoding, to 3184f4364dcfSSean Young * report a successfully decoded key press with *scancode*, 3185f4364dcfSSean Young * *toggle* value in the given *protocol*. The scancode will be 3186f4364dcfSSean Young * translated to a keycode using the rc keymap, and reported as 3187f4364dcfSSean Young * an input key down event. After a period a key up event is 3188f4364dcfSSean Young * generated. This period can be extended by calling either 318990b1023fSQuentin Monnet * **bpf_rc_keydown**\ () again with the same values, or calling 319090b1023fSQuentin Monnet * **bpf_rc_repeat**\ (). 3191f4364dcfSSean Young * 3192f4364dcfSSean Young * Some protocols include a toggle bit, in case the button was 3193f4364dcfSSean Young * released and pressed again between consecutive scancodes. 3194f4364dcfSSean Young * 3195f4364dcfSSean Young * The *ctx* should point to the lirc sample as passed into 3196f4364dcfSSean Young * the program. 3197f4364dcfSSean Young * 3198f4364dcfSSean Young * The *protocol* is the decoded protocol number (see 3199f4364dcfSSean Young * **enum rc_proto** for some predefined values). 3200f4364dcfSSean Young * 3201f4364dcfSSean Young * This helper is only available is the kernel was compiled with 3202f4364dcfSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3203f4364dcfSSean Young * "**y**". 3204f4364dcfSSean Young * Return 3205f4364dcfSSean Young * 0 3206f4364dcfSSean Young * 320762369db2SQuentin Monnet * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3208cb20b08eSDaniel Borkmann * Description 3209cb20b08eSDaniel Borkmann * Return the cgroup v2 id of the socket associated with the *skb*. 3210cb20b08eSDaniel Borkmann * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3211cb20b08eSDaniel Borkmann * helper for cgroup v1 by providing a tag resp. identifier that 3212cb20b08eSDaniel Borkmann * can be matched on or used for map lookups e.g. to implement 3213cb20b08eSDaniel Borkmann * policy. The cgroup v2 id of a given path in the hierarchy is 3214cb20b08eSDaniel Borkmann * exposed in user space through the f_handle API in order to get 3215cb20b08eSDaniel Borkmann * to the same 64-bit id. 3216cb20b08eSDaniel Borkmann * 3217cb20b08eSDaniel Borkmann * This helper can be used on TC egress path, but not on ingress, 3218cb20b08eSDaniel Borkmann * and is available only if the kernel was compiled with the 3219cb20b08eSDaniel Borkmann * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3220cb20b08eSDaniel Borkmann * Return 3221cb20b08eSDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 3222bf6fa2c8SYonghong Song * 3223bf6fa2c8SYonghong Song * u64 bpf_get_current_cgroup_id(void) 3224bf6fa2c8SYonghong Song * Return 3225bf6fa2c8SYonghong Song * A 64-bit integer containing the current cgroup id based 3226bf6fa2c8SYonghong Song * on the cgroup within which the current task is running. 3227cd339431SRoman Gushchin * 322862369db2SQuentin Monnet * void *bpf_get_local_storage(void *map, u64 flags) 3229cd339431SRoman Gushchin * Description 3230cd339431SRoman Gushchin * Get the pointer to the local storage area. 3231cd339431SRoman Gushchin * The type and the size of the local storage is defined 3232cd339431SRoman Gushchin * by the *map* argument. 3233cd339431SRoman Gushchin * The *flags* meaning is specific for each map type, 3234cd339431SRoman Gushchin * and has to be 0 for cgroup local storage. 3235cd339431SRoman Gushchin * 323690b1023fSQuentin Monnet * Depending on the BPF program type, a local storage area 323790b1023fSQuentin Monnet * can be shared between multiple instances of the BPF program, 3238cd339431SRoman Gushchin * running simultaneously. 3239cd339431SRoman Gushchin * 3240cd339431SRoman Gushchin * A user should care about the synchronization by himself. 324191c960b0SBrendan Jackman * For example, by using the **BPF_ATOMIC** instructions to alter 3242cd339431SRoman Gushchin * the shared data. 3243cd339431SRoman Gushchin * Return 324490b1023fSQuentin Monnet * A pointer to the local storage area. 32452dbb9b9eSMartin KaFai Lau * 3246bdb7b79bSAndrii Nakryiko * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 32472dbb9b9eSMartin KaFai Lau * Description 324890b1023fSQuentin Monnet * Select a **SO_REUSEPORT** socket from a 3249f170acdaSKuniyuki Iwashima * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 325090b1023fSQuentin Monnet * It checks the selected socket is matching the incoming 325190b1023fSQuentin Monnet * request in the socket buffer. 32522dbb9b9eSMartin KaFai Lau * Return 32532dbb9b9eSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 32546acc9b43SJoe Stringer * 325562369db2SQuentin Monnet * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 325662369db2SQuentin Monnet * Description 325762369db2SQuentin Monnet * Return id of cgroup v2 that is ancestor of cgroup associated 325862369db2SQuentin Monnet * with the *skb* at the *ancestor_level*. The root cgroup is at 325962369db2SQuentin Monnet * *ancestor_level* zero and each step down the hierarchy 326062369db2SQuentin Monnet * increments the level. If *ancestor_level* == level of cgroup 326162369db2SQuentin Monnet * associated with *skb*, then return value will be same as that 326262369db2SQuentin Monnet * of **bpf_skb_cgroup_id**\ (). 326362369db2SQuentin Monnet * 326462369db2SQuentin Monnet * The helper is useful to implement policies based on cgroups 326562369db2SQuentin Monnet * that are upper in hierarchy than immediate cgroup associated 326662369db2SQuentin Monnet * with *skb*. 326762369db2SQuentin Monnet * 326862369db2SQuentin Monnet * The format of returned id and helper limitations are same as in 326962369db2SQuentin Monnet * **bpf_skb_cgroup_id**\ (). 327062369db2SQuentin Monnet * Return 327162369db2SQuentin Monnet * The id is returned or 0 in case the id could not be retrieved. 327262369db2SQuentin Monnet * 3273f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 32746acc9b43SJoe Stringer * Description 32756acc9b43SJoe Stringer * Look for TCP socket matching *tuple*, optionally in a child 32766acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 327790b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 32786acc9b43SJoe Stringer * 32796acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 32806acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 32816acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 32826acc9b43SJoe Stringer * 32836acc9b43SJoe Stringer * *tuple_size* must be one of: 32846acc9b43SJoe Stringer * 32856acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 32866acc9b43SJoe Stringer * Look for an IPv4 socket. 32876acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 32886acc9b43SJoe Stringer * Look for an IPv6 socket. 32896acc9b43SJoe Stringer * 3290f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 3291bfdfa517SRandy Dunlap * socket lookup table in the netns associated with the *ctx* 3292f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 3293f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 3294f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 3295f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 3296f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 3297f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 32986acc9b43SJoe Stringer * 32996acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 33006acc9b43SJoe Stringer * be left at zero. 33016acc9b43SJoe Stringer * 33026acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 33036acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 33046acc9b43SJoe Stringer * Return 33050bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 33060bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 330780867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 330880867c5eSQuentin Monnet * tuple. 33096acc9b43SJoe Stringer * 3310f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 33116acc9b43SJoe Stringer * Description 33126acc9b43SJoe Stringer * Look for UDP socket matching *tuple*, optionally in a child 33136acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 331490b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 33156acc9b43SJoe Stringer * 33166acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 33176acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 33186acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 33196acc9b43SJoe Stringer * 33206acc9b43SJoe Stringer * *tuple_size* must be one of: 33216acc9b43SJoe Stringer * 33226acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 33236acc9b43SJoe Stringer * Look for an IPv4 socket. 33246acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 33256acc9b43SJoe Stringer * Look for an IPv6 socket. 33266acc9b43SJoe Stringer * 3327f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 3328bfdfa517SRandy Dunlap * socket lookup table in the netns associated with the *ctx* 3329f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 3330f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 3331f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 3332f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 3333f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 3334f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 33356acc9b43SJoe Stringer * 33366acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 33376acc9b43SJoe Stringer * be left at zero. 33386acc9b43SJoe Stringer * 33396acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 33406acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 33416acc9b43SJoe Stringer * Return 33420bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 33430bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 334480867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 334580867c5eSQuentin Monnet * tuple. 33466acc9b43SJoe Stringer * 3347a5fa25adSMartin KaFai Lau * long bpf_sk_release(void *sock) 33486acc9b43SJoe Stringer * Description 334990b1023fSQuentin Monnet * Release the reference held by *sock*. *sock* must be a 335090b1023fSQuentin Monnet * non-**NULL** pointer that was returned from 335190b1023fSQuentin Monnet * **bpf_sk_lookup_xxx**\ (). 335290b1023fSQuentin Monnet * Return 335390b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 335490b1023fSQuentin Monnet * 3355bdb7b79bSAndrii Nakryiko * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 335662369db2SQuentin Monnet * Description 335762369db2SQuentin Monnet * Push an element *value* in *map*. *flags* is one of: 335862369db2SQuentin Monnet * 335962369db2SQuentin Monnet * **BPF_EXIST** 336062369db2SQuentin Monnet * If the queue/stack is full, the oldest element is 336162369db2SQuentin Monnet * removed to make room for this. 336262369db2SQuentin Monnet * Return 336362369db2SQuentin Monnet * 0 on success, or a negative error in case of failure. 336462369db2SQuentin Monnet * 3365bdb7b79bSAndrii Nakryiko * long bpf_map_pop_elem(struct bpf_map *map, void *value) 336690b1023fSQuentin Monnet * Description 336790b1023fSQuentin Monnet * Pop an element from *map*. 336890b1023fSQuentin Monnet * Return 336990b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 337090b1023fSQuentin Monnet * 3371bdb7b79bSAndrii Nakryiko * long bpf_map_peek_elem(struct bpf_map *map, void *value) 337290b1023fSQuentin Monnet * Description 337390b1023fSQuentin Monnet * Get an element from *map* without removing it. 33746acc9b43SJoe Stringer * Return 33756acc9b43SJoe Stringer * 0 on success, or a negative error in case of failure. 33766fff607eSJohn Fastabend * 3377bdb7b79bSAndrii Nakryiko * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 33786fff607eSJohn Fastabend * Description 337990b1023fSQuentin Monnet * For socket policies, insert *len* bytes into *msg* at offset 33806fff607eSJohn Fastabend * *start*. 33816fff607eSJohn Fastabend * 33826fff607eSJohn Fastabend * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 338390b1023fSQuentin Monnet * *msg* it may want to insert metadata or options into the *msg*. 33846fff607eSJohn Fastabend * This can later be read and used by any of the lower layer BPF 33856fff607eSJohn Fastabend * hooks. 33866fff607eSJohn Fastabend * 33876fff607eSJohn Fastabend * This helper may fail if under memory pressure (a malloc 33886fff607eSJohn Fastabend * fails) in these cases BPF programs will get an appropriate 33896fff607eSJohn Fastabend * error and BPF programs will need to handle them. 33906fff607eSJohn Fastabend * Return 33916fff607eSJohn Fastabend * 0 on success, or a negative error in case of failure. 33927246d8edSJohn Fastabend * 3393bdb7b79bSAndrii Nakryiko * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 33947246d8edSJohn Fastabend * Description 33955f0e5412SAndrii Nakryiko * Will remove *len* bytes from a *msg* starting at byte *start*. 33967246d8edSJohn Fastabend * This may result in **ENOMEM** errors under certain situations if 33977246d8edSJohn Fastabend * an allocation and copy are required due to a full ring buffer. 33987246d8edSJohn Fastabend * However, the helper will try to avoid doing the allocation 33997246d8edSJohn Fastabend * if possible. Other errors can occur if input parameters are 340090b1023fSQuentin Monnet * invalid either due to *start* byte not being valid part of *msg* 34017246d8edSJohn Fastabend * payload and/or *pop* value being to large. 34027246d8edSJohn Fastabend * Return 340390b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 340401d3240aSSean Young * 3405bdb7b79bSAndrii Nakryiko * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 340601d3240aSSean Young * Description 340701d3240aSSean Young * This helper is used in programs implementing IR decoding, to 340801d3240aSSean Young * report a successfully decoded pointer movement. 340901d3240aSSean Young * 341001d3240aSSean Young * The *ctx* should point to the lirc sample as passed into 341101d3240aSSean Young * the program. 341201d3240aSSean Young * 341301d3240aSSean Young * This helper is only available is the kernel was compiled with 341401d3240aSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 341501d3240aSSean Young * "**y**". 341601d3240aSSean Young * Return 341701d3240aSSean Young * 0 341846f8bc92SMartin KaFai Lau * 3419bdb7b79bSAndrii Nakryiko * long bpf_spin_lock(struct bpf_spin_lock *lock) 34200eb09785SQuentin Monnet * Description 34210eb09785SQuentin Monnet * Acquire a spinlock represented by the pointer *lock*, which is 34220eb09785SQuentin Monnet * stored as part of a value of a map. Taking the lock allows to 34230eb09785SQuentin Monnet * safely update the rest of the fields in that value. The 34240eb09785SQuentin Monnet * spinlock can (and must) later be released with a call to 34250eb09785SQuentin Monnet * **bpf_spin_unlock**\ (\ *lock*\ ). 34260eb09785SQuentin Monnet * 34270eb09785SQuentin Monnet * Spinlocks in BPF programs come with a number of restrictions 34280eb09785SQuentin Monnet * and constraints: 34290eb09785SQuentin Monnet * 34300eb09785SQuentin Monnet * * **bpf_spin_lock** objects are only allowed inside maps of 34310eb09785SQuentin Monnet * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 34320eb09785SQuentin Monnet * list could be extended in the future). 34330eb09785SQuentin Monnet * * BTF description of the map is mandatory. 34340eb09785SQuentin Monnet * * The BPF program can take ONE lock at a time, since taking two 34350eb09785SQuentin Monnet * or more could cause dead locks. 34360eb09785SQuentin Monnet * * Only one **struct bpf_spin_lock** is allowed per map element. 34370eb09785SQuentin Monnet * * When the lock is taken, calls (either BPF to BPF or helpers) 34380eb09785SQuentin Monnet * are not allowed. 34390eb09785SQuentin Monnet * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 34400eb09785SQuentin Monnet * allowed inside a spinlock-ed region. 34410eb09785SQuentin Monnet * * The BPF program MUST call **bpf_spin_unlock**\ () to release 34420eb09785SQuentin Monnet * the lock, on all execution paths, before it returns. 34430eb09785SQuentin Monnet * * The BPF program can access **struct bpf_spin_lock** only via 34440eb09785SQuentin Monnet * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 34450eb09785SQuentin Monnet * helpers. Loading or storing data into the **struct 34460eb09785SQuentin Monnet * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 34470eb09785SQuentin Monnet * * To use the **bpf_spin_lock**\ () helper, the BTF description 34480eb09785SQuentin Monnet * of the map value must be a struct and have **struct 34490eb09785SQuentin Monnet * bpf_spin_lock** *anyname*\ **;** field at the top level. 34500eb09785SQuentin Monnet * Nested lock inside another struct is not allowed. 34510eb09785SQuentin Monnet * * The **struct bpf_spin_lock** *lock* field in a map value must 34520eb09785SQuentin Monnet * be aligned on a multiple of 4 bytes in that value. 34530eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 34540eb09785SQuentin Monnet * the **bpf_spin_lock** field to user space. 34550eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 34560eb09785SQuentin Monnet * a BPF program, do not update the **bpf_spin_lock** field. 34570eb09785SQuentin Monnet * * **bpf_spin_lock** cannot be on the stack or inside a 34580eb09785SQuentin Monnet * networking packet (it can only be inside of a map values). 34590eb09785SQuentin Monnet * * **bpf_spin_lock** is available to root only. 34600eb09785SQuentin Monnet * * Tracing programs and socket filter programs cannot use 34610eb09785SQuentin Monnet * **bpf_spin_lock**\ () due to insufficient preemption checks 34620eb09785SQuentin Monnet * (but this may change in the future). 34630eb09785SQuentin Monnet * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 34640eb09785SQuentin Monnet * Return 34650eb09785SQuentin Monnet * 0 34660eb09785SQuentin Monnet * 3467bdb7b79bSAndrii Nakryiko * long bpf_spin_unlock(struct bpf_spin_lock *lock) 34680eb09785SQuentin Monnet * Description 34690eb09785SQuentin Monnet * Release the *lock* previously locked by a call to 34700eb09785SQuentin Monnet * **bpf_spin_lock**\ (\ *lock*\ ). 34710eb09785SQuentin Monnet * Return 34720eb09785SQuentin Monnet * 0 34730eb09785SQuentin Monnet * 347446f8bc92SMartin KaFai Lau * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 347546f8bc92SMartin KaFai Lau * Description 347646f8bc92SMartin KaFai Lau * This helper gets a **struct bpf_sock** pointer such 347762369db2SQuentin Monnet * that all the fields in this **bpf_sock** can be accessed. 347846f8bc92SMartin KaFai Lau * Return 347962369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 348046f8bc92SMartin KaFai Lau * case of failure. 3481655a51e5SMartin KaFai Lau * 3482655a51e5SMartin KaFai Lau * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3483655a51e5SMartin KaFai Lau * Description 3484655a51e5SMartin KaFai Lau * This helper gets a **struct bpf_tcp_sock** pointer from a 3485655a51e5SMartin KaFai Lau * **struct bpf_sock** pointer. 3486655a51e5SMartin KaFai Lau * Return 348762369db2SQuentin Monnet * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3488655a51e5SMartin KaFai Lau * case of failure. 3489f7c917baSbrakmo * 3490bdb7b79bSAndrii Nakryiko * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3491f7c917baSbrakmo * Description 349262369db2SQuentin Monnet * Set ECN (Explicit Congestion Notification) field of IP header 349362369db2SQuentin Monnet * to **CE** (Congestion Encountered) if current value is **ECT** 349462369db2SQuentin Monnet * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 349562369db2SQuentin Monnet * and IPv4. 3496f7c917baSbrakmo * Return 349762369db2SQuentin Monnet * 1 if the **CE** flag is set (either by the current helper call 349862369db2SQuentin Monnet * or because it was already present), 0 if it is not set. 3499dbafd7ddSMartin KaFai Lau * 3500dbafd7ddSMartin KaFai Lau * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3501dbafd7ddSMartin KaFai Lau * Description 350262369db2SQuentin Monnet * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 350362369db2SQuentin Monnet * **bpf_sk_release**\ () is unnecessary and not allowed. 3504dbafd7ddSMartin KaFai Lau * Return 350562369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 3506dbafd7ddSMartin KaFai Lau * case of failure. 3507edbf8c01SLorenz Bauer * 3508edbf8c01SLorenz Bauer * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3509edbf8c01SLorenz Bauer * Description 3510edbf8c01SLorenz Bauer * Look for TCP socket matching *tuple*, optionally in a child 3511edbf8c01SLorenz Bauer * network namespace *netns*. The return value must be checked, 3512edbf8c01SLorenz Bauer * and if non-**NULL**, released via **bpf_sk_release**\ (). 3513edbf8c01SLorenz Bauer * 351480867c5eSQuentin Monnet * This function is identical to **bpf_sk_lookup_tcp**\ (), except 351580867c5eSQuentin Monnet * that it also returns timewait or request sockets. Use 351680867c5eSQuentin Monnet * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 351780867c5eSQuentin Monnet * full structure. 3518edbf8c01SLorenz Bauer * 3519edbf8c01SLorenz Bauer * This helper is available only if the kernel was compiled with 3520edbf8c01SLorenz Bauer * **CONFIG_NET** configuration option. 3521edbf8c01SLorenz Bauer * Return 3522edbf8c01SLorenz Bauer * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3523edbf8c01SLorenz Bauer * For sockets with reuseport option, the **struct bpf_sock** 352480867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 352580867c5eSQuentin Monnet * tuple. 352639904084SLorenz Bauer * 3527c0df236eSMartin KaFai Lau * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 352839904084SLorenz Bauer * Description 352980867c5eSQuentin Monnet * Check whether *iph* and *th* contain a valid SYN cookie ACK for 353080867c5eSQuentin Monnet * the listening socket in *sk*. 353139904084SLorenz Bauer * 353280867c5eSQuentin Monnet * *iph* points to the start of the IPv4 or IPv6 header, while 353380867c5eSQuentin Monnet * *iph_len* contains **sizeof**\ (**struct iphdr**) or 353480867c5eSQuentin Monnet * **sizeof**\ (**struct ip6hdr**). 353539904084SLorenz Bauer * 353680867c5eSQuentin Monnet * *th* points to the start of the TCP header, while *th_len* 353780867c5eSQuentin Monnet * contains **sizeof**\ (**struct tcphdr**). 353839904084SLorenz Bauer * Return 353980867c5eSQuentin Monnet * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 354080867c5eSQuentin Monnet * error otherwise. 3541808649fbSAndrey Ignatov * 3542bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3543808649fbSAndrey Ignatov * Description 3544808649fbSAndrey Ignatov * Get name of sysctl in /proc/sys/ and copy it into provided by 3545808649fbSAndrey Ignatov * program buffer *buf* of size *buf_len*. 3546808649fbSAndrey Ignatov * 3547808649fbSAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 3548808649fbSAndrey Ignatov * 3549808649fbSAndrey Ignatov * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3550808649fbSAndrey Ignatov * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3551808649fbSAndrey Ignatov * only (e.g. "tcp_mem"). 3552808649fbSAndrey Ignatov * Return 3553808649fbSAndrey Ignatov * Number of character copied (not including the trailing NUL). 3554808649fbSAndrey Ignatov * 3555808649fbSAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3556808649fbSAndrey Ignatov * truncated name in this case). 35571d11b301SAndrey Ignatov * 3558bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 35591d11b301SAndrey Ignatov * Description 35601d11b301SAndrey Ignatov * Get current value of sysctl as it is presented in /proc/sys 35611d11b301SAndrey Ignatov * (incl. newline, etc), and copy it as a string into provided 35621d11b301SAndrey Ignatov * by program buffer *buf* of size *buf_len*. 35631d11b301SAndrey Ignatov * 35641d11b301SAndrey Ignatov * The whole value is copied, no matter what file position user 35651d11b301SAndrey Ignatov * space issued e.g. sys_read at. 35661d11b301SAndrey Ignatov * 35671d11b301SAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 35681d11b301SAndrey Ignatov * Return 35691d11b301SAndrey Ignatov * Number of character copied (not including the trailing NUL). 35701d11b301SAndrey Ignatov * 35711d11b301SAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 35721d11b301SAndrey Ignatov * truncated name in this case). 35731d11b301SAndrey Ignatov * 35741d11b301SAndrey Ignatov * **-EINVAL** if current value was unavailable, e.g. because 35751d11b301SAndrey Ignatov * sysctl is uninitialized and read returns -EIO for it. 35764e63acdfSAndrey Ignatov * 3577bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 35784e63acdfSAndrey Ignatov * Description 35794e63acdfSAndrey Ignatov * Get new value being written by user space to sysctl (before 35804e63acdfSAndrey Ignatov * the actual write happens) and copy it as a string into 35814e63acdfSAndrey Ignatov * provided by program buffer *buf* of size *buf_len*. 35824e63acdfSAndrey Ignatov * 35834e63acdfSAndrey Ignatov * User space may write new value at file position > 0. 35844e63acdfSAndrey Ignatov * 35854e63acdfSAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 35864e63acdfSAndrey Ignatov * Return 35874e63acdfSAndrey Ignatov * Number of character copied (not including the trailing NUL). 35884e63acdfSAndrey Ignatov * 35894e63acdfSAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 35904e63acdfSAndrey Ignatov * truncated name in this case). 35914e63acdfSAndrey Ignatov * 35924e63acdfSAndrey Ignatov * **-EINVAL** if sysctl is being read. 35934e63acdfSAndrey Ignatov * 3594bdb7b79bSAndrii Nakryiko * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 35954e63acdfSAndrey Ignatov * Description 35964e63acdfSAndrey Ignatov * Override new value being written by user space to sysctl with 35974e63acdfSAndrey Ignatov * value provided by program in buffer *buf* of size *buf_len*. 35984e63acdfSAndrey Ignatov * 35994e63acdfSAndrey Ignatov * *buf* should contain a string in same form as provided by user 36004e63acdfSAndrey Ignatov * space on sysctl write. 36014e63acdfSAndrey Ignatov * 36024e63acdfSAndrey Ignatov * User space may write new value at file position > 0. To override 36034e63acdfSAndrey Ignatov * the whole sysctl value file position should be set to zero. 36044e63acdfSAndrey Ignatov * Return 36054e63acdfSAndrey Ignatov * 0 on success. 36064e63acdfSAndrey Ignatov * 36074e63acdfSAndrey Ignatov * **-E2BIG** if the *buf_len* is too big. 36084e63acdfSAndrey Ignatov * 36094e63acdfSAndrey Ignatov * **-EINVAL** if sysctl is being read. 3610d7a4cb9bSAndrey Ignatov * 3611bdb7b79bSAndrii Nakryiko * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3612d7a4cb9bSAndrey Ignatov * Description 3613d7a4cb9bSAndrey Ignatov * Convert the initial part of the string from buffer *buf* of 3614d7a4cb9bSAndrey Ignatov * size *buf_len* to a long integer according to the given base 3615d7a4cb9bSAndrey Ignatov * and save the result in *res*. 3616d7a4cb9bSAndrey Ignatov * 3617d7a4cb9bSAndrey Ignatov * The string may begin with an arbitrary amount of white space 361880867c5eSQuentin Monnet * (as determined by **isspace**\ (3)) followed by a single 361980867c5eSQuentin Monnet * optional '**-**' sign. 3620d7a4cb9bSAndrey Ignatov * 3621d7a4cb9bSAndrey Ignatov * Five least significant bits of *flags* encode base, other bits 3622d7a4cb9bSAndrey Ignatov * are currently unused. 3623d7a4cb9bSAndrey Ignatov * 3624d7a4cb9bSAndrey Ignatov * Base must be either 8, 10, 16 or 0 to detect it automatically 362580867c5eSQuentin Monnet * similar to user space **strtol**\ (3). 3626d7a4cb9bSAndrey Ignatov * Return 3627d7a4cb9bSAndrey Ignatov * Number of characters consumed on success. Must be positive but 362880867c5eSQuentin Monnet * no more than *buf_len*. 3629d7a4cb9bSAndrey Ignatov * 3630d7a4cb9bSAndrey Ignatov * **-EINVAL** if no valid digits were found or unsupported base 3631d7a4cb9bSAndrey Ignatov * was provided. 3632d7a4cb9bSAndrey Ignatov * 3633d7a4cb9bSAndrey Ignatov * **-ERANGE** if resulting value was out of range. 3634d7a4cb9bSAndrey Ignatov * 3635bdb7b79bSAndrii Nakryiko * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3636d7a4cb9bSAndrey Ignatov * Description 3637d7a4cb9bSAndrey Ignatov * Convert the initial part of the string from buffer *buf* of 3638d7a4cb9bSAndrey Ignatov * size *buf_len* to an unsigned long integer according to the 3639d7a4cb9bSAndrey Ignatov * given base and save the result in *res*. 3640d7a4cb9bSAndrey Ignatov * 3641d7a4cb9bSAndrey Ignatov * The string may begin with an arbitrary amount of white space 364280867c5eSQuentin Monnet * (as determined by **isspace**\ (3)). 3643d7a4cb9bSAndrey Ignatov * 3644d7a4cb9bSAndrey Ignatov * Five least significant bits of *flags* encode base, other bits 3645d7a4cb9bSAndrey Ignatov * are currently unused. 3646d7a4cb9bSAndrey Ignatov * 3647d7a4cb9bSAndrey Ignatov * Base must be either 8, 10, 16 or 0 to detect it automatically 364880867c5eSQuentin Monnet * similar to user space **strtoul**\ (3). 3649d7a4cb9bSAndrey Ignatov * Return 3650d7a4cb9bSAndrey Ignatov * Number of characters consumed on success. Must be positive but 365180867c5eSQuentin Monnet * no more than *buf_len*. 3652d7a4cb9bSAndrey Ignatov * 3653d7a4cb9bSAndrey Ignatov * **-EINVAL** if no valid digits were found or unsupported base 3654d7a4cb9bSAndrey Ignatov * was provided. 3655d7a4cb9bSAndrey Ignatov * 3656d7a4cb9bSAndrey Ignatov * **-ERANGE** if resulting value was out of range. 36576ac99e8fSMartin KaFai Lau * 365830897832SKP Singh * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 36596ac99e8fSMartin KaFai Lau * Description 366080867c5eSQuentin Monnet * Get a bpf-local-storage from a *sk*. 36616ac99e8fSMartin KaFai Lau * 36626ac99e8fSMartin KaFai Lau * Logically, it could be thought of getting the value from 36636ac99e8fSMartin KaFai Lau * a *map* with *sk* as the **key**. From this 36646ac99e8fSMartin KaFai Lau * perspective, the usage is not much different from 366580867c5eSQuentin Monnet * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 366680867c5eSQuentin Monnet * helper enforces the key must be a full socket and the map must 366780867c5eSQuentin Monnet * be a **BPF_MAP_TYPE_SK_STORAGE** also. 36686ac99e8fSMartin KaFai Lau * 36696ac99e8fSMartin KaFai Lau * Underneath, the value is stored locally at *sk* instead of 367080867c5eSQuentin Monnet * the *map*. The *map* is used as the bpf-local-storage 367180867c5eSQuentin Monnet * "type". The bpf-local-storage "type" (i.e. the *map*) is 367280867c5eSQuentin Monnet * searched against all bpf-local-storages residing at *sk*. 36736ac99e8fSMartin KaFai Lau * 367430897832SKP Singh * *sk* is a kernel **struct sock** pointer for LSM program. 367530897832SKP Singh * *sk* is a **struct bpf_sock** pointer for other program types. 367630897832SKP Singh * 367780867c5eSQuentin Monnet * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 36786ac99e8fSMartin KaFai Lau * used such that a new bpf-local-storage will be 36796ac99e8fSMartin KaFai Lau * created if one does not exist. *value* can be used 368080867c5eSQuentin Monnet * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 36816ac99e8fSMartin KaFai Lau * the initial value of a bpf-local-storage. If *value* is 368280867c5eSQuentin Monnet * **NULL**, the new bpf-local-storage will be zero initialized. 36836ac99e8fSMartin KaFai Lau * Return 36846ac99e8fSMartin KaFai Lau * A bpf-local-storage pointer is returned on success. 36856ac99e8fSMartin KaFai Lau * 36866ac99e8fSMartin KaFai Lau * **NULL** if not found or there was an error in adding 36876ac99e8fSMartin KaFai Lau * a new bpf-local-storage. 36886ac99e8fSMartin KaFai Lau * 368930897832SKP Singh * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 36906ac99e8fSMartin KaFai Lau * Description 369180867c5eSQuentin Monnet * Delete a bpf-local-storage from a *sk*. 36926ac99e8fSMartin KaFai Lau * Return 36936ac99e8fSMartin KaFai Lau * 0 on success. 36946ac99e8fSMartin KaFai Lau * 36956ac99e8fSMartin KaFai Lau * **-ENOENT** if the bpf-local-storage cannot be found. 3696592a3498SMartin KaFai Lau * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 36978b401f9eSYonghong Song * 3698bdb7b79bSAndrii Nakryiko * long bpf_send_signal(u32 sig) 36998b401f9eSYonghong Song * Description 37008482941fSYonghong Song * Send signal *sig* to the process of the current task. 37018482941fSYonghong Song * The signal may be delivered to any of this process's threads. 37028b401f9eSYonghong Song * Return 37038b401f9eSYonghong Song * 0 on success or successfully queued. 37048b401f9eSYonghong Song * 37058b401f9eSYonghong Song * **-EBUSY** if work queue under nmi is full. 37068b401f9eSYonghong Song * 37078b401f9eSYonghong Song * **-EINVAL** if *sig* is invalid. 37088b401f9eSYonghong Song * 37098b401f9eSYonghong Song * **-EPERM** if no permission to send the *sig*. 37108b401f9eSYonghong Song * 37118b401f9eSYonghong Song * **-EAGAIN** if bpf program can try again. 371270d66244SPetar Penkov * 3713c0df236eSMartin KaFai Lau * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 371470d66244SPetar Penkov * Description 371570d66244SPetar Penkov * Try to issue a SYN cookie for the packet with corresponding 371670d66244SPetar Penkov * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 371770d66244SPetar Penkov * 371870d66244SPetar Penkov * *iph* points to the start of the IPv4 or IPv6 header, while 371970d66244SPetar Penkov * *iph_len* contains **sizeof**\ (**struct iphdr**) or 372070d66244SPetar Penkov * **sizeof**\ (**struct ip6hdr**). 372170d66244SPetar Penkov * 372270d66244SPetar Penkov * *th* points to the start of the TCP header, while *th_len* 372370d66244SPetar Penkov * contains the length of the TCP header. 372470d66244SPetar Penkov * Return 372570d66244SPetar Penkov * On success, lower 32 bits hold the generated SYN cookie in 372670d66244SPetar Penkov * followed by 16 bits which hold the MSS value for that cookie, 372770d66244SPetar Penkov * and the top 16 bits are unused. 372870d66244SPetar Penkov * 372970d66244SPetar Penkov * On failure, the returned value is one of the following: 373070d66244SPetar Penkov * 373170d66244SPetar Penkov * **-EINVAL** SYN cookie cannot be issued due to error 373270d66244SPetar Penkov * 373370d66244SPetar Penkov * **-ENOENT** SYN cookie should not be issued (no SYN flood) 373470d66244SPetar Penkov * 373570d66244SPetar Penkov * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 373670d66244SPetar Penkov * 373770d66244SPetar Penkov * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 3738a7658e1aSAlexei Starovoitov * 3739bdb7b79bSAndrii Nakryiko * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3740a7658e1aSAlexei Starovoitov * Description 3741a7658e1aSAlexei Starovoitov * Write raw *data* blob into a special BPF perf event held by 3742a7658e1aSAlexei Starovoitov * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3743a7658e1aSAlexei Starovoitov * event must have the following attributes: **PERF_SAMPLE_RAW** 3744a7658e1aSAlexei Starovoitov * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3745a7658e1aSAlexei Starovoitov * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3746a7658e1aSAlexei Starovoitov * 3747a7658e1aSAlexei Starovoitov * The *flags* are used to indicate the index in *map* for which 3748a7658e1aSAlexei Starovoitov * the value must be put, masked with **BPF_F_INDEX_MASK**. 3749a7658e1aSAlexei Starovoitov * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3750a7658e1aSAlexei Starovoitov * to indicate that the index of the current CPU core should be 3751a7658e1aSAlexei Starovoitov * used. 3752a7658e1aSAlexei Starovoitov * 3753a7658e1aSAlexei Starovoitov * The value to write, of *size*, is passed through eBPF stack and 3754a7658e1aSAlexei Starovoitov * pointed by *data*. 3755a7658e1aSAlexei Starovoitov * 3756a7658e1aSAlexei Starovoitov * *ctx* is a pointer to in-kernel struct sk_buff. 3757a7658e1aSAlexei Starovoitov * 3758a7658e1aSAlexei Starovoitov * This helper is similar to **bpf_perf_event_output**\ () but 3759a7658e1aSAlexei Starovoitov * restricted to raw_tracepoint bpf programs. 3760a7658e1aSAlexei Starovoitov * Return 3761a7658e1aSAlexei Starovoitov * 0 on success, or a negative error in case of failure. 37626ae08ae3SDaniel Borkmann * 3763bdb7b79bSAndrii Nakryiko * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 37646ae08ae3SDaniel Borkmann * Description 37656ae08ae3SDaniel Borkmann * Safely attempt to read *size* bytes from user space address 37666ae08ae3SDaniel Borkmann * *unsafe_ptr* and store the data in *dst*. 37676ae08ae3SDaniel Borkmann * Return 37686ae08ae3SDaniel Borkmann * 0 on success, or a negative error in case of failure. 37696ae08ae3SDaniel Borkmann * 3770bdb7b79bSAndrii Nakryiko * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 37716ae08ae3SDaniel Borkmann * Description 37726ae08ae3SDaniel Borkmann * Safely attempt to read *size* bytes from kernel space address 37736ae08ae3SDaniel Borkmann * *unsafe_ptr* and store the data in *dst*. 37746ae08ae3SDaniel Borkmann * Return 37756ae08ae3SDaniel Borkmann * 0 on success, or a negative error in case of failure. 37766ae08ae3SDaniel Borkmann * 3777bdb7b79bSAndrii Nakryiko * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 37786ae08ae3SDaniel Borkmann * Description 37796ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe user address 37806ae08ae3SDaniel Borkmann * *unsafe_ptr* to *dst*. The *size* should include the 37816ae08ae3SDaniel Borkmann * terminating NUL byte. In case the string length is smaller than 37826ae08ae3SDaniel Borkmann * *size*, the target is not padded with further NUL bytes. If the 37836ae08ae3SDaniel Borkmann * string length is larger than *size*, just *size*-1 bytes are 37846ae08ae3SDaniel Borkmann * copied and the last byte is set to NUL. 37856ae08ae3SDaniel Borkmann * 3786c6458e72SBrendan Jackman * On success, returns the number of bytes that were written, 3787c6458e72SBrendan Jackman * including the terminal NUL. This makes this helper useful in 3788c6458e72SBrendan Jackman * tracing programs for reading strings, and more importantly to 3789c6458e72SBrendan Jackman * get its length at runtime. See the following snippet: 37906ae08ae3SDaniel Borkmann * 37916ae08ae3SDaniel Borkmann * :: 37926ae08ae3SDaniel Borkmann * 37936ae08ae3SDaniel Borkmann * SEC("kprobe/sys_open") 37946ae08ae3SDaniel Borkmann * void bpf_sys_open(struct pt_regs *ctx) 37956ae08ae3SDaniel Borkmann * { 37966ae08ae3SDaniel Borkmann * char buf[PATHLEN]; // PATHLEN is defined to 256 37976ae08ae3SDaniel Borkmann * int res = bpf_probe_read_user_str(buf, sizeof(buf), 37986ae08ae3SDaniel Borkmann * ctx->di); 37996ae08ae3SDaniel Borkmann * 38006ae08ae3SDaniel Borkmann * // Consume buf, for example push it to 38016ae08ae3SDaniel Borkmann * // userspace via bpf_perf_event_output(); we 38026ae08ae3SDaniel Borkmann * // can use res (the string length) as event 38036ae08ae3SDaniel Borkmann * // size, after checking its boundaries. 38046ae08ae3SDaniel Borkmann * } 38056ae08ae3SDaniel Borkmann * 3806ab8d7809SQuentin Monnet * In comparison, using **bpf_probe_read_user**\ () helper here 38076ae08ae3SDaniel Borkmann * instead to read the string would require to estimate the length 38086ae08ae3SDaniel Borkmann * at compile time, and would often result in copying more memory 38096ae08ae3SDaniel Borkmann * than necessary. 38106ae08ae3SDaniel Borkmann * 38116ae08ae3SDaniel Borkmann * Another useful use case is when parsing individual process 38126ae08ae3SDaniel Borkmann * arguments or individual environment variables navigating 38136ae08ae3SDaniel Borkmann * *current*\ **->mm->arg_start** and *current*\ 38146ae08ae3SDaniel Borkmann * **->mm->env_start**: using this helper and the return value, 38156ae08ae3SDaniel Borkmann * one can quickly iterate at the right offset of the memory area. 38166ae08ae3SDaniel Borkmann * Return 3817c6458e72SBrendan Jackman * On success, the strictly positive length of the output string, 38186ae08ae3SDaniel Borkmann * including the trailing NUL character. On error, a negative 38196ae08ae3SDaniel Borkmann * value. 38206ae08ae3SDaniel Borkmann * 3821bdb7b79bSAndrii Nakryiko * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 38226ae08ae3SDaniel Borkmann * Description 38236ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 3824ab8d7809SQuentin Monnet * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 38256ae08ae3SDaniel Borkmann * Return 38266ae08ae3SDaniel Borkmann * On success, the strictly positive length of the string, including 38276ae08ae3SDaniel Borkmann * the trailing NUL character. On error, a negative value. 3828206057feSMartin KaFai Lau * 3829bdb7b79bSAndrii Nakryiko * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 3830206057feSMartin KaFai Lau * Description 3831ab8d7809SQuentin Monnet * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 3832206057feSMartin KaFai Lau * *rcv_nxt* is the ack_seq to be sent out. 3833206057feSMartin KaFai Lau * Return 3834206057feSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 3835206057feSMartin KaFai Lau * 3836bdb7b79bSAndrii Nakryiko * long bpf_send_signal_thread(u32 sig) 38378482941fSYonghong Song * Description 38388482941fSYonghong Song * Send signal *sig* to the thread corresponding to the current task. 38398482941fSYonghong Song * Return 38408482941fSYonghong Song * 0 on success or successfully queued. 38418482941fSYonghong Song * 38428482941fSYonghong Song * **-EBUSY** if work queue under nmi is full. 38438482941fSYonghong Song * 38448482941fSYonghong Song * **-EINVAL** if *sig* is invalid. 38458482941fSYonghong Song * 38468482941fSYonghong Song * **-EPERM** if no permission to send the *sig*. 38478482941fSYonghong Song * 38488482941fSYonghong Song * **-EAGAIN** if bpf program can try again. 38495576b991SMartin KaFai Lau * 38505576b991SMartin KaFai Lau * u64 bpf_jiffies64(void) 38515576b991SMartin KaFai Lau * Description 38525576b991SMartin KaFai Lau * Obtain the 64bit jiffies 38535576b991SMartin KaFai Lau * Return 38545576b991SMartin KaFai Lau * The 64 bit jiffies 3855fff7b643SDaniel Xu * 3856bdb7b79bSAndrii Nakryiko * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3857fff7b643SDaniel Xu * Description 3858fff7b643SDaniel Xu * For an eBPF program attached to a perf event, retrieve the 3859ab8d7809SQuentin Monnet * branch records (**struct perf_branch_entry**) associated to *ctx* 3860fff7b643SDaniel Xu * and store it in the buffer pointed by *buf* up to size 3861fff7b643SDaniel Xu * *size* bytes. 3862fff7b643SDaniel Xu * Return 3863fff7b643SDaniel Xu * On success, number of bytes written to *buf*. On error, a 3864fff7b643SDaniel Xu * negative value. 3865fff7b643SDaniel Xu * 3866fff7b643SDaniel Xu * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3867fff7b643SDaniel Xu * instead return the number of bytes required to store all the 3868fff7b643SDaniel Xu * branch entries. If this flag is set, *buf* may be NULL. 3869fff7b643SDaniel Xu * 3870fff7b643SDaniel Xu * **-EINVAL** if arguments invalid or **size** not a multiple 3871ab8d7809SQuentin Monnet * of **sizeof**\ (**struct perf_branch_entry**\ ). 3872fff7b643SDaniel Xu * 3873fff7b643SDaniel Xu * **-ENOENT** if architecture does not support branch records. 3874b4490c5cSCarlos Neira * 3875bdb7b79bSAndrii Nakryiko * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3876b4490c5cSCarlos Neira * Description 3877b4490c5cSCarlos Neira * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3878b4490c5cSCarlos Neira * *namespace* will be returned in *nsdata*. 3879ab8d7809SQuentin Monnet * Return 3880ab8d7809SQuentin Monnet * 0 on success, or one of the following in case of failure: 3881b4490c5cSCarlos Neira * 3882b4490c5cSCarlos Neira * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3883b4490c5cSCarlos Neira * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3884b4490c5cSCarlos Neira * 3885b4490c5cSCarlos Neira * **-ENOENT** if pidns does not exists for the current task. 3886b4490c5cSCarlos Neira * 3887bdb7b79bSAndrii Nakryiko * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3888d831ee84SEelco Chaudron * Description 3889d831ee84SEelco Chaudron * Write raw *data* blob into a special BPF perf event held by 3890d831ee84SEelco Chaudron * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3891d831ee84SEelco Chaudron * event must have the following attributes: **PERF_SAMPLE_RAW** 3892d831ee84SEelco Chaudron * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3893d831ee84SEelco Chaudron * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3894d831ee84SEelco Chaudron * 3895d831ee84SEelco Chaudron * The *flags* are used to indicate the index in *map* for which 3896d831ee84SEelco Chaudron * the value must be put, masked with **BPF_F_INDEX_MASK**. 3897d831ee84SEelco Chaudron * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3898d831ee84SEelco Chaudron * to indicate that the index of the current CPU core should be 3899d831ee84SEelco Chaudron * used. 3900d831ee84SEelco Chaudron * 3901d831ee84SEelco Chaudron * The value to write, of *size*, is passed through eBPF stack and 3902d831ee84SEelco Chaudron * pointed by *data*. 3903d831ee84SEelco Chaudron * 3904d831ee84SEelco Chaudron * *ctx* is a pointer to in-kernel struct xdp_buff. 3905d831ee84SEelco Chaudron * 3906d831ee84SEelco Chaudron * This helper is similar to **bpf_perf_eventoutput**\ () but 3907d831ee84SEelco Chaudron * restricted to raw_tracepoint bpf programs. 3908d831ee84SEelco Chaudron * Return 3909d831ee84SEelco Chaudron * 0 on success, or a negative error in case of failure. 3910f318903cSDaniel Borkmann * 3911f318903cSDaniel Borkmann * u64 bpf_get_netns_cookie(void *ctx) 3912f318903cSDaniel Borkmann * Description 3913f318903cSDaniel Borkmann * Retrieve the cookie (generated by the kernel) of the network 3914f318903cSDaniel Borkmann * namespace the input *ctx* is associated with. The network 3915f318903cSDaniel Borkmann * namespace cookie remains stable for its lifetime and provides 3916f318903cSDaniel Borkmann * a global identifier that can be assumed unique. If *ctx* is 3917f318903cSDaniel Borkmann * NULL, then the helper returns the cookie for the initial 3918f318903cSDaniel Borkmann * network namespace. The cookie itself is very similar to that 3919ab8d7809SQuentin Monnet * of **bpf_get_socket_cookie**\ () helper, but for network 3920ab8d7809SQuentin Monnet * namespaces instead of sockets. 3921f318903cSDaniel Borkmann * Return 3922f318903cSDaniel Borkmann * A 8-byte long opaque number. 39230f09abd1SDaniel Borkmann * 39240f09abd1SDaniel Borkmann * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 39250f09abd1SDaniel Borkmann * Description 39260f09abd1SDaniel Borkmann * Return id of cgroup v2 that is ancestor of the cgroup associated 39270f09abd1SDaniel Borkmann * with the current task at the *ancestor_level*. The root cgroup 39280f09abd1SDaniel Borkmann * is at *ancestor_level* zero and each step down the hierarchy 39290f09abd1SDaniel Borkmann * increments the level. If *ancestor_level* == level of cgroup 39300f09abd1SDaniel Borkmann * associated with the current task, then return value will be the 39310f09abd1SDaniel Borkmann * same as that of **bpf_get_current_cgroup_id**\ (). 39320f09abd1SDaniel Borkmann * 39330f09abd1SDaniel Borkmann * The helper is useful to implement policies based on cgroups 39340f09abd1SDaniel Borkmann * that are upper in hierarchy than immediate cgroup associated 39350f09abd1SDaniel Borkmann * with the current task. 39360f09abd1SDaniel Borkmann * 39370f09abd1SDaniel Borkmann * The format of returned id and helper limitations are same as in 39380f09abd1SDaniel Borkmann * **bpf_get_current_cgroup_id**\ (). 39390f09abd1SDaniel Borkmann * Return 39400f09abd1SDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 3941cf7fbe66SJoe Stringer * 394227e5203bSMartin KaFai Lau * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 3943cf7fbe66SJoe Stringer * Description 3944e9ddbb77SJakub Sitnicki * Helper is overloaded depending on BPF program type. This 3945e9ddbb77SJakub Sitnicki * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 3946e9ddbb77SJakub Sitnicki * **BPF_PROG_TYPE_SCHED_ACT** programs. 3947e9ddbb77SJakub Sitnicki * 3948cf7fbe66SJoe Stringer * Assign the *sk* to the *skb*. When combined with appropriate 3949cf7fbe66SJoe Stringer * routing configuration to receive the packet towards the socket, 3950cf7fbe66SJoe Stringer * will cause *skb* to be delivered to the specified socket. 3951cf7fbe66SJoe Stringer * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3952cf7fbe66SJoe Stringer * **bpf_clone_redirect**\ () or other methods outside of BPF may 3953cf7fbe66SJoe Stringer * interfere with successful delivery to the socket. 3954cf7fbe66SJoe Stringer * 3955cf7fbe66SJoe Stringer * This operation is only valid from TC ingress path. 3956cf7fbe66SJoe Stringer * 3957cf7fbe66SJoe Stringer * The *flags* argument must be zero. 3958cf7fbe66SJoe Stringer * Return 3959ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 3960cf7fbe66SJoe Stringer * 3961ab8d7809SQuentin Monnet * **-EINVAL** if specified *flags* are not supported. 3962ab8d7809SQuentin Monnet * 3963ab8d7809SQuentin Monnet * **-ENOENT** if the socket is unavailable for assignment. 3964ab8d7809SQuentin Monnet * 3965ab8d7809SQuentin Monnet * **-ENETUNREACH** if the socket is unreachable (wrong netns). 3966ab8d7809SQuentin Monnet * 3967ab8d7809SQuentin Monnet * **-EOPNOTSUPP** if the operation is not supported, for example 3968ab8d7809SQuentin Monnet * a call from outside of TC ingress. 3969ab8d7809SQuentin Monnet * 3970ab8d7809SQuentin Monnet * **-ESOCKTNOSUPPORT** if the socket type is not supported 3971ab8d7809SQuentin Monnet * (reuseport). 397271d19214SMaciej Żenczykowski * 3973e9ddbb77SJakub Sitnicki * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 3974e9ddbb77SJakub Sitnicki * Description 3975e9ddbb77SJakub Sitnicki * Helper is overloaded depending on BPF program type. This 3976e9ddbb77SJakub Sitnicki * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 3977e9ddbb77SJakub Sitnicki * 3978e9ddbb77SJakub Sitnicki * Select the *sk* as a result of a socket lookup. 3979e9ddbb77SJakub Sitnicki * 3980e9ddbb77SJakub Sitnicki * For the operation to succeed passed socket must be compatible 3981e9ddbb77SJakub Sitnicki * with the packet description provided by the *ctx* object. 3982e9ddbb77SJakub Sitnicki * 3983e9ddbb77SJakub Sitnicki * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 3984e9ddbb77SJakub Sitnicki * be an exact match. While IP family (**AF_INET** or 3985e9ddbb77SJakub Sitnicki * **AF_INET6**) must be compatible, that is IPv6 sockets 3986e9ddbb77SJakub Sitnicki * that are not v6-only can be selected for IPv4 packets. 3987e9ddbb77SJakub Sitnicki * 3988e9ddbb77SJakub Sitnicki * Only TCP listeners and UDP unconnected sockets can be 3989e9ddbb77SJakub Sitnicki * selected. *sk* can also be NULL to reset any previous 3990e9ddbb77SJakub Sitnicki * selection. 3991e9ddbb77SJakub Sitnicki * 3992e9ddbb77SJakub Sitnicki * *flags* argument can combination of following values: 3993e9ddbb77SJakub Sitnicki * 3994e9ddbb77SJakub Sitnicki * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 3995e9ddbb77SJakub Sitnicki * socket selection, potentially done by a BPF program 3996e9ddbb77SJakub Sitnicki * that ran before us. 3997e9ddbb77SJakub Sitnicki * 3998e9ddbb77SJakub Sitnicki * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 3999e9ddbb77SJakub Sitnicki * load-balancing within reuseport group for the socket 4000e9ddbb77SJakub Sitnicki * being selected. 4001e9ddbb77SJakub Sitnicki * 4002e9ddbb77SJakub Sitnicki * On success *ctx->sk* will point to the selected socket. 4003e9ddbb77SJakub Sitnicki * 4004e9ddbb77SJakub Sitnicki * Return 4005e9ddbb77SJakub Sitnicki * 0 on success, or a negative errno in case of failure. 4006e9ddbb77SJakub Sitnicki * 4007e9ddbb77SJakub Sitnicki * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4008e9ddbb77SJakub Sitnicki * not compatible with packet family (*ctx->family*). 4009e9ddbb77SJakub Sitnicki * 4010e9ddbb77SJakub Sitnicki * * **-EEXIST** if socket has been already selected, 4011e9ddbb77SJakub Sitnicki * potentially by another program, and 4012e9ddbb77SJakub Sitnicki * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4013e9ddbb77SJakub Sitnicki * 4014e9ddbb77SJakub Sitnicki * * **-EINVAL** if unsupported flags were specified. 4015e9ddbb77SJakub Sitnicki * 4016e9ddbb77SJakub Sitnicki * * **-EPROTOTYPE** if socket L4 protocol 4017e9ddbb77SJakub Sitnicki * (*sk->protocol*) doesn't match packet protocol 4018e9ddbb77SJakub Sitnicki * (*ctx->protocol*). 4019e9ddbb77SJakub Sitnicki * 4020e9ddbb77SJakub Sitnicki * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4021e9ddbb77SJakub Sitnicki * state (TCP listening or UDP unconnected). 4022e9ddbb77SJakub Sitnicki * 402371d19214SMaciej Żenczykowski * u64 bpf_ktime_get_boot_ns(void) 402471d19214SMaciej Żenczykowski * Description 402571d19214SMaciej Żenczykowski * Return the time elapsed since system boot, in nanoseconds. 402671d19214SMaciej Żenczykowski * Does include the time the system was suspended. 4027ab8d7809SQuentin Monnet * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 402871d19214SMaciej Żenczykowski * Return 402971d19214SMaciej Żenczykowski * Current *ktime*. 4030492e639fSYonghong Song * 4031bdb7b79bSAndrii Nakryiko * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4032492e639fSYonghong Song * Description 4033ab8d7809SQuentin Monnet * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4034ab8d7809SQuentin Monnet * out the format string. 4035492e639fSYonghong Song * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4036492e639fSYonghong Song * the format string itself. The *data* and *data_len* are format string 4037ab8d7809SQuentin Monnet * arguments. The *data* are a **u64** array and corresponding format string 4038492e639fSYonghong Song * values are stored in the array. For strings and pointers where pointees 4039492e639fSYonghong Song * are accessed, only the pointer values are stored in the *data* array. 4040ab8d7809SQuentin Monnet * The *data_len* is the size of *data* in bytes. 4041492e639fSYonghong Song * 4042492e639fSYonghong Song * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4043492e639fSYonghong Song * Reading kernel memory may fail due to either invalid address or 4044492e639fSYonghong Song * valid address but requiring a major memory fault. If reading kernel memory 4045492e639fSYonghong Song * fails, the string for **%s** will be an empty string, and the ip 4046492e639fSYonghong Song * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4047ab8d7809SQuentin Monnet * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4048492e639fSYonghong Song * Return 4049ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 4050492e639fSYonghong Song * 4051ab8d7809SQuentin Monnet * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4052492e639fSYonghong Song * by returning 1 from bpf program. 4053ab8d7809SQuentin Monnet * 4054ab8d7809SQuentin Monnet * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4055ab8d7809SQuentin Monnet * 4056ab8d7809SQuentin Monnet * **-E2BIG** if *fmt* contains too many format specifiers. 4057ab8d7809SQuentin Monnet * 4058ab8d7809SQuentin Monnet * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4059492e639fSYonghong Song * 4060bdb7b79bSAndrii Nakryiko * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4061492e639fSYonghong Song * Description 4062ab8d7809SQuentin Monnet * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4063492e639fSYonghong Song * The *m* represents the seq_file. The *data* and *len* represent the 4064492e639fSYonghong Song * data to write in bytes. 4065492e639fSYonghong Song * Return 4066ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 4067492e639fSYonghong Song * 4068ab8d7809SQuentin Monnet * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4069f307fa2cSAndrey Ignatov * 4070a5fa25adSMartin KaFai Lau * u64 bpf_sk_cgroup_id(void *sk) 4071f307fa2cSAndrey Ignatov * Description 4072f307fa2cSAndrey Ignatov * Return the cgroup v2 id of the socket *sk*. 4073f307fa2cSAndrey Ignatov * 4074a5fa25adSMartin KaFai Lau * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4075f307fa2cSAndrey Ignatov * returned from **bpf_sk_lookup_xxx**\ (), 4076f307fa2cSAndrey Ignatov * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4077f307fa2cSAndrey Ignatov * same as in **bpf_skb_cgroup_id**\ (). 4078f307fa2cSAndrey Ignatov * 4079f307fa2cSAndrey Ignatov * This helper is available only if the kernel was compiled with 4080f307fa2cSAndrey Ignatov * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4081f307fa2cSAndrey Ignatov * Return 4082f307fa2cSAndrey Ignatov * The id is returned or 0 in case the id could not be retrieved. 4083f307fa2cSAndrey Ignatov * 4084a5fa25adSMartin KaFai Lau * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4085f307fa2cSAndrey Ignatov * Description 4086f307fa2cSAndrey Ignatov * Return id of cgroup v2 that is ancestor of cgroup associated 4087f307fa2cSAndrey Ignatov * with the *sk* at the *ancestor_level*. The root cgroup is at 4088f307fa2cSAndrey Ignatov * *ancestor_level* zero and each step down the hierarchy 4089f307fa2cSAndrey Ignatov * increments the level. If *ancestor_level* == level of cgroup 4090f307fa2cSAndrey Ignatov * associated with *sk*, then return value will be same as that 4091f307fa2cSAndrey Ignatov * of **bpf_sk_cgroup_id**\ (). 4092f307fa2cSAndrey Ignatov * 4093f307fa2cSAndrey Ignatov * The helper is useful to implement policies based on cgroups 4094f307fa2cSAndrey Ignatov * that are upper in hierarchy than immediate cgroup associated 4095f307fa2cSAndrey Ignatov * with *sk*. 4096f307fa2cSAndrey Ignatov * 4097f307fa2cSAndrey Ignatov * The format of returned id and helper limitations are same as in 4098f307fa2cSAndrey Ignatov * **bpf_sk_cgroup_id**\ (). 4099f307fa2cSAndrey Ignatov * Return 4100f307fa2cSAndrey Ignatov * The id is returned or 0 in case the id could not be retrieved. 4101457f4436SAndrii Nakryiko * 4102e1613b57SAndrii Nakryiko * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4103457f4436SAndrii Nakryiko * Description 4104457f4436SAndrii Nakryiko * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4105bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4106bcc7f554SQuentin Monnet * of new data availability is sent. 4107bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4108bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 41095c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 41105c507329SPedro Tammela * of new data availability is sent. 41115c507329SPedro Tammela * 41125c507329SPedro Tammela * An adaptive notification is a notification sent whenever the user-space 41135c507329SPedro Tammela * process has caught up and consumed all available payloads. In case the user-space 41145c507329SPedro Tammela * process is still processing a previous payload, then no notification is needed 41155c507329SPedro Tammela * as it will process the newly added payload automatically. 4116457f4436SAndrii Nakryiko * Return 4117bcc7f554SQuentin Monnet * 0 on success, or a negative error in case of failure. 4118457f4436SAndrii Nakryiko * 4119457f4436SAndrii Nakryiko * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4120457f4436SAndrii Nakryiko * Description 4121457f4436SAndrii Nakryiko * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 41225c507329SPedro Tammela * *flags* must be 0. 4123457f4436SAndrii Nakryiko * Return 4124457f4436SAndrii Nakryiko * Valid pointer with *size* bytes of memory available; NULL, 4125457f4436SAndrii Nakryiko * otherwise. 4126457f4436SAndrii Nakryiko * 4127457f4436SAndrii Nakryiko * void bpf_ringbuf_submit(void *data, u64 flags) 4128457f4436SAndrii Nakryiko * Description 4129457f4436SAndrii Nakryiko * Submit reserved ring buffer sample, pointed to by *data*. 4130bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4131bcc7f554SQuentin Monnet * of new data availability is sent. 4132bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4133bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 41345c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 41355c507329SPedro Tammela * of new data availability is sent. 41365c507329SPedro Tammela * 41375c507329SPedro Tammela * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4138457f4436SAndrii Nakryiko * Return 4139457f4436SAndrii Nakryiko * Nothing. Always succeeds. 4140457f4436SAndrii Nakryiko * 4141457f4436SAndrii Nakryiko * void bpf_ringbuf_discard(void *data, u64 flags) 4142457f4436SAndrii Nakryiko * Description 4143457f4436SAndrii Nakryiko * Discard reserved ring buffer sample, pointed to by *data*. 4144bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4145bcc7f554SQuentin Monnet * of new data availability is sent. 4146bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4147bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 41485c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 41495c507329SPedro Tammela * of new data availability is sent. 41505c507329SPedro Tammela * 41515c507329SPedro Tammela * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4152457f4436SAndrii Nakryiko * Return 4153457f4436SAndrii Nakryiko * Nothing. Always succeeds. 4154457f4436SAndrii Nakryiko * 4155457f4436SAndrii Nakryiko * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4156457f4436SAndrii Nakryiko * Description 4157457f4436SAndrii Nakryiko * Query various characteristics of provided ring buffer. What 4158457f4436SAndrii Nakryiko * exactly is queries is determined by *flags*: 4159bcc7f554SQuentin Monnet * 4160bcc7f554SQuentin Monnet * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4161bcc7f554SQuentin Monnet * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4162bcc7f554SQuentin Monnet * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4163bcc7f554SQuentin Monnet * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4164bcc7f554SQuentin Monnet * 4165bcc7f554SQuentin Monnet * Data returned is just a momentary snapshot of actual values 4166457f4436SAndrii Nakryiko * and could be inaccurate, so this facility should be used to 4167457f4436SAndrii Nakryiko * power heuristics and for reporting, not to make 100% correct 4168457f4436SAndrii Nakryiko * calculation. 4169457f4436SAndrii Nakryiko * Return 4170bcc7f554SQuentin Monnet * Requested value, or 0, if *flags* are not recognized. 41717cdec54fSDaniel Borkmann * 4172bdb7b79bSAndrii Nakryiko * long bpf_csum_level(struct sk_buff *skb, u64 level) 41737cdec54fSDaniel Borkmann * Description 41747cdec54fSDaniel Borkmann * Change the skbs checksum level by one layer up or down, or 41757cdec54fSDaniel Borkmann * reset it entirely to none in order to have the stack perform 41767cdec54fSDaniel Borkmann * checksum validation. The level is applicable to the following 41777cdec54fSDaniel Borkmann * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 41787cdec54fSDaniel Borkmann * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 41797cdec54fSDaniel Borkmann * through **bpf_skb_adjust_room**\ () helper with passing in 41807cdec54fSDaniel Borkmann * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 41817cdec54fSDaniel Borkmann * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 41827cdec54fSDaniel Borkmann * the UDP header is removed. Similarly, an encap of the latter 41837cdec54fSDaniel Borkmann * into the former could be accompanied by a helper call to 41847cdec54fSDaniel Borkmann * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 41857cdec54fSDaniel Borkmann * skb is still intended to be processed in higher layers of the 41867cdec54fSDaniel Borkmann * stack instead of just egressing at tc. 41877cdec54fSDaniel Borkmann * 41887cdec54fSDaniel Borkmann * There are three supported level settings at this time: 41897cdec54fSDaniel Borkmann * 41907cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 41917cdec54fSDaniel Borkmann * with CHECKSUM_UNNECESSARY. 41927cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 41937cdec54fSDaniel Borkmann * with CHECKSUM_UNNECESSARY. 41947cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 41957cdec54fSDaniel Borkmann * sets CHECKSUM_NONE to force checksum validation by the stack. 41967cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 41977cdec54fSDaniel Borkmann * skb->csum_level. 41987cdec54fSDaniel Borkmann * Return 41997cdec54fSDaniel Borkmann * 0 on success, or a negative error in case of failure. In the 42007cdec54fSDaniel Borkmann * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 42017cdec54fSDaniel Borkmann * is returned or the error code -EACCES in case the skb is not 42027cdec54fSDaniel Borkmann * subject to CHECKSUM_UNNECESSARY. 4203af7ec138SYonghong Song * 4204af7ec138SYonghong Song * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4205af7ec138SYonghong Song * Description 4206af7ec138SYonghong Song * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4207af7ec138SYonghong Song * Return 4208938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4209478cfbdfSYonghong Song * 4210478cfbdfSYonghong Song * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4211478cfbdfSYonghong Song * Description 4212478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4213478cfbdfSYonghong Song * Return 4214938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4215478cfbdfSYonghong Song * 4216478cfbdfSYonghong Song * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4217478cfbdfSYonghong Song * Description 4218478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4219478cfbdfSYonghong Song * Return 4220938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4221478cfbdfSYonghong Song * 4222478cfbdfSYonghong Song * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4223478cfbdfSYonghong Song * Description 4224478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4225478cfbdfSYonghong Song * Return 4226938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 42270d4fad3eSYonghong Song * 42280d4fad3eSYonghong Song * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 42290d4fad3eSYonghong Song * Description 42300d4fad3eSYonghong Song * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 42310d4fad3eSYonghong Song * Return 4232938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4233fa28dcb8SSong Liu * 4234fa28dcb8SSong Liu * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4235fa28dcb8SSong Liu * Description 4236fa28dcb8SSong Liu * Return a user or a kernel stack in bpf program provided buffer. 4237fa28dcb8SSong Liu * To achieve this, the helper needs *task*, which is a valid 4238938c3efdSQuentin Monnet * pointer to **struct task_struct**. To store the stacktrace, the 4239fa28dcb8SSong Liu * bpf program provides *buf* with a nonnegative *size*. 4240fa28dcb8SSong Liu * 4241fa28dcb8SSong Liu * The last argument, *flags*, holds the number of stack frames to 4242fa28dcb8SSong Liu * skip (from 0 to 255), masked with 4243fa28dcb8SSong Liu * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4244fa28dcb8SSong Liu * the following flags: 4245fa28dcb8SSong Liu * 4246fa28dcb8SSong Liu * **BPF_F_USER_STACK** 4247fa28dcb8SSong Liu * Collect a user space stack instead of a kernel stack. 4248fa28dcb8SSong Liu * **BPF_F_USER_BUILD_ID** 4249fa28dcb8SSong Liu * Collect buildid+offset instead of ips for user stack, 4250fa28dcb8SSong Liu * only valid if **BPF_F_USER_STACK** is also specified. 4251fa28dcb8SSong Liu * 4252fa28dcb8SSong Liu * **bpf_get_task_stack**\ () can collect up to 4253fa28dcb8SSong Liu * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4254fa28dcb8SSong Liu * to sufficient large buffer size. Note that 4255fa28dcb8SSong Liu * this limit can be controlled with the **sysctl** program, and 4256fa28dcb8SSong Liu * that it should be manually increased in order to profile long 4257fa28dcb8SSong Liu * user stacks (such as stacks for Java programs). To do so, use: 4258fa28dcb8SSong Liu * 4259fa28dcb8SSong Liu * :: 4260fa28dcb8SSong Liu * 4261fa28dcb8SSong Liu * # sysctl kernel.perf_event_max_stack=<new value> 4262fa28dcb8SSong Liu * Return 4263fa28dcb8SSong Liu * A non-negative value equal to or less than *size* on success, 4264fa28dcb8SSong Liu * or a negative error in case of failure. 4265fa28dcb8SSong Liu * 42660813a841SMartin KaFai Lau * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 42670813a841SMartin KaFai Lau * Description 42680813a841SMartin KaFai Lau * Load header option. Support reading a particular TCP header 4269938c3efdSQuentin Monnet * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 42700813a841SMartin KaFai Lau * 42710813a841SMartin KaFai Lau * If *flags* is 0, it will search the option from the 4272938c3efdSQuentin Monnet * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 42730813a841SMartin KaFai Lau * has details on what skb_data contains under different 4274938c3efdSQuentin Monnet * *skops*\ **->op**. 42750813a841SMartin KaFai Lau * 42760813a841SMartin KaFai Lau * The first byte of the *searchby_res* specifies the 42770813a841SMartin KaFai Lau * kind that it wants to search. 42780813a841SMartin KaFai Lau * 42790813a841SMartin KaFai Lau * If the searching kind is an experimental kind 42800813a841SMartin KaFai Lau * (i.e. 253 or 254 according to RFC6994). It also 42810813a841SMartin KaFai Lau * needs to specify the "magic" which is either 42820813a841SMartin KaFai Lau * 2 bytes or 4 bytes. It then also needs to 42830813a841SMartin KaFai Lau * specify the size of the magic by using 42840813a841SMartin KaFai Lau * the 2nd byte which is "kind-length" of a TCP 42850813a841SMartin KaFai Lau * header option and the "kind-length" also 42860813a841SMartin KaFai Lau * includes the first 2 bytes "kind" and "kind-length" 42870813a841SMartin KaFai Lau * itself as a normal TCP header option also does. 42880813a841SMartin KaFai Lau * 42890813a841SMartin KaFai Lau * For example, to search experimental kind 254 with 42900813a841SMartin KaFai Lau * 2 byte magic 0xeB9F, the searchby_res should be 42910813a841SMartin KaFai Lau * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 42920813a841SMartin KaFai Lau * 42930813a841SMartin KaFai Lau * To search for the standard window scale option (3), 4294938c3efdSQuentin Monnet * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 42950813a841SMartin KaFai Lau * Note, kind-length must be 0 for regular option. 42960813a841SMartin KaFai Lau * 42970813a841SMartin KaFai Lau * Searching for No-Op (0) and End-of-Option-List (1) are 42980813a841SMartin KaFai Lau * not supported. 42990813a841SMartin KaFai Lau * 43000813a841SMartin KaFai Lau * *len* must be at least 2 bytes which is the minimal size 43010813a841SMartin KaFai Lau * of a header option. 43020813a841SMartin KaFai Lau * 43030813a841SMartin KaFai Lau * Supported flags: 4304938c3efdSQuentin Monnet * 43050813a841SMartin KaFai Lau * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 43060813a841SMartin KaFai Lau * saved_syn packet or the just-received syn packet. 43070813a841SMartin KaFai Lau * 43080813a841SMartin KaFai Lau * Return 43090813a841SMartin KaFai Lau * > 0 when found, the header option is copied to *searchby_res*. 4310938c3efdSQuentin Monnet * The return value is the total length copied. On failure, a 4311938c3efdSQuentin Monnet * negative error code is returned: 43120813a841SMartin KaFai Lau * 4313938c3efdSQuentin Monnet * **-EINVAL** if a parameter is invalid. 43140813a841SMartin KaFai Lau * 4315938c3efdSQuentin Monnet * **-ENOMSG** if the option is not found. 43160813a841SMartin KaFai Lau * 4317938c3efdSQuentin Monnet * **-ENOENT** if no syn packet is available when 4318938c3efdSQuentin Monnet * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 43190813a841SMartin KaFai Lau * 4320938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space. Only *len* number of 43210813a841SMartin KaFai Lau * bytes are copied. 43220813a841SMartin KaFai Lau * 4323938c3efdSQuentin Monnet * **-EFAULT** on failure to parse the header options in the 4324938c3efdSQuentin Monnet * packet. 43250813a841SMartin KaFai Lau * 4326938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4327938c3efdSQuentin Monnet * *skops*\ **->op**. 43280813a841SMartin KaFai Lau * 43290813a841SMartin KaFai Lau * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 43300813a841SMartin KaFai Lau * Description 43310813a841SMartin KaFai Lau * Store header option. The data will be copied 43320813a841SMartin KaFai Lau * from buffer *from* with length *len* to the TCP header. 43330813a841SMartin KaFai Lau * 43340813a841SMartin KaFai Lau * The buffer *from* should have the whole option that 43350813a841SMartin KaFai Lau * includes the kind, kind-length, and the actual 43360813a841SMartin KaFai Lau * option data. The *len* must be at least kind-length 43370813a841SMartin KaFai Lau * long. The kind-length does not have to be 4 byte 43380813a841SMartin KaFai Lau * aligned. The kernel will take care of the padding 43390813a841SMartin KaFai Lau * and setting the 4 bytes aligned value to th->doff. 43400813a841SMartin KaFai Lau * 43410813a841SMartin KaFai Lau * This helper will check for duplicated option 43420813a841SMartin KaFai Lau * by searching the same option in the outgoing skb. 43430813a841SMartin KaFai Lau * 43440813a841SMartin KaFai Lau * This helper can only be called during 4345938c3efdSQuentin Monnet * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 43460813a841SMartin KaFai Lau * 43470813a841SMartin KaFai Lau * Return 43480813a841SMartin KaFai Lau * 0 on success, or negative error in case of failure: 43490813a841SMartin KaFai Lau * 4350938c3efdSQuentin Monnet * **-EINVAL** If param is invalid. 43510813a841SMartin KaFai Lau * 4352938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space in the header. 43530813a841SMartin KaFai Lau * Nothing has been written 43540813a841SMartin KaFai Lau * 4355938c3efdSQuentin Monnet * **-EEXIST** if the option already exists. 43560813a841SMartin KaFai Lau * 4357938c3efdSQuentin Monnet * **-EFAULT** on failrue to parse the existing header options. 43580813a841SMartin KaFai Lau * 4359938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4360938c3efdSQuentin Monnet * *skops*\ **->op**. 43610813a841SMartin KaFai Lau * 43620813a841SMartin KaFai Lau * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 43630813a841SMartin KaFai Lau * Description 43640813a841SMartin KaFai Lau * Reserve *len* bytes for the bpf header option. The 4365938c3efdSQuentin Monnet * space will be used by **bpf_store_hdr_opt**\ () later in 4366938c3efdSQuentin Monnet * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 43670813a841SMartin KaFai Lau * 4368938c3efdSQuentin Monnet * If **bpf_reserve_hdr_opt**\ () is called multiple times, 43690813a841SMartin KaFai Lau * the total number of bytes will be reserved. 43700813a841SMartin KaFai Lau * 43710813a841SMartin KaFai Lau * This helper can only be called during 4372938c3efdSQuentin Monnet * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 43730813a841SMartin KaFai Lau * 43740813a841SMartin KaFai Lau * Return 43750813a841SMartin KaFai Lau * 0 on success, or negative error in case of failure: 43760813a841SMartin KaFai Lau * 4377938c3efdSQuentin Monnet * **-EINVAL** if a parameter is invalid. 43780813a841SMartin KaFai Lau * 4379938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space in the header. 43800813a841SMartin KaFai Lau * 4381938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4382938c3efdSQuentin Monnet * *skops*\ **->op**. 43836e22ab9dSJiri Olsa * 43848ea63684SKP Singh * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 43858ea63684SKP Singh * Description 43868ea63684SKP Singh * Get a bpf_local_storage from an *inode*. 43878ea63684SKP Singh * 43888ea63684SKP Singh * Logically, it could be thought of as getting the value from 43898ea63684SKP Singh * a *map* with *inode* as the **key**. From this 43908ea63684SKP Singh * perspective, the usage is not much different from 43918ea63684SKP Singh * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 43928ea63684SKP Singh * helper enforces the key must be an inode and the map must also 43938ea63684SKP Singh * be a **BPF_MAP_TYPE_INODE_STORAGE**. 43948ea63684SKP Singh * 43958ea63684SKP Singh * Underneath, the value is stored locally at *inode* instead of 43968ea63684SKP Singh * the *map*. The *map* is used as the bpf-local-storage 43978ea63684SKP Singh * "type". The bpf-local-storage "type" (i.e. the *map*) is 43988ea63684SKP Singh * searched against all bpf_local_storage residing at *inode*. 43998ea63684SKP Singh * 44008ea63684SKP Singh * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 44018ea63684SKP Singh * used such that a new bpf_local_storage will be 44028ea63684SKP Singh * created if one does not exist. *value* can be used 44038ea63684SKP Singh * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 44048ea63684SKP Singh * the initial value of a bpf_local_storage. If *value* is 44058ea63684SKP Singh * **NULL**, the new bpf_local_storage will be zero initialized. 44068ea63684SKP Singh * Return 44078ea63684SKP Singh * A bpf_local_storage pointer is returned on success. 44088ea63684SKP Singh * 44098ea63684SKP Singh * **NULL** if not found or there was an error in adding 44108ea63684SKP Singh * a new bpf_local_storage. 44118ea63684SKP Singh * 44128ea63684SKP Singh * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 44138ea63684SKP Singh * Description 44148ea63684SKP Singh * Delete a bpf_local_storage from an *inode*. 44158ea63684SKP Singh * Return 44168ea63684SKP Singh * 0 on success. 44178ea63684SKP Singh * 44188ea63684SKP Singh * **-ENOENT** if the bpf_local_storage cannot be found. 44196e22ab9dSJiri Olsa * 44206e22ab9dSJiri Olsa * long bpf_d_path(struct path *path, char *buf, u32 sz) 44216e22ab9dSJiri Olsa * Description 4422938c3efdSQuentin Monnet * Return full path for given **struct path** object, which 4423938c3efdSQuentin Monnet * needs to be the kernel BTF *path* object. The path is 4424938c3efdSQuentin Monnet * returned in the provided buffer *buf* of size *sz* and 44256e22ab9dSJiri Olsa * is zero terminated. 44266e22ab9dSJiri Olsa * 44276e22ab9dSJiri Olsa * Return 44286e22ab9dSJiri Olsa * On success, the strictly positive length of the string, 44296e22ab9dSJiri Olsa * including the trailing NUL character. On error, a negative 44306e22ab9dSJiri Olsa * value. 443107be4c4aSAlexei Starovoitov * 443207be4c4aSAlexei Starovoitov * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 443307be4c4aSAlexei Starovoitov * Description 443407be4c4aSAlexei Starovoitov * Read *size* bytes from user space address *user_ptr* and store 4435938c3efdSQuentin Monnet * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 443607be4c4aSAlexei Starovoitov * Return 443707be4c4aSAlexei Starovoitov * 0 on success, or a negative error in case of failure. 4438c4d0bfb4SAlan Maguire * 4439c4d0bfb4SAlan Maguire * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4440c4d0bfb4SAlan Maguire * Description 4441c4d0bfb4SAlan Maguire * Use BTF to store a string representation of *ptr*->ptr in *str*, 4442c4d0bfb4SAlan Maguire * using *ptr*->type_id. This value should specify the type 4443c4d0bfb4SAlan Maguire * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4444c4d0bfb4SAlan Maguire * can be used to look up vmlinux BTF type ids. Traversing the 4445c4d0bfb4SAlan Maguire * data structure using BTF, the type information and values are 4446c4d0bfb4SAlan Maguire * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4447c4d0bfb4SAlan Maguire * the pointer data is carried out to avoid kernel crashes during 4448c4d0bfb4SAlan Maguire * operation. Smaller types can use string space on the stack; 4449c4d0bfb4SAlan Maguire * larger programs can use map data to store the string 4450c4d0bfb4SAlan Maguire * representation. 4451c4d0bfb4SAlan Maguire * 4452c4d0bfb4SAlan Maguire * The string can be subsequently shared with userspace via 4453c4d0bfb4SAlan Maguire * bpf_perf_event_output() or ring buffer interfaces. 4454c4d0bfb4SAlan Maguire * bpf_trace_printk() is to be avoided as it places too small 4455c4d0bfb4SAlan Maguire * a limit on string size to be useful. 4456c4d0bfb4SAlan Maguire * 4457c4d0bfb4SAlan Maguire * *flags* is a combination of 4458c4d0bfb4SAlan Maguire * 4459c4d0bfb4SAlan Maguire * **BTF_F_COMPACT** 4460c4d0bfb4SAlan Maguire * no formatting around type information 4461c4d0bfb4SAlan Maguire * **BTF_F_NONAME** 4462c4d0bfb4SAlan Maguire * no struct/union member names/types 4463c4d0bfb4SAlan Maguire * **BTF_F_PTR_RAW** 4464c4d0bfb4SAlan Maguire * show raw (unobfuscated) pointer values; 4465c4d0bfb4SAlan Maguire * equivalent to printk specifier %px. 4466c4d0bfb4SAlan Maguire * **BTF_F_ZERO** 4467c4d0bfb4SAlan Maguire * show zero-valued struct/union members; they 4468c4d0bfb4SAlan Maguire * are not displayed by default 4469c4d0bfb4SAlan Maguire * 4470c4d0bfb4SAlan Maguire * Return 4471c4d0bfb4SAlan Maguire * The number of bytes that were written (or would have been 4472c4d0bfb4SAlan Maguire * written if output had to be truncated due to string size), 4473c4d0bfb4SAlan Maguire * or a negative error in cases of failure. 4474eb411377SAlan Maguire * 4475eb411377SAlan Maguire * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4476eb411377SAlan Maguire * Description 4477eb411377SAlan Maguire * Use BTF to write to seq_write a string representation of 4478eb411377SAlan Maguire * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4479eb411377SAlan Maguire * *flags* are identical to those used for bpf_snprintf_btf. 4480eb411377SAlan Maguire * Return 4481eb411377SAlan Maguire * 0 on success or a negative error in case of failure. 4482b426ce83SDaniel Borkmann * 4483b426ce83SDaniel Borkmann * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4484b426ce83SDaniel Borkmann * Description 4485b426ce83SDaniel Borkmann * See **bpf_get_cgroup_classid**\ () for the main description. 4486b426ce83SDaniel Borkmann * This helper differs from **bpf_get_cgroup_classid**\ () in that 4487b426ce83SDaniel Borkmann * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4488b426ce83SDaniel Borkmann * associated socket instead of the current process. 4489b426ce83SDaniel Borkmann * Return 4490b426ce83SDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 4491b4ab3141SDaniel Borkmann * 4492ba452c9eSToke Høiland-Jørgensen * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4493b4ab3141SDaniel Borkmann * Description 4494b4ab3141SDaniel Borkmann * Redirect the packet to another net device of index *ifindex* 4495b4ab3141SDaniel Borkmann * and fill in L2 addresses from neighboring subsystem. This helper 4496b4ab3141SDaniel Borkmann * is somewhat similar to **bpf_redirect**\ (), except that it 4497dd2ce6a5SDaniel Borkmann * populates L2 addresses as well, meaning, internally, the helper 4498ba452c9eSToke Høiland-Jørgensen * relies on the neighbor lookup for the L2 address of the nexthop. 4499ba452c9eSToke Høiland-Jørgensen * 4500ba452c9eSToke Høiland-Jørgensen * The helper will perform a FIB lookup based on the skb's 4501ba452c9eSToke Høiland-Jørgensen * networking header to get the address of the next hop, unless 4502ba452c9eSToke Høiland-Jørgensen * this is supplied by the caller in the *params* argument. The 4503ba452c9eSToke Høiland-Jørgensen * *plen* argument indicates the len of *params* and should be set 4504ba452c9eSToke Høiland-Jørgensen * to 0 if *params* is NULL. 4505dd2ce6a5SDaniel Borkmann * 4506b4ab3141SDaniel Borkmann * The *flags* argument is reserved and must be 0. The helper is 4507dd2ce6a5SDaniel Borkmann * currently only supported for tc BPF program types, and enabled 4508dd2ce6a5SDaniel Borkmann * for IPv4 and IPv6 protocols. 4509b4ab3141SDaniel Borkmann * Return 4510b4ab3141SDaniel Borkmann * The helper returns **TC_ACT_REDIRECT** on success or 4511b4ab3141SDaniel Borkmann * **TC_ACT_SHOT** on error. 4512eaa6bcb7SHao Luo * 4513eaa6bcb7SHao Luo * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4514eaa6bcb7SHao Luo * Description 4515eaa6bcb7SHao Luo * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4516eaa6bcb7SHao Luo * pointer to the percpu kernel variable on *cpu*. A ksym is an 4517eaa6bcb7SHao Luo * extern variable decorated with '__ksym'. For ksym, there is a 4518eaa6bcb7SHao Luo * global var (either static or global) defined of the same name 4519eaa6bcb7SHao Luo * in the kernel. The ksym is percpu if the global var is percpu. 4520eaa6bcb7SHao Luo * The returned pointer points to the global percpu var on *cpu*. 4521eaa6bcb7SHao Luo * 4522eaa6bcb7SHao Luo * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4523eaa6bcb7SHao Luo * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4524eaa6bcb7SHao Luo * happens if *cpu* is larger than nr_cpu_ids. The caller of 4525eaa6bcb7SHao Luo * bpf_per_cpu_ptr() must check the returned value. 4526eaa6bcb7SHao Luo * Return 4527eaa6bcb7SHao Luo * A pointer pointing to the kernel percpu variable on *cpu*, or 4528eaa6bcb7SHao Luo * NULL, if *cpu* is invalid. 452963d9b80dSHao Luo * 453063d9b80dSHao Luo * void *bpf_this_cpu_ptr(const void *percpu_ptr) 453163d9b80dSHao Luo * Description 453263d9b80dSHao Luo * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 453363d9b80dSHao Luo * pointer to the percpu kernel variable on this cpu. See the 453463d9b80dSHao Luo * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 453563d9b80dSHao Luo * 453663d9b80dSHao Luo * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 453763d9b80dSHao Luo * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 453863d9b80dSHao Luo * never return NULL. 453963d9b80dSHao Luo * Return 454063d9b80dSHao Luo * A pointer pointing to the kernel percpu variable on this cpu. 45419aa1206eSDaniel Borkmann * 45429aa1206eSDaniel Borkmann * long bpf_redirect_peer(u32 ifindex, u64 flags) 45439aa1206eSDaniel Borkmann * Description 45449aa1206eSDaniel Borkmann * Redirect the packet to another net device of index *ifindex*. 45459aa1206eSDaniel Borkmann * This helper is somewhat similar to **bpf_redirect**\ (), except 45469aa1206eSDaniel Borkmann * that the redirection happens to the *ifindex*' peer device and 45479aa1206eSDaniel Borkmann * the netns switch takes place from ingress to ingress without 45489aa1206eSDaniel Borkmann * going through the CPU's backlog queue. 45499aa1206eSDaniel Borkmann * 45509aa1206eSDaniel Borkmann * The *flags* argument is reserved and must be 0. The helper is 45519aa1206eSDaniel Borkmann * currently only supported for tc BPF program types at the ingress 45529aa1206eSDaniel Borkmann * hook and for veth device types. The peer device must reside in a 45539aa1206eSDaniel Borkmann * different network namespace. 45549aa1206eSDaniel Borkmann * Return 45559aa1206eSDaniel Borkmann * The helper returns **TC_ACT_REDIRECT** on success or 45569aa1206eSDaniel Borkmann * **TC_ACT_SHOT** on error. 45574cf1bc1fSKP Singh * 45584cf1bc1fSKP Singh * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 45594cf1bc1fSKP Singh * Description 45604cf1bc1fSKP Singh * Get a bpf_local_storage from the *task*. 45614cf1bc1fSKP Singh * 45624cf1bc1fSKP Singh * Logically, it could be thought of as getting the value from 45634cf1bc1fSKP Singh * a *map* with *task* as the **key**. From this 45644cf1bc1fSKP Singh * perspective, the usage is not much different from 45654cf1bc1fSKP Singh * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 45664cf1bc1fSKP Singh * helper enforces the key must be an task_struct and the map must also 45674cf1bc1fSKP Singh * be a **BPF_MAP_TYPE_TASK_STORAGE**. 45684cf1bc1fSKP Singh * 45694cf1bc1fSKP Singh * Underneath, the value is stored locally at *task* instead of 45704cf1bc1fSKP Singh * the *map*. The *map* is used as the bpf-local-storage 45714cf1bc1fSKP Singh * "type". The bpf-local-storage "type" (i.e. the *map*) is 45724cf1bc1fSKP Singh * searched against all bpf_local_storage residing at *task*. 45734cf1bc1fSKP Singh * 45744cf1bc1fSKP Singh * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 45754cf1bc1fSKP Singh * used such that a new bpf_local_storage will be 45764cf1bc1fSKP Singh * created if one does not exist. *value* can be used 45774cf1bc1fSKP Singh * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 45784cf1bc1fSKP Singh * the initial value of a bpf_local_storage. If *value* is 45794cf1bc1fSKP Singh * **NULL**, the new bpf_local_storage will be zero initialized. 45804cf1bc1fSKP Singh * Return 45814cf1bc1fSKP Singh * A bpf_local_storage pointer is returned on success. 45824cf1bc1fSKP Singh * 45834cf1bc1fSKP Singh * **NULL** if not found or there was an error in adding 45844cf1bc1fSKP Singh * a new bpf_local_storage. 45854cf1bc1fSKP Singh * 45864cf1bc1fSKP Singh * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 45874cf1bc1fSKP Singh * Description 45884cf1bc1fSKP Singh * Delete a bpf_local_storage from a *task*. 45894cf1bc1fSKP Singh * Return 45904cf1bc1fSKP Singh * 0 on success. 45914cf1bc1fSKP Singh * 45924cf1bc1fSKP Singh * **-ENOENT** if the bpf_local_storage cannot be found. 45933ca1032aSKP Singh * 45943ca1032aSKP Singh * struct task_struct *bpf_get_current_task_btf(void) 45953ca1032aSKP Singh * Description 45963ca1032aSKP Singh * Return a BTF pointer to the "current" task. 45973ca1032aSKP Singh * This pointer can also be used in helpers that accept an 45983ca1032aSKP Singh * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 45993ca1032aSKP Singh * Return 46003ca1032aSKP Singh * Pointer to the current task. 46013f6719c7SKP Singh * 46023f6719c7SKP Singh * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 46033f6719c7SKP Singh * Description 46043f6719c7SKP Singh * Set or clear certain options on *bprm*: 46053f6719c7SKP Singh * 46063f6719c7SKP Singh * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 46073f6719c7SKP Singh * which sets the **AT_SECURE** auxv for glibc. The bit 46083f6719c7SKP Singh * is cleared if the flag is not specified. 46093f6719c7SKP Singh * Return 46103f6719c7SKP Singh * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4611d0551261SDmitrii Banshchikov * 4612d0551261SDmitrii Banshchikov * u64 bpf_ktime_get_coarse_ns(void) 4613d0551261SDmitrii Banshchikov * Description 4614d0551261SDmitrii Banshchikov * Return a coarse-grained version of the time elapsed since 4615d0551261SDmitrii Banshchikov * system boot, in nanoseconds. Does not include time the system 4616d0551261SDmitrii Banshchikov * was suspended. 4617d0551261SDmitrii Banshchikov * 4618d0551261SDmitrii Banshchikov * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4619d0551261SDmitrii Banshchikov * Return 4620d0551261SDmitrii Banshchikov * Current *ktime*. 462127672f0dSKP Singh * 462227672f0dSKP Singh * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 462327672f0dSKP Singh * Description 462427672f0dSKP Singh * Returns the stored IMA hash of the *inode* (if it's avaialable). 462527672f0dSKP Singh * If the hash is larger than *size*, then only *size* 462627672f0dSKP Singh * bytes will be copied to *dst* 462727672f0dSKP Singh * Return 462827672f0dSKP Singh * The **hash_algo** is returned on success, 462927672f0dSKP Singh * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 463027672f0dSKP Singh * invalid arguments are passed. 46314f19cab7SFlorent Revest * 46324f19cab7SFlorent Revest * struct socket *bpf_sock_from_file(struct file *file) 46334f19cab7SFlorent Revest * Description 46344f19cab7SFlorent Revest * If the given file represents a socket, returns the associated 46354f19cab7SFlorent Revest * socket. 46364f19cab7SFlorent Revest * Return 46374f19cab7SFlorent Revest * A pointer to a struct socket on success or NULL if the file is 46384f19cab7SFlorent Revest * not a socket. 463934b2021cSJesper Dangaard Brouer * 464034b2021cSJesper Dangaard Brouer * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 464134b2021cSJesper Dangaard Brouer * Description 4642e5e35e75SJesper Dangaard Brouer * Check packet size against exceeding MTU of net device (based 464334b2021cSJesper Dangaard Brouer * on *ifindex*). This helper will likely be used in combination 464434b2021cSJesper Dangaard Brouer * with helpers that adjust/change the packet size. 464534b2021cSJesper Dangaard Brouer * 464634b2021cSJesper Dangaard Brouer * The argument *len_diff* can be used for querying with a planned 464734b2021cSJesper Dangaard Brouer * size change. This allows to check MTU prior to changing packet 464834b2021cSJesper Dangaard Brouer * ctx. Providing an *len_diff* adjustment that is larger than the 464934b2021cSJesper Dangaard Brouer * actual packet size (resulting in negative packet size) will in 465034b2021cSJesper Dangaard Brouer * principle not exceed the MTU, why it is not considered a 465134b2021cSJesper Dangaard Brouer * failure. Other BPF-helpers are needed for performing the 465234b2021cSJesper Dangaard Brouer * planned size change, why the responsability for catch a negative 465334b2021cSJesper Dangaard Brouer * packet size belong in those helpers. 465434b2021cSJesper Dangaard Brouer * 465534b2021cSJesper Dangaard Brouer * Specifying *ifindex* zero means the MTU check is performed 465634b2021cSJesper Dangaard Brouer * against the current net device. This is practical if this isn't 465734b2021cSJesper Dangaard Brouer * used prior to redirect. 465834b2021cSJesper Dangaard Brouer * 4659e5e35e75SJesper Dangaard Brouer * On input *mtu_len* must be a valid pointer, else verifier will 4660e5e35e75SJesper Dangaard Brouer * reject BPF program. If the value *mtu_len* is initialized to 4661e5e35e75SJesper Dangaard Brouer * zero then the ctx packet size is use. When value *mtu_len* is 4662e5e35e75SJesper Dangaard Brouer * provided as input this specify the L3 length that the MTU check 4663e5e35e75SJesper Dangaard Brouer * is done against. Remember XDP and TC length operate at L2, but 4664e5e35e75SJesper Dangaard Brouer * this value is L3 as this correlate to MTU and IP-header tot_len 4665e5e35e75SJesper Dangaard Brouer * values which are L3 (similar behavior as bpf_fib_lookup). 4666e5e35e75SJesper Dangaard Brouer * 466734b2021cSJesper Dangaard Brouer * The Linux kernel route table can configure MTUs on a more 466834b2021cSJesper Dangaard Brouer * specific per route level, which is not provided by this helper. 466934b2021cSJesper Dangaard Brouer * For route level MTU checks use the **bpf_fib_lookup**\ () 467034b2021cSJesper Dangaard Brouer * helper. 467134b2021cSJesper Dangaard Brouer * 467234b2021cSJesper Dangaard Brouer * *ctx* is either **struct xdp_md** for XDP programs or 467334b2021cSJesper Dangaard Brouer * **struct sk_buff** for tc cls_act programs. 467434b2021cSJesper Dangaard Brouer * 467534b2021cSJesper Dangaard Brouer * The *flags* argument can be a combination of one or more of the 467634b2021cSJesper Dangaard Brouer * following values: 467734b2021cSJesper Dangaard Brouer * 467834b2021cSJesper Dangaard Brouer * **BPF_MTU_CHK_SEGS** 467934b2021cSJesper Dangaard Brouer * This flag will only works for *ctx* **struct sk_buff**. 468034b2021cSJesper Dangaard Brouer * If packet context contains extra packet segment buffers 468134b2021cSJesper Dangaard Brouer * (often knows as GSO skb), then MTU check is harder to 468234b2021cSJesper Dangaard Brouer * check at this point, because in transmit path it is 468334b2021cSJesper Dangaard Brouer * possible for the skb packet to get re-segmented 468434b2021cSJesper Dangaard Brouer * (depending on net device features). This could still be 468534b2021cSJesper Dangaard Brouer * a MTU violation, so this flag enables performing MTU 468634b2021cSJesper Dangaard Brouer * check against segments, with a different violation 468734b2021cSJesper Dangaard Brouer * return code to tell it apart. Check cannot use len_diff. 468834b2021cSJesper Dangaard Brouer * 468934b2021cSJesper Dangaard Brouer * On return *mtu_len* pointer contains the MTU value of the net 469034b2021cSJesper Dangaard Brouer * device. Remember the net device configured MTU is the L3 size, 4691e5e35e75SJesper Dangaard Brouer * which is returned here and XDP and TC length operate at L2. 469234b2021cSJesper Dangaard Brouer * Helper take this into account for you, but remember when using 4693e5e35e75SJesper Dangaard Brouer * MTU value in your BPF-code. 469434b2021cSJesper Dangaard Brouer * 469534b2021cSJesper Dangaard Brouer * Return 469634b2021cSJesper Dangaard Brouer * * 0 on success, and populate MTU value in *mtu_len* pointer. 469734b2021cSJesper Dangaard Brouer * 469834b2021cSJesper Dangaard Brouer * * < 0 if any input argument is invalid (*mtu_len* not updated) 469934b2021cSJesper Dangaard Brouer * 470034b2021cSJesper Dangaard Brouer * MTU violations return positive values, but also populate MTU 470134b2021cSJesper Dangaard Brouer * value in *mtu_len* pointer, as this can be needed for 470234b2021cSJesper Dangaard Brouer * implementing PMTU handing: 470334b2021cSJesper Dangaard Brouer * 470434b2021cSJesper Dangaard Brouer * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 470534b2021cSJesper Dangaard Brouer * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 470634b2021cSJesper Dangaard Brouer * 470769c087baSYonghong Song * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 470869c087baSYonghong Song * Description 470969c087baSYonghong Song * For each element in **map**, call **callback_fn** function with 471069c087baSYonghong Song * **map**, **callback_ctx** and other map-specific parameters. 471169c087baSYonghong Song * The **callback_fn** should be a static function and 471269c087baSYonghong Song * the **callback_ctx** should be a pointer to the stack. 471369c087baSYonghong Song * The **flags** is used to control certain aspects of the helper. 471469c087baSYonghong Song * Currently, the **flags** must be 0. 471569c087baSYonghong Song * 471669c087baSYonghong Song * The following are a list of supported map types and their 471769c087baSYonghong Song * respective expected callback signatures: 471869c087baSYonghong Song * 471969c087baSYonghong Song * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 472069c087baSYonghong Song * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 472169c087baSYonghong Song * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 472269c087baSYonghong Song * 472369c087baSYonghong Song * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 472469c087baSYonghong Song * 472569c087baSYonghong Song * For per_cpu maps, the map_value is the value on the cpu where the 472669c087baSYonghong Song * bpf_prog is running. 472769c087baSYonghong Song * 472869c087baSYonghong Song * If **callback_fn** return 0, the helper will continue to the next 472969c087baSYonghong Song * element. If return value is 1, the helper will skip the rest of 473069c087baSYonghong Song * elements and return. Other return values are not used now. 473169c087baSYonghong Song * 473269c087baSYonghong Song * Return 473369c087baSYonghong Song * The number of traversed map elements for success, **-EINVAL** for 473469c087baSYonghong Song * invalid **flags**. 47357b15523aSFlorent Revest * 47367b15523aSFlorent Revest * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 47377b15523aSFlorent Revest * Description 47387b15523aSFlorent Revest * Outputs a string into the **str** buffer of size **str_size** 47397b15523aSFlorent Revest * based on a format string stored in a read-only map pointed by 47407b15523aSFlorent Revest * **fmt**. 47417b15523aSFlorent Revest * 47427b15523aSFlorent Revest * Each format specifier in **fmt** corresponds to one u64 element 47437b15523aSFlorent Revest * in the **data** array. For strings and pointers where pointees 47447b15523aSFlorent Revest * are accessed, only the pointer values are stored in the *data* 47457b15523aSFlorent Revest * array. The *data_len* is the size of *data* in bytes. 47467b15523aSFlorent Revest * 47477b15523aSFlorent Revest * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 47487b15523aSFlorent Revest * memory. Reading kernel memory may fail due to either invalid 47497b15523aSFlorent Revest * address or valid address but requiring a major memory fault. If 47507b15523aSFlorent Revest * reading kernel memory fails, the string for **%s** will be an 47517b15523aSFlorent Revest * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 47527b15523aSFlorent Revest * Not returning error to bpf program is consistent with what 47537b15523aSFlorent Revest * **bpf_trace_printk**\ () does for now. 47547b15523aSFlorent Revest * 47557b15523aSFlorent Revest * Return 47567b15523aSFlorent Revest * The strictly positive length of the formatted string, including 47577b15523aSFlorent Revest * the trailing zero character. If the return value is greater than 47587b15523aSFlorent Revest * **str_size**, **str** contains a truncated string, guaranteed to 47597b15523aSFlorent Revest * be zero-terminated except when **str_size** is 0. 47607b15523aSFlorent Revest * 47617b15523aSFlorent Revest * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 476279a7f8bdSAlexei Starovoitov * 476379a7f8bdSAlexei Starovoitov * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 476479a7f8bdSAlexei Starovoitov * Description 476579a7f8bdSAlexei Starovoitov * Execute bpf syscall with given arguments. 476679a7f8bdSAlexei Starovoitov * Return 476779a7f8bdSAlexei Starovoitov * A syscall result. 47683d78417bSAlexei Starovoitov * 47693d78417bSAlexei Starovoitov * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 47703d78417bSAlexei Starovoitov * Description 47713d78417bSAlexei Starovoitov * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 47723d78417bSAlexei Starovoitov * Return 47733d78417bSAlexei Starovoitov * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 47743abea089SAlexei Starovoitov * 47753abea089SAlexei Starovoitov * long bpf_sys_close(u32 fd) 47763abea089SAlexei Starovoitov * Description 47773abea089SAlexei Starovoitov * Execute close syscall for given FD. 47783abea089SAlexei Starovoitov * Return 47793abea089SAlexei Starovoitov * A syscall result. 4780*b00628b1SAlexei Starovoitov * 4781*b00628b1SAlexei Starovoitov * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 4782*b00628b1SAlexei Starovoitov * Description 4783*b00628b1SAlexei Starovoitov * Initialize the timer. 4784*b00628b1SAlexei Starovoitov * First 4 bits of *flags* specify clockid. 4785*b00628b1SAlexei Starovoitov * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 4786*b00628b1SAlexei Starovoitov * All other bits of *flags* are reserved. 4787*b00628b1SAlexei Starovoitov * The verifier will reject the program if *timer* is not from 4788*b00628b1SAlexei Starovoitov * the same *map*. 4789*b00628b1SAlexei Starovoitov * Return 4790*b00628b1SAlexei Starovoitov * 0 on success. 4791*b00628b1SAlexei Starovoitov * **-EBUSY** if *timer* is already initialized. 4792*b00628b1SAlexei Starovoitov * **-EINVAL** if invalid *flags* are passed. 4793*b00628b1SAlexei Starovoitov * **-EPERM** if *timer* is in a map that doesn't have any user references. 4794*b00628b1SAlexei Starovoitov * The user space should either hold a file descriptor to a map with timers 4795*b00628b1SAlexei Starovoitov * or pin such map in bpffs. When map is unpinned or file descriptor is 4796*b00628b1SAlexei Starovoitov * closed all timers in the map will be cancelled and freed. 4797*b00628b1SAlexei Starovoitov * 4798*b00628b1SAlexei Starovoitov * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 4799*b00628b1SAlexei Starovoitov * Description 4800*b00628b1SAlexei Starovoitov * Configure the timer to call *callback_fn* static function. 4801*b00628b1SAlexei Starovoitov * Return 4802*b00628b1SAlexei Starovoitov * 0 on success. 4803*b00628b1SAlexei Starovoitov * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4804*b00628b1SAlexei Starovoitov * **-EPERM** if *timer* is in a map that doesn't have any user references. 4805*b00628b1SAlexei Starovoitov * The user space should either hold a file descriptor to a map with timers 4806*b00628b1SAlexei Starovoitov * or pin such map in bpffs. When map is unpinned or file descriptor is 4807*b00628b1SAlexei Starovoitov * closed all timers in the map will be cancelled and freed. 4808*b00628b1SAlexei Starovoitov * 4809*b00628b1SAlexei Starovoitov * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 4810*b00628b1SAlexei Starovoitov * Description 4811*b00628b1SAlexei Starovoitov * Set timer expiration N nanoseconds from the current time. The 4812*b00628b1SAlexei Starovoitov * configured callback will be invoked in soft irq context on some cpu 4813*b00628b1SAlexei Starovoitov * and will not repeat unless another bpf_timer_start() is made. 4814*b00628b1SAlexei Starovoitov * In such case the next invocation can migrate to a different cpu. 4815*b00628b1SAlexei Starovoitov * Since struct bpf_timer is a field inside map element the map 4816*b00628b1SAlexei Starovoitov * owns the timer. The bpf_timer_set_callback() will increment refcnt 4817*b00628b1SAlexei Starovoitov * of BPF program to make sure that callback_fn code stays valid. 4818*b00628b1SAlexei Starovoitov * When user space reference to a map reaches zero all timers 4819*b00628b1SAlexei Starovoitov * in a map are cancelled and corresponding program's refcnts are 4820*b00628b1SAlexei Starovoitov * decremented. This is done to make sure that Ctrl-C of a user 4821*b00628b1SAlexei Starovoitov * process doesn't leave any timers running. If map is pinned in 4822*b00628b1SAlexei Starovoitov * bpffs the callback_fn can re-arm itself indefinitely. 4823*b00628b1SAlexei Starovoitov * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 4824*b00628b1SAlexei Starovoitov * cancel and free the timer in the given map element. 4825*b00628b1SAlexei Starovoitov * The map can contain timers that invoke callback_fn-s from different 4826*b00628b1SAlexei Starovoitov * programs. The same callback_fn can serve different timers from 4827*b00628b1SAlexei Starovoitov * different maps if key/value layout matches across maps. 4828*b00628b1SAlexei Starovoitov * Every bpf_timer_set_callback() can have different callback_fn. 4829*b00628b1SAlexei Starovoitov * 4830*b00628b1SAlexei Starovoitov * Return 4831*b00628b1SAlexei Starovoitov * 0 on success. 4832*b00628b1SAlexei Starovoitov * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 4833*b00628b1SAlexei Starovoitov * or invalid *flags* are passed. 4834*b00628b1SAlexei Starovoitov * 4835*b00628b1SAlexei Starovoitov * long bpf_timer_cancel(struct bpf_timer *timer) 4836*b00628b1SAlexei Starovoitov * Description 4837*b00628b1SAlexei Starovoitov * Cancel the timer and wait for callback_fn to finish if it was running. 4838*b00628b1SAlexei Starovoitov * Return 4839*b00628b1SAlexei Starovoitov * 0 if the timer was not active. 4840*b00628b1SAlexei Starovoitov * 1 if the timer was active. 4841*b00628b1SAlexei Starovoitov * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4842*b00628b1SAlexei Starovoitov * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 4843*b00628b1SAlexei Starovoitov * own timer which would have led to a deadlock otherwise. 48447a4b28c6SDaniel Borkmann */ 4845ebb676daSThomas Graf #define __BPF_FUNC_MAPPER(FN) \ 4846ebb676daSThomas Graf FN(unspec), \ 4847ebb676daSThomas Graf FN(map_lookup_elem), \ 4848ebb676daSThomas Graf FN(map_update_elem), \ 4849ebb676daSThomas Graf FN(map_delete_elem), \ 4850ebb676daSThomas Graf FN(probe_read), \ 4851ebb676daSThomas Graf FN(ktime_get_ns), \ 4852ebb676daSThomas Graf FN(trace_printk), \ 4853ebb676daSThomas Graf FN(get_prandom_u32), \ 4854ebb676daSThomas Graf FN(get_smp_processor_id), \ 4855ebb676daSThomas Graf FN(skb_store_bytes), \ 4856ebb676daSThomas Graf FN(l3_csum_replace), \ 4857ebb676daSThomas Graf FN(l4_csum_replace), \ 4858ebb676daSThomas Graf FN(tail_call), \ 4859ebb676daSThomas Graf FN(clone_redirect), \ 4860ebb676daSThomas Graf FN(get_current_pid_tgid), \ 4861ebb676daSThomas Graf FN(get_current_uid_gid), \ 4862ebb676daSThomas Graf FN(get_current_comm), \ 4863ebb676daSThomas Graf FN(get_cgroup_classid), \ 4864ebb676daSThomas Graf FN(skb_vlan_push), \ 4865ebb676daSThomas Graf FN(skb_vlan_pop), \ 4866ebb676daSThomas Graf FN(skb_get_tunnel_key), \ 4867ebb676daSThomas Graf FN(skb_set_tunnel_key), \ 4868ebb676daSThomas Graf FN(perf_event_read), \ 4869ebb676daSThomas Graf FN(redirect), \ 4870ebb676daSThomas Graf FN(get_route_realm), \ 4871ebb676daSThomas Graf FN(perf_event_output), \ 4872ebb676daSThomas Graf FN(skb_load_bytes), \ 4873ebb676daSThomas Graf FN(get_stackid), \ 4874ebb676daSThomas Graf FN(csum_diff), \ 4875ebb676daSThomas Graf FN(skb_get_tunnel_opt), \ 4876ebb676daSThomas Graf FN(skb_set_tunnel_opt), \ 4877ebb676daSThomas Graf FN(skb_change_proto), \ 4878ebb676daSThomas Graf FN(skb_change_type), \ 4879ebb676daSThomas Graf FN(skb_under_cgroup), \ 4880ebb676daSThomas Graf FN(get_hash_recalc), \ 4881ebb676daSThomas Graf FN(get_current_task), \ 4882ebb676daSThomas Graf FN(probe_write_user), \ 4883ebb676daSThomas Graf FN(current_task_under_cgroup), \ 4884ebb676daSThomas Graf FN(skb_change_tail), \ 4885ebb676daSThomas Graf FN(skb_pull_data), \ 4886ebb676daSThomas Graf FN(csum_update), \ 4887ebb676daSThomas Graf FN(set_hash_invalid), \ 48883a0af8fdSThomas Graf FN(get_numa_node_id), \ 488917bedab2SMartin KaFai Lau FN(skb_change_head), \ 4890a5e8c070SGianluca Borello FN(xdp_adjust_head), \ 489191b8270fSChenbo Feng FN(probe_read_str), \ 48926acc5c29SChenbo Feng FN(get_socket_cookie), \ 4893ded092cdSDaniel Borkmann FN(get_socket_uid), \ 48948c4b4c7eSLawrence Brakmo FN(set_hash), \ 48952be7e212SDaniel Borkmann FN(setsockopt), \ 489697f91a7cSJohn Fastabend FN(skb_adjust_room), \ 4897174a79ffSJohn Fastabend FN(redirect_map), \ 4898174a79ffSJohn Fastabend FN(sk_redirect_map), \ 4899174a79ffSJohn Fastabend FN(sock_map_update), \ 4900908432caSYonghong Song FN(xdp_adjust_meta), \ 49014bebdc7aSYonghong Song FN(perf_event_read_value), \ 4902cd86d1fdSLawrence Brakmo FN(perf_prog_read_value), \ 49039802d865SJosef Bacik FN(getsockopt), \ 4904b13d8807SLawrence Brakmo FN(override_return), \ 49054f738adbSJohn Fastabend FN(sock_ops_cb_flags_set), \ 49062a100317SJohn Fastabend FN(msg_redirect_map), \ 490791843d54SJohn Fastabend FN(msg_apply_bytes), \ 4908015632bbSJohn Fastabend FN(msg_cork_bytes), \ 4909d74bad4eSAndrey Ignatov FN(msg_pull_data), \ 4910b32cc5b9SNikita V. Shirokov FN(bind), \ 491112bed760SEyal Birger FN(xdp_adjust_tail), \ 4912c195651eSYonghong Song FN(skb_get_xfrm_state), \ 49134e1ec56cSDaniel Borkmann FN(get_stack), \ 491487f5fc7eSDavid Ahern FN(skb_load_bytes_relative), \ 491581110384SJohn Fastabend FN(fib_lookup), \ 491681110384SJohn Fastabend FN(sock_hash_update), \ 491781110384SJohn Fastabend FN(msg_redirect_hash), \ 4918fe94cc29SMathieu Xhonneux FN(sk_redirect_hash), \ 4919fe94cc29SMathieu Xhonneux FN(lwt_push_encap), \ 4920fe94cc29SMathieu Xhonneux FN(lwt_seg6_store_bytes), \ 4921fe94cc29SMathieu Xhonneux FN(lwt_seg6_adjust_srh), \ 4922f4364dcfSSean Young FN(lwt_seg6_action), \ 4923f4364dcfSSean Young FN(rc_repeat), \ 4924cb20b08eSDaniel Borkmann FN(rc_keydown), \ 4925bf6fa2c8SYonghong Song FN(skb_cgroup_id), \ 4926cd339431SRoman Gushchin FN(get_current_cgroup_id), \ 49272dbb9b9eSMartin KaFai Lau FN(get_local_storage), \ 492877236281SAndrey Ignatov FN(sk_select_reuseport), \ 49296acc9b43SJoe Stringer FN(skb_ancestor_cgroup_id), \ 49306acc9b43SJoe Stringer FN(sk_lookup_tcp), \ 49316acc9b43SJoe Stringer FN(sk_lookup_udp), \ 4932f1a2e44aSMauricio Vasquez B FN(sk_release), \ 4933f1a2e44aSMauricio Vasquez B FN(map_push_elem), \ 4934f1a2e44aSMauricio Vasquez B FN(map_pop_elem), \ 49356fff607eSJohn Fastabend FN(map_peek_elem), \ 49367246d8edSJohn Fastabend FN(msg_push_data), \ 493701d3240aSSean Young FN(msg_pop_data), \ 4938d83525caSAlexei Starovoitov FN(rc_pointer_rel), \ 4939d83525caSAlexei Starovoitov FN(spin_lock), \ 494046f8bc92SMartin KaFai Lau FN(spin_unlock), \ 4941655a51e5SMartin KaFai Lau FN(sk_fullsock), \ 4942f7c917baSbrakmo FN(tcp_sock), \ 4943dbafd7ddSMartin KaFai Lau FN(skb_ecn_set_ce), \ 4944edbf8c01SLorenz Bauer FN(get_listener_sock), \ 494539904084SLorenz Bauer FN(skc_lookup_tcp), \ 4946808649fbSAndrey Ignatov FN(tcp_check_syncookie), \ 49471d11b301SAndrey Ignatov FN(sysctl_get_name), \ 49484e63acdfSAndrey Ignatov FN(sysctl_get_current_value), \ 49494e63acdfSAndrey Ignatov FN(sysctl_get_new_value), \ 4950d7a4cb9bSAndrey Ignatov FN(sysctl_set_new_value), \ 4951d7a4cb9bSAndrey Ignatov FN(strtol), \ 49526ac99e8fSMartin KaFai Lau FN(strtoul), \ 49536ac99e8fSMartin KaFai Lau FN(sk_storage_get), \ 49548b401f9eSYonghong Song FN(sk_storage_delete), \ 495570d66244SPetar Penkov FN(send_signal), \ 4956a7658e1aSAlexei Starovoitov FN(tcp_gen_syncookie), \ 49576ae08ae3SDaniel Borkmann FN(skb_output), \ 49586ae08ae3SDaniel Borkmann FN(probe_read_user), \ 49596ae08ae3SDaniel Borkmann FN(probe_read_kernel), \ 49606ae08ae3SDaniel Borkmann FN(probe_read_user_str), \ 4961206057feSMartin KaFai Lau FN(probe_read_kernel_str), \ 49628482941fSYonghong Song FN(tcp_send_ack), \ 49635576b991SMartin KaFai Lau FN(send_signal_thread), \ 4964fff7b643SDaniel Xu FN(jiffies64), \ 4965b4490c5cSCarlos Neira FN(read_branch_records), \ 4966d831ee84SEelco Chaudron FN(get_ns_current_pid_tgid), \ 4967f318903cSDaniel Borkmann FN(xdp_output), \ 49680f09abd1SDaniel Borkmann FN(get_netns_cookie), \ 4969cf7fbe66SJoe Stringer FN(get_current_ancestor_cgroup_id), \ 497071d19214SMaciej Żenczykowski FN(sk_assign), \ 4971492e639fSYonghong Song FN(ktime_get_boot_ns), \ 4972492e639fSYonghong Song FN(seq_printf), \ 4973f307fa2cSAndrey Ignatov FN(seq_write), \ 4974f307fa2cSAndrey Ignatov FN(sk_cgroup_id), \ 4975457f4436SAndrii Nakryiko FN(sk_ancestor_cgroup_id), \ 4976457f4436SAndrii Nakryiko FN(ringbuf_output), \ 4977457f4436SAndrii Nakryiko FN(ringbuf_reserve), \ 4978457f4436SAndrii Nakryiko FN(ringbuf_submit), \ 4979457f4436SAndrii Nakryiko FN(ringbuf_discard), \ 49807cdec54fSDaniel Borkmann FN(ringbuf_query), \ 4981af7ec138SYonghong Song FN(csum_level), \ 4982478cfbdfSYonghong Song FN(skc_to_tcp6_sock), \ 4983478cfbdfSYonghong Song FN(skc_to_tcp_sock), \ 4984478cfbdfSYonghong Song FN(skc_to_tcp_timewait_sock), \ 49850d4fad3eSYonghong Song FN(skc_to_tcp_request_sock), \ 4986fa28dcb8SSong Liu FN(skc_to_udp6_sock), \ 4987fa28dcb8SSong Liu FN(get_task_stack), \ 49880813a841SMartin KaFai Lau FN(load_hdr_opt), \ 49890813a841SMartin KaFai Lau FN(store_hdr_opt), \ 49908ea63684SKP Singh FN(reserve_hdr_opt), \ 49918ea63684SKP Singh FN(inode_storage_get), \ 49928ea63684SKP Singh FN(inode_storage_delete), \ 49936e22ab9dSJiri Olsa FN(d_path), \ 499407be4c4aSAlexei Starovoitov FN(copy_from_user), \ 4995c4d0bfb4SAlan Maguire FN(snprintf_btf), \ 4996eb411377SAlan Maguire FN(seq_printf_btf), \ 4997b426ce83SDaniel Borkmann FN(skb_cgroup_classid), \ 4998b4ab3141SDaniel Borkmann FN(redirect_neigh), \ 4999b7906b70SAndrii Nakryiko FN(per_cpu_ptr), \ 5000b7906b70SAndrii Nakryiko FN(this_cpu_ptr), \ 50019aa1206eSDaniel Borkmann FN(redirect_peer), \ 50024cf1bc1fSKP Singh FN(task_storage_get), \ 50034cf1bc1fSKP Singh FN(task_storage_delete), \ 50043ca1032aSKP Singh FN(get_current_task_btf), \ 50053f6719c7SKP Singh FN(bprm_opts_set), \ 5006d0551261SDmitrii Banshchikov FN(ktime_get_coarse_ns), \ 500727672f0dSKP Singh FN(ima_inode_hash), \ 50084f19cab7SFlorent Revest FN(sock_from_file), \ 500934b2021cSJesper Dangaard Brouer FN(check_mtu), \ 501069c087baSYonghong Song FN(for_each_map_elem), \ 50117b15523aSFlorent Revest FN(snprintf), \ 501279a7f8bdSAlexei Starovoitov FN(sys_bpf), \ 50133d78417bSAlexei Starovoitov FN(btf_find_by_name_kind), \ 50143abea089SAlexei Starovoitov FN(sys_close), \ 5015*b00628b1SAlexei Starovoitov FN(timer_init), \ 5016*b00628b1SAlexei Starovoitov FN(timer_set_callback), \ 5017*b00628b1SAlexei Starovoitov FN(timer_start), \ 5018*b00628b1SAlexei Starovoitov FN(timer_cancel), \ 5019fa28dcb8SSong Liu /* */ 50207a4b28c6SDaniel Borkmann 5021ebb676daSThomas Graf /* integer value in 'imm' field of BPF_CALL instruction selects which helper 5022ebb676daSThomas Graf * function eBPF program intends to call 50232d0e30c3SDaniel Borkmann */ 5024ebb676daSThomas Graf #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 5025ebb676daSThomas Graf enum bpf_func_id { 5026ebb676daSThomas Graf __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 502709756af4SAlexei Starovoitov __BPF_FUNC_MAX_ID, 502809756af4SAlexei Starovoitov }; 5029ebb676daSThomas Graf #undef __BPF_ENUM_FN 503009756af4SAlexei Starovoitov 5031781c53bcSDaniel Borkmann /* All flags used by eBPF helper functions, placed here. */ 5032781c53bcSDaniel Borkmann 5033781c53bcSDaniel Borkmann /* BPF_FUNC_skb_store_bytes flags. */ 50341aae4bddSAndrii Nakryiko enum { 50351aae4bddSAndrii Nakryiko BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 50361aae4bddSAndrii Nakryiko BPF_F_INVALIDATE_HASH = (1ULL << 1), 50371aae4bddSAndrii Nakryiko }; 5038781c53bcSDaniel Borkmann 5039781c53bcSDaniel Borkmann /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 5040781c53bcSDaniel Borkmann * First 4 bits are for passing the header field size. 5041781c53bcSDaniel Borkmann */ 50421aae4bddSAndrii Nakryiko enum { 50431aae4bddSAndrii Nakryiko BPF_F_HDR_FIELD_MASK = 0xfULL, 50441aae4bddSAndrii Nakryiko }; 5045781c53bcSDaniel Borkmann 5046781c53bcSDaniel Borkmann /* BPF_FUNC_l4_csum_replace flags. */ 50471aae4bddSAndrii Nakryiko enum { 50481aae4bddSAndrii Nakryiko BPF_F_PSEUDO_HDR = (1ULL << 4), 50491aae4bddSAndrii Nakryiko BPF_F_MARK_MANGLED_0 = (1ULL << 5), 50501aae4bddSAndrii Nakryiko BPF_F_MARK_ENFORCE = (1ULL << 6), 50511aae4bddSAndrii Nakryiko }; 5052781c53bcSDaniel Borkmann 5053781c53bcSDaniel Borkmann /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 50541aae4bddSAndrii Nakryiko enum { 50551aae4bddSAndrii Nakryiko BPF_F_INGRESS = (1ULL << 0), 50561aae4bddSAndrii Nakryiko }; 5057781c53bcSDaniel Borkmann 5058c6c33454SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 50591aae4bddSAndrii Nakryiko enum { 50601aae4bddSAndrii Nakryiko BPF_F_TUNINFO_IPV6 = (1ULL << 0), 50611aae4bddSAndrii Nakryiko }; 5062c6c33454SDaniel Borkmann 5063c195651eSYonghong Song /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 50641aae4bddSAndrii Nakryiko enum { 50651aae4bddSAndrii Nakryiko BPF_F_SKIP_FIELD_MASK = 0xffULL, 50661aae4bddSAndrii Nakryiko BPF_F_USER_STACK = (1ULL << 8), 5067c195651eSYonghong Song /* flags used by BPF_FUNC_get_stackid only. */ 50681aae4bddSAndrii Nakryiko BPF_F_FAST_STACK_CMP = (1ULL << 9), 50691aae4bddSAndrii Nakryiko BPF_F_REUSE_STACKID = (1ULL << 10), 5070c195651eSYonghong Song /* flags used by BPF_FUNC_get_stack only. */ 50711aae4bddSAndrii Nakryiko BPF_F_USER_BUILD_ID = (1ULL << 11), 50721aae4bddSAndrii Nakryiko }; 5073d5a3b1f6SAlexei Starovoitov 50742da897e5SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key flags. */ 50751aae4bddSAndrii Nakryiko enum { 50761aae4bddSAndrii Nakryiko BPF_F_ZERO_CSUM_TX = (1ULL << 1), 50771aae4bddSAndrii Nakryiko BPF_F_DONT_FRAGMENT = (1ULL << 2), 50781aae4bddSAndrii Nakryiko BPF_F_SEQ_NUMBER = (1ULL << 3), 50791aae4bddSAndrii Nakryiko }; 50802da897e5SDaniel Borkmann 5081908432caSYonghong Song /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 5082908432caSYonghong Song * BPF_FUNC_perf_event_read_value flags. 5083908432caSYonghong Song */ 50841aae4bddSAndrii Nakryiko enum { 50851aae4bddSAndrii Nakryiko BPF_F_INDEX_MASK = 0xffffffffULL, 50861aae4bddSAndrii Nakryiko BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 5087555c8a86SDaniel Borkmann /* BPF_FUNC_perf_event_output for sk_buff input context. */ 50881aae4bddSAndrii Nakryiko BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 50891aae4bddSAndrii Nakryiko }; 50901e33759cSDaniel Borkmann 5091f71c6143SJoe Stringer /* Current network namespace */ 50921aae4bddSAndrii Nakryiko enum { 50931aae4bddSAndrii Nakryiko BPF_F_CURRENT_NETNS = (-1L), 50941aae4bddSAndrii Nakryiko }; 5095f71c6143SJoe Stringer 50967cdec54fSDaniel Borkmann /* BPF_FUNC_csum_level level values. */ 50977cdec54fSDaniel Borkmann enum { 50987cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_QUERY, 50997cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_INC, 51007cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_DEC, 51017cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_RESET, 51027cdec54fSDaniel Borkmann }; 51037cdec54fSDaniel Borkmann 51042278f6ccSWillem de Bruijn /* BPF_FUNC_skb_adjust_room flags. */ 51051aae4bddSAndrii Nakryiko enum { 51061aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 51071aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 51081aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 51091aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 51101aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 5111836e66c2SDaniel Borkmann BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 5112d01b59c9SXuesen Huang BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 51131aae4bddSAndrii Nakryiko }; 51142278f6ccSWillem de Bruijn 51151aae4bddSAndrii Nakryiko enum { 51161aae4bddSAndrii Nakryiko BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 51171aae4bddSAndrii Nakryiko BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 51181aae4bddSAndrii Nakryiko }; 511958dfc900SAlan Maguire 512058dfc900SAlan Maguire #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 512158dfc900SAlan Maguire BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 512258dfc900SAlan Maguire << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 5123868d5235SWillem de Bruijn 5124808649fbSAndrey Ignatov /* BPF_FUNC_sysctl_get_name flags. */ 51251aae4bddSAndrii Nakryiko enum { 51261aae4bddSAndrii Nakryiko BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 51271aae4bddSAndrii Nakryiko }; 5128808649fbSAndrey Ignatov 5129f836a56eSKP Singh /* BPF_FUNC_<kernel_obj>_storage_get flags */ 51301aae4bddSAndrii Nakryiko enum { 5131f836a56eSKP Singh BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 5132f836a56eSKP Singh /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 5133f836a56eSKP Singh * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 5134f836a56eSKP Singh */ 5135f836a56eSKP Singh BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 51361aae4bddSAndrii Nakryiko }; 51376ac99e8fSMartin KaFai Lau 5138fff7b643SDaniel Xu /* BPF_FUNC_read_branch_records flags. */ 51391aae4bddSAndrii Nakryiko enum { 51401aae4bddSAndrii Nakryiko BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 51411aae4bddSAndrii Nakryiko }; 5142fff7b643SDaniel Xu 5143457f4436SAndrii Nakryiko /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 5144457f4436SAndrii Nakryiko * BPF_FUNC_bpf_ringbuf_output flags. 5145457f4436SAndrii Nakryiko */ 5146457f4436SAndrii Nakryiko enum { 5147457f4436SAndrii Nakryiko BPF_RB_NO_WAKEUP = (1ULL << 0), 5148457f4436SAndrii Nakryiko BPF_RB_FORCE_WAKEUP = (1ULL << 1), 5149457f4436SAndrii Nakryiko }; 5150457f4436SAndrii Nakryiko 5151457f4436SAndrii Nakryiko /* BPF_FUNC_bpf_ringbuf_query flags */ 5152457f4436SAndrii Nakryiko enum { 5153457f4436SAndrii Nakryiko BPF_RB_AVAIL_DATA = 0, 5154457f4436SAndrii Nakryiko BPF_RB_RING_SIZE = 1, 5155457f4436SAndrii Nakryiko BPF_RB_CONS_POS = 2, 5156457f4436SAndrii Nakryiko BPF_RB_PROD_POS = 3, 5157457f4436SAndrii Nakryiko }; 5158457f4436SAndrii Nakryiko 5159457f4436SAndrii Nakryiko /* BPF ring buffer constants */ 5160457f4436SAndrii Nakryiko enum { 5161457f4436SAndrii Nakryiko BPF_RINGBUF_BUSY_BIT = (1U << 31), 5162457f4436SAndrii Nakryiko BPF_RINGBUF_DISCARD_BIT = (1U << 30), 5163457f4436SAndrii Nakryiko BPF_RINGBUF_HDR_SZ = 8, 5164457f4436SAndrii Nakryiko }; 5165457f4436SAndrii Nakryiko 5166e9ddbb77SJakub Sitnicki /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 5167e9ddbb77SJakub Sitnicki enum { 5168e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 5169e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 5170e9ddbb77SJakub Sitnicki }; 5171e9ddbb77SJakub Sitnicki 51722be7e212SDaniel Borkmann /* Mode for BPF_FUNC_skb_adjust_room helper. */ 51732be7e212SDaniel Borkmann enum bpf_adj_room_mode { 51742be7e212SDaniel Borkmann BPF_ADJ_ROOM_NET, 517514aa3192SWillem de Bruijn BPF_ADJ_ROOM_MAC, 51762be7e212SDaniel Borkmann }; 51772be7e212SDaniel Borkmann 51784e1ec56cSDaniel Borkmann /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 51794e1ec56cSDaniel Borkmann enum bpf_hdr_start_off { 51804e1ec56cSDaniel Borkmann BPF_HDR_START_MAC, 51814e1ec56cSDaniel Borkmann BPF_HDR_START_NET, 51824e1ec56cSDaniel Borkmann }; 51834e1ec56cSDaniel Borkmann 5184fe94cc29SMathieu Xhonneux /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 5185fe94cc29SMathieu Xhonneux enum bpf_lwt_encap_mode { 5186fe94cc29SMathieu Xhonneux BPF_LWT_ENCAP_SEG6, 51873e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_SEG6_INLINE, 51883e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_IP, 5189fe94cc29SMathieu Xhonneux }; 5190fe94cc29SMathieu Xhonneux 51913f6719c7SKP Singh /* Flags for bpf_bprm_opts_set helper */ 51923f6719c7SKP Singh enum { 51933f6719c7SKP Singh BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 51943f6719c7SKP Singh }; 51953f6719c7SKP Singh 5196e624d4edSHangbin Liu /* Flags for bpf_redirect_map helper */ 5197e624d4edSHangbin Liu enum { 5198e624d4edSHangbin Liu BPF_F_BROADCAST = (1ULL << 3), 5199e624d4edSHangbin Liu BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 5200e624d4edSHangbin Liu }; 5201e624d4edSHangbin Liu 5202b7df9adaSDaniel Borkmann #define __bpf_md_ptr(type, name) \ 5203b7df9adaSDaniel Borkmann union { \ 5204b7df9adaSDaniel Borkmann type name; \ 5205b7df9adaSDaniel Borkmann __u64 :64; \ 5206b7df9adaSDaniel Borkmann } __attribute__((aligned(8))) 5207b7df9adaSDaniel Borkmann 52089bac3d6dSAlexei Starovoitov /* user accessible mirror of in-kernel sk_buff. 52099bac3d6dSAlexei Starovoitov * new fields can only be added to the end of this structure 52109bac3d6dSAlexei Starovoitov */ 52119bac3d6dSAlexei Starovoitov struct __sk_buff { 52129bac3d6dSAlexei Starovoitov __u32 len; 52139bac3d6dSAlexei Starovoitov __u32 pkt_type; 52149bac3d6dSAlexei Starovoitov __u32 mark; 52159bac3d6dSAlexei Starovoitov __u32 queue_mapping; 5216c2497395SAlexei Starovoitov __u32 protocol; 5217c2497395SAlexei Starovoitov __u32 vlan_present; 5218c2497395SAlexei Starovoitov __u32 vlan_tci; 521927cd5452SMichal Sekletar __u32 vlan_proto; 5220bcad5718SDaniel Borkmann __u32 priority; 522137e82c2fSAlexei Starovoitov __u32 ingress_ifindex; 522237e82c2fSAlexei Starovoitov __u32 ifindex; 5223d691f9e8SAlexei Starovoitov __u32 tc_index; 5224d691f9e8SAlexei Starovoitov __u32 cb[5]; 5225ba7591d8SDaniel Borkmann __u32 hash; 5226045efa82SDaniel Borkmann __u32 tc_classid; 5227969bf05eSAlexei Starovoitov __u32 data; 5228969bf05eSAlexei Starovoitov __u32 data_end; 5229b1d9fc41SDaniel Borkmann __u32 napi_id; 52308a31db56SJohn Fastabend 5231de8f3a83SDaniel Borkmann /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 52328a31db56SJohn Fastabend __u32 family; 52338a31db56SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 52348a31db56SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 52358a31db56SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 52368a31db56SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 52378a31db56SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 52388a31db56SJohn Fastabend __u32 local_port; /* stored in host byte order */ 5239de8f3a83SDaniel Borkmann /* ... here. */ 5240de8f3a83SDaniel Borkmann 5241de8f3a83SDaniel Borkmann __u32 data_meta; 5242b7df9adaSDaniel Borkmann __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 5243f11216b2SVlad Dumitrescu __u64 tstamp; 5244e3da08d0SPetar Penkov __u32 wire_len; 5245d9ff286aSEric Dumazet __u32 gso_segs; 524646f8bc92SMartin KaFai Lau __bpf_md_ptr(struct bpf_sock *, sk); 5247cf62089bSWillem de Bruijn __u32 gso_size; 52489bac3d6dSAlexei Starovoitov }; 52499bac3d6dSAlexei Starovoitov 5250d3aa45ceSAlexei Starovoitov struct bpf_tunnel_key { 5251d3aa45ceSAlexei Starovoitov __u32 tunnel_id; 5252c6c33454SDaniel Borkmann union { 5253d3aa45ceSAlexei Starovoitov __u32 remote_ipv4; 5254c6c33454SDaniel Borkmann __u32 remote_ipv6[4]; 5255c6c33454SDaniel Borkmann }; 5256c6c33454SDaniel Borkmann __u8 tunnel_tos; 5257c6c33454SDaniel Borkmann __u8 tunnel_ttl; 52581fbc2e0cSDaniel Borkmann __u16 tunnel_ext; /* Padding, future use. */ 52594018ab18SDaniel Borkmann __u32 tunnel_label; 5260d3aa45ceSAlexei Starovoitov }; 5261d3aa45ceSAlexei Starovoitov 526212bed760SEyal Birger /* user accessible mirror of in-kernel xfrm_state. 526312bed760SEyal Birger * new fields can only be added to the end of this structure 526412bed760SEyal Birger */ 526512bed760SEyal Birger struct bpf_xfrm_state { 526612bed760SEyal Birger __u32 reqid; 526712bed760SEyal Birger __u32 spi; /* Stored in network byte order */ 526812bed760SEyal Birger __u16 family; 52691fbc2e0cSDaniel Borkmann __u16 ext; /* Padding, future use. */ 527012bed760SEyal Birger union { 527112bed760SEyal Birger __u32 remote_ipv4; /* Stored in network byte order */ 527212bed760SEyal Birger __u32 remote_ipv6[4]; /* Stored in network byte order */ 527312bed760SEyal Birger }; 527412bed760SEyal Birger }; 527512bed760SEyal Birger 52763a0af8fdSThomas Graf /* Generic BPF return codes which all BPF program types may support. 52773a0af8fdSThomas Graf * The values are binary compatible with their TC_ACT_* counter-part to 52783a0af8fdSThomas Graf * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 52793a0af8fdSThomas Graf * programs. 52803a0af8fdSThomas Graf * 52813a0af8fdSThomas Graf * XDP is handled seprately, see XDP_*. 52823a0af8fdSThomas Graf */ 52833a0af8fdSThomas Graf enum bpf_ret_code { 52843a0af8fdSThomas Graf BPF_OK = 0, 52853a0af8fdSThomas Graf /* 1 reserved */ 52863a0af8fdSThomas Graf BPF_DROP = 2, 52873a0af8fdSThomas Graf /* 3-6 reserved */ 52883a0af8fdSThomas Graf BPF_REDIRECT = 7, 52893e0bd37cSPeter Oskolkov /* >127 are reserved for prog type specific return codes. 52903e0bd37cSPeter Oskolkov * 52913e0bd37cSPeter Oskolkov * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 52923e0bd37cSPeter Oskolkov * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 52933e0bd37cSPeter Oskolkov * changed and should be routed based on its new L3 header. 52943e0bd37cSPeter Oskolkov * (This is an L3 redirect, as opposed to L2 redirect 52953e0bd37cSPeter Oskolkov * represented by BPF_REDIRECT above). 52963e0bd37cSPeter Oskolkov */ 52973e0bd37cSPeter Oskolkov BPF_LWT_REROUTE = 128, 52983a0af8fdSThomas Graf }; 52993a0af8fdSThomas Graf 530061023658SDavid Ahern struct bpf_sock { 530161023658SDavid Ahern __u32 bound_dev_if; 5302aa4c1037SDavid Ahern __u32 family; 5303aa4c1037SDavid Ahern __u32 type; 5304aa4c1037SDavid Ahern __u32 protocol; 5305482dca93SDavid Ahern __u32 mark; 5306482dca93SDavid Ahern __u32 priority; 5307aa65d696SMartin KaFai Lau /* IP address also allows 1 and 2 bytes access */ 5308aa65d696SMartin KaFai Lau __u32 src_ip4; 5309aa65d696SMartin KaFai Lau __u32 src_ip6[4]; 5310aa65d696SMartin KaFai Lau __u32 src_port; /* host byte order */ 5311aa65d696SMartin KaFai Lau __u32 dst_port; /* network byte order */ 5312aa65d696SMartin KaFai Lau __u32 dst_ip4; 5313aa65d696SMartin KaFai Lau __u32 dst_ip6[4]; 5314aa65d696SMartin KaFai Lau __u32 state; 5315c3c16f2eSAmritha Nambiar __s32 rx_queue_mapping; 531661023658SDavid Ahern }; 531761023658SDavid Ahern 5318655a51e5SMartin KaFai Lau struct bpf_tcp_sock { 5319655a51e5SMartin KaFai Lau __u32 snd_cwnd; /* Sending congestion window */ 5320655a51e5SMartin KaFai Lau __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 5321655a51e5SMartin KaFai Lau __u32 rtt_min; 5322655a51e5SMartin KaFai Lau __u32 snd_ssthresh; /* Slow start size threshold */ 5323655a51e5SMartin KaFai Lau __u32 rcv_nxt; /* What we want to receive next */ 5324655a51e5SMartin KaFai Lau __u32 snd_nxt; /* Next sequence we send */ 5325655a51e5SMartin KaFai Lau __u32 snd_una; /* First byte we want an ack for */ 5326655a51e5SMartin KaFai Lau __u32 mss_cache; /* Cached effective mss, not including SACKS */ 5327655a51e5SMartin KaFai Lau __u32 ecn_flags; /* ECN status bits. */ 5328655a51e5SMartin KaFai Lau __u32 rate_delivered; /* saved rate sample: packets delivered */ 5329655a51e5SMartin KaFai Lau __u32 rate_interval_us; /* saved rate sample: time elapsed */ 5330655a51e5SMartin KaFai Lau __u32 packets_out; /* Packets which are "in flight" */ 5331655a51e5SMartin KaFai Lau __u32 retrans_out; /* Retransmitted packets out */ 5332655a51e5SMartin KaFai Lau __u32 total_retrans; /* Total retransmits for entire connection */ 5333655a51e5SMartin KaFai Lau __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 5334655a51e5SMartin KaFai Lau * total number of segments in. 5335655a51e5SMartin KaFai Lau */ 5336655a51e5SMartin KaFai Lau __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 5337655a51e5SMartin KaFai Lau * total number of data segments in. 5338655a51e5SMartin KaFai Lau */ 5339655a51e5SMartin KaFai Lau __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 5340655a51e5SMartin KaFai Lau * The total number of segments sent. 5341655a51e5SMartin KaFai Lau */ 5342655a51e5SMartin KaFai Lau __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 5343655a51e5SMartin KaFai Lau * total number of data segments sent. 5344655a51e5SMartin KaFai Lau */ 5345655a51e5SMartin KaFai Lau __u32 lost_out; /* Lost packets */ 5346655a51e5SMartin KaFai Lau __u32 sacked_out; /* SACK'd packets */ 5347655a51e5SMartin KaFai Lau __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 5348655a51e5SMartin KaFai Lau * sum(delta(rcv_nxt)), or how many bytes 5349655a51e5SMartin KaFai Lau * were acked. 5350655a51e5SMartin KaFai Lau */ 5351655a51e5SMartin KaFai Lau __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 5352655a51e5SMartin KaFai Lau * sum(delta(snd_una)), or how many bytes 5353655a51e5SMartin KaFai Lau * were acked. 5354655a51e5SMartin KaFai Lau */ 53550357746dSStanislav Fomichev __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 53560357746dSStanislav Fomichev * total number of DSACK blocks received 53570357746dSStanislav Fomichev */ 53580357746dSStanislav Fomichev __u32 delivered; /* Total data packets delivered incl. rexmits */ 53590357746dSStanislav Fomichev __u32 delivered_ce; /* Like the above but only ECE marked packets */ 5360c2cb5e82SStanislav Fomichev __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 5361655a51e5SMartin KaFai Lau }; 5362655a51e5SMartin KaFai Lau 53636acc9b43SJoe Stringer struct bpf_sock_tuple { 53646acc9b43SJoe Stringer union { 53656acc9b43SJoe Stringer struct { 53666acc9b43SJoe Stringer __be32 saddr; 53676acc9b43SJoe Stringer __be32 daddr; 53686acc9b43SJoe Stringer __be16 sport; 53696acc9b43SJoe Stringer __be16 dport; 53706acc9b43SJoe Stringer } ipv4; 53716acc9b43SJoe Stringer struct { 53726acc9b43SJoe Stringer __be32 saddr[4]; 53736acc9b43SJoe Stringer __be32 daddr[4]; 53746acc9b43SJoe Stringer __be16 sport; 53756acc9b43SJoe Stringer __be16 dport; 53766acc9b43SJoe Stringer } ipv6; 53776acc9b43SJoe Stringer }; 53786acc9b43SJoe Stringer }; 53796acc9b43SJoe Stringer 5380fada7fdcSJonathan Lemon struct bpf_xdp_sock { 5381fada7fdcSJonathan Lemon __u32 queue_id; 5382fada7fdcSJonathan Lemon }; 5383fada7fdcSJonathan Lemon 538417bedab2SMartin KaFai Lau #define XDP_PACKET_HEADROOM 256 538517bedab2SMartin KaFai Lau 53866a773a15SBrenden Blanco /* User return codes for XDP prog type. 53876a773a15SBrenden Blanco * A valid XDP program must return one of these defined values. All other 53889beb8bedSDaniel Borkmann * return codes are reserved for future use. Unknown return codes will 53899beb8bedSDaniel Borkmann * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 53906a773a15SBrenden Blanco */ 53916a773a15SBrenden Blanco enum xdp_action { 53926a773a15SBrenden Blanco XDP_ABORTED = 0, 53936a773a15SBrenden Blanco XDP_DROP, 53946a773a15SBrenden Blanco XDP_PASS, 53956ce96ca3SBrenden Blanco XDP_TX, 5396814abfabSJohn Fastabend XDP_REDIRECT, 53976a773a15SBrenden Blanco }; 53986a773a15SBrenden Blanco 53996a773a15SBrenden Blanco /* user accessible metadata for XDP packet hook 54006a773a15SBrenden Blanco * new fields must be added to the end of this structure 54016a773a15SBrenden Blanco */ 54026a773a15SBrenden Blanco struct xdp_md { 54036a773a15SBrenden Blanco __u32 data; 54046a773a15SBrenden Blanco __u32 data_end; 5405de8f3a83SDaniel Borkmann __u32 data_meta; 5406daaf24c6SJesper Dangaard Brouer /* Below access go through struct xdp_rxq_info */ 540702dd3291SJesper Dangaard Brouer __u32 ingress_ifindex; /* rxq->dev->ifindex */ 540802dd3291SJesper Dangaard Brouer __u32 rx_queue_index; /* rxq->queue_index */ 540964b59025SDavid Ahern 541064b59025SDavid Ahern __u32 egress_ifindex; /* txq->dev->ifindex */ 54116a773a15SBrenden Blanco }; 54126a773a15SBrenden Blanco 5413281920b7SJesper Dangaard Brouer /* DEVMAP map-value layout 5414281920b7SJesper Dangaard Brouer * 5415281920b7SJesper Dangaard Brouer * The struct data-layout of map-value is a configuration interface. 5416281920b7SJesper Dangaard Brouer * New members can only be added to the end of this structure. 5417281920b7SJesper Dangaard Brouer */ 5418281920b7SJesper Dangaard Brouer struct bpf_devmap_val { 5419281920b7SJesper Dangaard Brouer __u32 ifindex; /* device index */ 5420281920b7SJesper Dangaard Brouer union { 5421281920b7SJesper Dangaard Brouer int fd; /* prog fd on map write */ 5422281920b7SJesper Dangaard Brouer __u32 id; /* prog id on map read */ 5423281920b7SJesper Dangaard Brouer } bpf_prog; 5424281920b7SJesper Dangaard Brouer }; 5425281920b7SJesper Dangaard Brouer 5426644bfe51SLorenzo Bianconi /* CPUMAP map-value layout 5427644bfe51SLorenzo Bianconi * 5428644bfe51SLorenzo Bianconi * The struct data-layout of map-value is a configuration interface. 5429644bfe51SLorenzo Bianconi * New members can only be added to the end of this structure. 5430644bfe51SLorenzo Bianconi */ 5431644bfe51SLorenzo Bianconi struct bpf_cpumap_val { 5432644bfe51SLorenzo Bianconi __u32 qsize; /* queue size to remote target CPU */ 543392164774SLorenzo Bianconi union { 543492164774SLorenzo Bianconi int fd; /* prog fd on map write */ 543592164774SLorenzo Bianconi __u32 id; /* prog id on map read */ 543692164774SLorenzo Bianconi } bpf_prog; 5437644bfe51SLorenzo Bianconi }; 5438644bfe51SLorenzo Bianconi 5439174a79ffSJohn Fastabend enum sk_action { 5440bfa64075SJohn Fastabend SK_DROP = 0, 5441bfa64075SJohn Fastabend SK_PASS, 5442174a79ffSJohn Fastabend }; 5443174a79ffSJohn Fastabend 54444f738adbSJohn Fastabend /* user accessible metadata for SK_MSG packet hook, new fields must 54454f738adbSJohn Fastabend * be added to the end of this structure 54464f738adbSJohn Fastabend */ 54474f738adbSJohn Fastabend struct sk_msg_md { 5448b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 5449b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 5450303def35SJohn Fastabend 5451303def35SJohn Fastabend __u32 family; 5452303def35SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 5453303def35SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 5454303def35SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 5455303def35SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 5456303def35SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 5457303def35SJohn Fastabend __u32 local_port; /* stored in host byte order */ 54583bdbd022SJohn Fastabend __u32 size; /* Total size of sk_msg */ 545913d70f5aSJohn Fastabend 546013d70f5aSJohn Fastabend __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 54614f738adbSJohn Fastabend }; 54624f738adbSJohn Fastabend 54632dbb9b9eSMartin KaFai Lau struct sk_reuseport_md { 54642dbb9b9eSMartin KaFai Lau /* 54652dbb9b9eSMartin KaFai Lau * Start of directly accessible data. It begins from 54662dbb9b9eSMartin KaFai Lau * the tcp/udp header. 54672dbb9b9eSMartin KaFai Lau */ 5468b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 5469b7df9adaSDaniel Borkmann /* End of directly accessible data */ 5470b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 54712dbb9b9eSMartin KaFai Lau /* 54722dbb9b9eSMartin KaFai Lau * Total length of packet (starting from the tcp/udp header). 54732dbb9b9eSMartin KaFai Lau * Note that the directly accessible bytes (data_end - data) 54742dbb9b9eSMartin KaFai Lau * could be less than this "len". Those bytes could be 54752dbb9b9eSMartin KaFai Lau * indirectly read by a helper "bpf_skb_load_bytes()". 54762dbb9b9eSMartin KaFai Lau */ 54772dbb9b9eSMartin KaFai Lau __u32 len; 54782dbb9b9eSMartin KaFai Lau /* 54792dbb9b9eSMartin KaFai Lau * Eth protocol in the mac header (network byte order). e.g. 54802dbb9b9eSMartin KaFai Lau * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 54812dbb9b9eSMartin KaFai Lau */ 54822dbb9b9eSMartin KaFai Lau __u32 eth_protocol; 54832dbb9b9eSMartin KaFai Lau __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 54842dbb9b9eSMartin KaFai Lau __u32 bind_inany; /* Is sock bound to an INANY address? */ 54852dbb9b9eSMartin KaFai Lau __u32 hash; /* A hash of the packet 4 tuples */ 5486d5e4ddaeSKuniyuki Iwashima /* When reuse->migrating_sk is NULL, it is selecting a sk for the 5487d5e4ddaeSKuniyuki Iwashima * new incoming connection request (e.g. selecting a listen sk for 5488d5e4ddaeSKuniyuki Iwashima * the received SYN in the TCP case). reuse->sk is one of the sk 5489d5e4ddaeSKuniyuki Iwashima * in the reuseport group. The bpf prog can use reuse->sk to learn 5490d5e4ddaeSKuniyuki Iwashima * the local listening ip/port without looking into the skb. 5491d5e4ddaeSKuniyuki Iwashima * 5492d5e4ddaeSKuniyuki Iwashima * When reuse->migrating_sk is not NULL, reuse->sk is closed and 5493d5e4ddaeSKuniyuki Iwashima * reuse->migrating_sk is the socket that needs to be migrated 5494d5e4ddaeSKuniyuki Iwashima * to another listening socket. migrating_sk could be a fullsock 5495d5e4ddaeSKuniyuki Iwashima * sk that is fully established or a reqsk that is in-the-middle 5496d5e4ddaeSKuniyuki Iwashima * of 3-way handshake. 5497d5e4ddaeSKuniyuki Iwashima */ 5498e0610476SKuniyuki Iwashima __bpf_md_ptr(struct bpf_sock *, sk); 5499d5e4ddaeSKuniyuki Iwashima __bpf_md_ptr(struct bpf_sock *, migrating_sk); 55002dbb9b9eSMartin KaFai Lau }; 55012dbb9b9eSMartin KaFai Lau 55021e270976SMartin KaFai Lau #define BPF_TAG_SIZE 8 55031e270976SMartin KaFai Lau 55041e270976SMartin KaFai Lau struct bpf_prog_info { 55051e270976SMartin KaFai Lau __u32 type; 55061e270976SMartin KaFai Lau __u32 id; 55071e270976SMartin KaFai Lau __u8 tag[BPF_TAG_SIZE]; 55081e270976SMartin KaFai Lau __u32 jited_prog_len; 55091e270976SMartin KaFai Lau __u32 xlated_prog_len; 55101e270976SMartin KaFai Lau __aligned_u64 jited_prog_insns; 55111e270976SMartin KaFai Lau __aligned_u64 xlated_prog_insns; 5512cb4d2b3fSMartin KaFai Lau __u64 load_time; /* ns since boottime */ 5513cb4d2b3fSMartin KaFai Lau __u32 created_by_uid; 5514cb4d2b3fSMartin KaFai Lau __u32 nr_map_ids; 5515cb4d2b3fSMartin KaFai Lau __aligned_u64 map_ids; 5516067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 5517675fc275SJakub Kicinski __u32 ifindex; 5518b85fab0eSJiri Olsa __u32 gpl_compatible:1; 55190472301aSBaruch Siach __u32 :31; /* alignment pad */ 5520675fc275SJakub Kicinski __u64 netns_dev; 5521675fc275SJakub Kicinski __u64 netns_ino; 5522dbecd738SSandipan Das __u32 nr_jited_ksyms; 5523815581c1SSandipan Das __u32 nr_jited_func_lens; 5524dbecd738SSandipan Das __aligned_u64 jited_ksyms; 5525815581c1SSandipan Das __aligned_u64 jited_func_lens; 5526838e9690SYonghong Song __u32 btf_id; 5527838e9690SYonghong Song __u32 func_info_rec_size; 5528838e9690SYonghong Song __aligned_u64 func_info; 552911d8b82dSYonghong Song __u32 nr_func_info; 553011d8b82dSYonghong Song __u32 nr_line_info; 5531c454a46bSMartin KaFai Lau __aligned_u64 line_info; 5532c454a46bSMartin KaFai Lau __aligned_u64 jited_line_info; 553311d8b82dSYonghong Song __u32 nr_jited_line_info; 5534c454a46bSMartin KaFai Lau __u32 line_info_rec_size; 5535c454a46bSMartin KaFai Lau __u32 jited_line_info_rec_size; 5536c872bdb3SSong Liu __u32 nr_prog_tags; 5537c872bdb3SSong Liu __aligned_u64 prog_tags; 55385f8f8b93SAlexei Starovoitov __u64 run_time_ns; 55395f8f8b93SAlexei Starovoitov __u64 run_cnt; 55409ed9e9baSAlexei Starovoitov __u64 recursion_misses; 55411e270976SMartin KaFai Lau } __attribute__((aligned(8))); 55421e270976SMartin KaFai Lau 55431e270976SMartin KaFai Lau struct bpf_map_info { 55441e270976SMartin KaFai Lau __u32 type; 55451e270976SMartin KaFai Lau __u32 id; 55461e270976SMartin KaFai Lau __u32 key_size; 55471e270976SMartin KaFai Lau __u32 value_size; 55481e270976SMartin KaFai Lau __u32 max_entries; 55491e270976SMartin KaFai Lau __u32 map_flags; 5550067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 555152775b33SJakub Kicinski __u32 ifindex; 555285d33df3SMartin KaFai Lau __u32 btf_vmlinux_value_type_id; 555352775b33SJakub Kicinski __u64 netns_dev; 555452775b33SJakub Kicinski __u64 netns_ino; 555578958fcaSMartin KaFai Lau __u32 btf_id; 55569b2cf328SMartin KaFai Lau __u32 btf_key_type_id; 55579b2cf328SMartin KaFai Lau __u32 btf_value_type_id; 55581e270976SMartin KaFai Lau } __attribute__((aligned(8))); 55591e270976SMartin KaFai Lau 556062dab84cSMartin KaFai Lau struct bpf_btf_info { 556162dab84cSMartin KaFai Lau __aligned_u64 btf; 556262dab84cSMartin KaFai Lau __u32 btf_size; 556362dab84cSMartin KaFai Lau __u32 id; 556453297220SAndrii Nakryiko __aligned_u64 name; 556553297220SAndrii Nakryiko __u32 name_len; 556653297220SAndrii Nakryiko __u32 kernel_btf; 556762dab84cSMartin KaFai Lau } __attribute__((aligned(8))); 556862dab84cSMartin KaFai Lau 5569f2e10bffSAndrii Nakryiko struct bpf_link_info { 5570f2e10bffSAndrii Nakryiko __u32 type; 5571f2e10bffSAndrii Nakryiko __u32 id; 5572f2e10bffSAndrii Nakryiko __u32 prog_id; 5573f2e10bffSAndrii Nakryiko union { 5574f2e10bffSAndrii Nakryiko struct { 5575f2e10bffSAndrii Nakryiko __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 5576f2e10bffSAndrii Nakryiko __u32 tp_name_len; /* in/out: tp_name buffer len */ 5577f2e10bffSAndrii Nakryiko } raw_tracepoint; 5578f2e10bffSAndrii Nakryiko struct { 5579f2e10bffSAndrii Nakryiko __u32 attach_type; 5580441e8c66SToke Høiland-Jørgensen __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 5581441e8c66SToke Høiland-Jørgensen __u32 target_btf_id; /* BTF type id inside the object */ 5582f2e10bffSAndrii Nakryiko } tracing; 5583f2e10bffSAndrii Nakryiko struct { 5584f2e10bffSAndrii Nakryiko __u64 cgroup_id; 5585f2e10bffSAndrii Nakryiko __u32 attach_type; 5586f2e10bffSAndrii Nakryiko } cgroup; 55877f045a49SJakub Sitnicki struct { 55886b0a249aSYonghong Song __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 55896b0a249aSYonghong Song __u32 target_name_len; /* in/out: target_name buffer len */ 55906b0a249aSYonghong Song union { 5591b0c9eb37SYonghong Song struct { 55926b0a249aSYonghong Song __u32 map_id; 55936b0a249aSYonghong Song } map; 5594b0c9eb37SYonghong Song }; 55956b0a249aSYonghong Song } iter; 55966b0a249aSYonghong Song struct { 55977f045a49SJakub Sitnicki __u32 netns_ino; 55987f045a49SJakub Sitnicki __u32 attach_type; 55997f045a49SJakub Sitnicki } netns; 5600c1931c97SAndrii Nakryiko struct { 5601c1931c97SAndrii Nakryiko __u32 ifindex; 5602c1931c97SAndrii Nakryiko } xdp; 5603f2e10bffSAndrii Nakryiko }; 5604f2e10bffSAndrii Nakryiko } __attribute__((aligned(8))); 5605f2e10bffSAndrii Nakryiko 56064fbac77dSAndrey Ignatov /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 56074fbac77dSAndrey Ignatov * by user and intended to be used by socket (e.g. to bind to, depends on 5608bfdfa517SRandy Dunlap * attach type). 56094fbac77dSAndrey Ignatov */ 56104fbac77dSAndrey Ignatov struct bpf_sock_addr { 56114fbac77dSAndrey Ignatov __u32 user_family; /* Allows 4-byte read, but no write. */ 56124fbac77dSAndrey Ignatov __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 56134fbac77dSAndrey Ignatov * Stored in network byte order. 56144fbac77dSAndrey Ignatov */ 5615d4ecfeb1SStanislav Fomichev __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 56164fbac77dSAndrey Ignatov * Stored in network byte order. 56174fbac77dSAndrey Ignatov */ 56187aebfa1bSAndrey Ignatov __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 56194fbac77dSAndrey Ignatov * Stored in network byte order 56204fbac77dSAndrey Ignatov */ 56214fbac77dSAndrey Ignatov __u32 family; /* Allows 4-byte read, but no write */ 56224fbac77dSAndrey Ignatov __u32 type; /* Allows 4-byte read, but no write */ 56234fbac77dSAndrey Ignatov __u32 protocol; /* Allows 4-byte read, but no write */ 5624600c70baSStanislav Fomichev __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 56251cedee13SAndrey Ignatov * Stored in network byte order. 56261cedee13SAndrey Ignatov */ 5627d4ecfeb1SStanislav Fomichev __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 56281cedee13SAndrey Ignatov * Stored in network byte order. 56291cedee13SAndrey Ignatov */ 5630fb85c4a7SStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 56314fbac77dSAndrey Ignatov }; 56324fbac77dSAndrey Ignatov 563340304b2aSLawrence Brakmo /* User bpf_sock_ops struct to access socket values and specify request ops 563440304b2aSLawrence Brakmo * and their replies. 563540304b2aSLawrence Brakmo * Some of this fields are in network (bigendian) byte order and may need 563640304b2aSLawrence Brakmo * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 563740304b2aSLawrence Brakmo * New fields can only be added at the end of this structure 563840304b2aSLawrence Brakmo */ 563940304b2aSLawrence Brakmo struct bpf_sock_ops { 564040304b2aSLawrence Brakmo __u32 op; 564140304b2aSLawrence Brakmo union { 5642de525be2SLawrence Brakmo __u32 args[4]; /* Optionally passed to bpf program */ 5643de525be2SLawrence Brakmo __u32 reply; /* Returned by bpf program */ 5644de525be2SLawrence Brakmo __u32 replylong[4]; /* Optionally returned by bpf prog */ 564540304b2aSLawrence Brakmo }; 564640304b2aSLawrence Brakmo __u32 family; 564740304b2aSLawrence Brakmo __u32 remote_ip4; /* Stored in network byte order */ 564840304b2aSLawrence Brakmo __u32 local_ip4; /* Stored in network byte order */ 564940304b2aSLawrence Brakmo __u32 remote_ip6[4]; /* Stored in network byte order */ 565040304b2aSLawrence Brakmo __u32 local_ip6[4]; /* Stored in network byte order */ 565140304b2aSLawrence Brakmo __u32 remote_port; /* Stored in network byte order */ 565240304b2aSLawrence Brakmo __u32 local_port; /* stored in host byte order */ 5653f19397a5SLawrence Brakmo __u32 is_fullsock; /* Some TCP fields are only valid if 5654f19397a5SLawrence Brakmo * there is a full socket. If not, the 5655f19397a5SLawrence Brakmo * fields read as zero. 5656f19397a5SLawrence Brakmo */ 5657f19397a5SLawrence Brakmo __u32 snd_cwnd; 5658f19397a5SLawrence Brakmo __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 5659b13d8807SLawrence Brakmo __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 566044f0e430SLawrence Brakmo __u32 state; 566144f0e430SLawrence Brakmo __u32 rtt_min; 566244f0e430SLawrence Brakmo __u32 snd_ssthresh; 566344f0e430SLawrence Brakmo __u32 rcv_nxt; 566444f0e430SLawrence Brakmo __u32 snd_nxt; 566544f0e430SLawrence Brakmo __u32 snd_una; 566644f0e430SLawrence Brakmo __u32 mss_cache; 566744f0e430SLawrence Brakmo __u32 ecn_flags; 566844f0e430SLawrence Brakmo __u32 rate_delivered; 566944f0e430SLawrence Brakmo __u32 rate_interval_us; 567044f0e430SLawrence Brakmo __u32 packets_out; 567144f0e430SLawrence Brakmo __u32 retrans_out; 567244f0e430SLawrence Brakmo __u32 total_retrans; 567344f0e430SLawrence Brakmo __u32 segs_in; 567444f0e430SLawrence Brakmo __u32 data_segs_in; 567544f0e430SLawrence Brakmo __u32 segs_out; 567644f0e430SLawrence Brakmo __u32 data_segs_out; 567744f0e430SLawrence Brakmo __u32 lost_out; 567844f0e430SLawrence Brakmo __u32 sacked_out; 567944f0e430SLawrence Brakmo __u32 sk_txhash; 568044f0e430SLawrence Brakmo __u64 bytes_received; 568144f0e430SLawrence Brakmo __u64 bytes_acked; 56821314ef56SStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 56830813a841SMartin KaFai Lau /* [skb_data, skb_data_end) covers the whole TCP header. 56840813a841SMartin KaFai Lau * 56850813a841SMartin KaFai Lau * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 56860813a841SMartin KaFai Lau * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 56870813a841SMartin KaFai Lau * header has not been written. 56880813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 56890813a841SMartin KaFai Lau * been written so far. 56900813a841SMartin KaFai Lau * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 56910813a841SMartin KaFai Lau * the 3WHS. 56920813a841SMartin KaFai Lau * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 56930813a841SMartin KaFai Lau * the 3WHS. 56940813a841SMartin KaFai Lau * 56950813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also be used to read a particular option. 56960813a841SMartin KaFai Lau */ 56970813a841SMartin KaFai Lau __bpf_md_ptr(void *, skb_data); 56980813a841SMartin KaFai Lau __bpf_md_ptr(void *, skb_data_end); 56990813a841SMartin KaFai Lau __u32 skb_len; /* The total length of a packet. 57000813a841SMartin KaFai Lau * It includes the header, options, 57010813a841SMartin KaFai Lau * and payload. 57020813a841SMartin KaFai Lau */ 57030813a841SMartin KaFai Lau __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 57040813a841SMartin KaFai Lau * an easy way to check for tcp_flags 57050813a841SMartin KaFai Lau * without parsing skb_data. 57060813a841SMartin KaFai Lau * 57070813a841SMartin KaFai Lau * In particular, the skb_tcp_flags 57080813a841SMartin KaFai Lau * will still be available in 57090813a841SMartin KaFai Lau * BPF_SOCK_OPS_HDR_OPT_LEN even though 57100813a841SMartin KaFai Lau * the outgoing header has not 57110813a841SMartin KaFai Lau * been written yet. 57120813a841SMartin KaFai Lau */ 571340304b2aSLawrence Brakmo }; 571440304b2aSLawrence Brakmo 5715b13d8807SLawrence Brakmo /* Definitions for bpf_sock_ops_cb_flags */ 57161aae4bddSAndrii Nakryiko enum { 57171aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 57181aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 57191aae4bddSAndrii Nakryiko BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 57201aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 57210813a841SMartin KaFai Lau /* Call bpf for all received TCP headers. The bpf prog will be 57220813a841SMartin KaFai Lau * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 57230813a841SMartin KaFai Lau * 57240813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 57250813a841SMartin KaFai Lau * for the header option related helpers that will be useful 57260813a841SMartin KaFai Lau * to the bpf programs. 57270813a841SMartin KaFai Lau * 57280813a841SMartin KaFai Lau * It could be used at the client/active side (i.e. connect() side) 57290813a841SMartin KaFai Lau * when the server told it that the server was in syncookie 57300813a841SMartin KaFai Lau * mode and required the active side to resend the bpf-written 57310813a841SMartin KaFai Lau * options. The active side can keep writing the bpf-options until 57320813a841SMartin KaFai Lau * it received a valid packet from the server side to confirm 57330813a841SMartin KaFai Lau * the earlier packet (and options) has been received. The later 57340813a841SMartin KaFai Lau * example patch is using it like this at the active side when the 57350813a841SMartin KaFai Lau * server is in syncookie mode. 57360813a841SMartin KaFai Lau * 57370813a841SMartin KaFai Lau * The bpf prog will usually turn this off in the common cases. 57380813a841SMartin KaFai Lau */ 573900d211a4SMartin KaFai Lau BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 57400813a841SMartin KaFai Lau /* Call bpf when kernel has received a header option that 57410813a841SMartin KaFai Lau * the kernel cannot handle. The bpf prog will be called under 57420813a841SMartin KaFai Lau * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 57430813a841SMartin KaFai Lau * 57440813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 57450813a841SMartin KaFai Lau * for the header option related helpers that will be useful 57460813a841SMartin KaFai Lau * to the bpf programs. 57470813a841SMartin KaFai Lau */ 574800d211a4SMartin KaFai Lau BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 57490813a841SMartin KaFai Lau /* Call bpf when the kernel is writing header options for the 57500813a841SMartin KaFai Lau * outgoing packet. The bpf prog will first be called 57510813a841SMartin KaFai Lau * to reserve space in a skb under 57520813a841SMartin KaFai Lau * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 57530813a841SMartin KaFai Lau * the bpf prog will be called to write the header option(s) 57540813a841SMartin KaFai Lau * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 57550813a841SMartin KaFai Lau * 57560813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 57570813a841SMartin KaFai Lau * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 57580813a841SMartin KaFai Lau * related helpers that will be useful to the bpf programs. 57590813a841SMartin KaFai Lau * 57600813a841SMartin KaFai Lau * The kernel gets its chance to reserve space and write 57610813a841SMartin KaFai Lau * options first before the BPF program does. 57620813a841SMartin KaFai Lau */ 5763331fca43SMartin KaFai Lau BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 57641aae4bddSAndrii Nakryiko /* Mask of all currently supported cb flags */ 5765331fca43SMartin KaFai Lau BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 57661aae4bddSAndrii Nakryiko }; 5767b13d8807SLawrence Brakmo 576840304b2aSLawrence Brakmo /* List of known BPF sock_ops operators. 576940304b2aSLawrence Brakmo * New entries can only be added at the end 577040304b2aSLawrence Brakmo */ 577140304b2aSLawrence Brakmo enum { 577240304b2aSLawrence Brakmo BPF_SOCK_OPS_VOID, 57738550f328SLawrence Brakmo BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 57748550f328SLawrence Brakmo * -1 if default value should be used 57758550f328SLawrence Brakmo */ 577613d3b1ebSLawrence Brakmo BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 577713d3b1ebSLawrence Brakmo * window (in packets) or -1 if default 577813d3b1ebSLawrence Brakmo * value should be used 577913d3b1ebSLawrence Brakmo */ 57809872a4bdSLawrence Brakmo BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 57819872a4bdSLawrence Brakmo * active connection is initialized 57829872a4bdSLawrence Brakmo */ 57839872a4bdSLawrence Brakmo BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 57849872a4bdSLawrence Brakmo * active connection is 57859872a4bdSLawrence Brakmo * established 57869872a4bdSLawrence Brakmo */ 57879872a4bdSLawrence Brakmo BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 57889872a4bdSLawrence Brakmo * passive connection is 57899872a4bdSLawrence Brakmo * established 57909872a4bdSLawrence Brakmo */ 579191b5b21cSLawrence Brakmo BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 579291b5b21cSLawrence Brakmo * needs ECN 579391b5b21cSLawrence Brakmo */ 5794e6546ef6SLawrence Brakmo BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 5795e6546ef6SLawrence Brakmo * based on the path and may be 5796e6546ef6SLawrence Brakmo * dependent on the congestion control 5797e6546ef6SLawrence Brakmo * algorithm. In general it indicates 5798e6546ef6SLawrence Brakmo * a congestion threshold. RTTs above 5799e6546ef6SLawrence Brakmo * this indicate congestion 5800e6546ef6SLawrence Brakmo */ 5801f89013f6SLawrence Brakmo BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 5802f89013f6SLawrence Brakmo * Arg1: value of icsk_retransmits 5803f89013f6SLawrence Brakmo * Arg2: value of icsk_rto 5804f89013f6SLawrence Brakmo * Arg3: whether RTO has expired 5805f89013f6SLawrence Brakmo */ 5806a31ad29eSLawrence Brakmo BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 5807a31ad29eSLawrence Brakmo * Arg1: sequence number of 1st byte 5808a31ad29eSLawrence Brakmo * Arg2: # segments 5809a31ad29eSLawrence Brakmo * Arg3: return value of 5810a31ad29eSLawrence Brakmo * tcp_transmit_skb (0 => success) 5811a31ad29eSLawrence Brakmo */ 5812d4487491SLawrence Brakmo BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 5813d4487491SLawrence Brakmo * Arg1: old_state 5814d4487491SLawrence Brakmo * Arg2: new_state 5815d4487491SLawrence Brakmo */ 5816f333ee0cSAndrey Ignatov BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 5817f333ee0cSAndrey Ignatov * socket transition to LISTEN state. 5818f333ee0cSAndrey Ignatov */ 581923729ff2SStanislav Fomichev BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 582023729ff2SStanislav Fomichev */ 58210813a841SMartin KaFai Lau BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 58220813a841SMartin KaFai Lau * It will be called to handle 58230813a841SMartin KaFai Lau * the packets received at 58240813a841SMartin KaFai Lau * an already established 58250813a841SMartin KaFai Lau * connection. 58260813a841SMartin KaFai Lau * 58270813a841SMartin KaFai Lau * sock_ops->skb_data: 58280813a841SMartin KaFai Lau * Referring to the received skb. 58290813a841SMartin KaFai Lau * It covers the TCP header only. 58300813a841SMartin KaFai Lau * 58310813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also 58320813a841SMartin KaFai Lau * be used to search for a 58330813a841SMartin KaFai Lau * particular option. 58340813a841SMartin KaFai Lau */ 58350813a841SMartin KaFai Lau BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 58360813a841SMartin KaFai Lau * header option later in 58370813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 58380813a841SMartin KaFai Lau * Arg1: bool want_cookie. (in 58390813a841SMartin KaFai Lau * writing SYNACK only) 58400813a841SMartin KaFai Lau * 58410813a841SMartin KaFai Lau * sock_ops->skb_data: 58420813a841SMartin KaFai Lau * Not available because no header has 58430813a841SMartin KaFai Lau * been written yet. 58440813a841SMartin KaFai Lau * 58450813a841SMartin KaFai Lau * sock_ops->skb_tcp_flags: 58460813a841SMartin KaFai Lau * The tcp_flags of the 58470813a841SMartin KaFai Lau * outgoing skb. (e.g. SYN, ACK, FIN). 58480813a841SMartin KaFai Lau * 58490813a841SMartin KaFai Lau * bpf_reserve_hdr_opt() should 58500813a841SMartin KaFai Lau * be used to reserve space. 58510813a841SMartin KaFai Lau */ 58520813a841SMartin KaFai Lau BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 58530813a841SMartin KaFai Lau * Arg1: bool want_cookie. (in 58540813a841SMartin KaFai Lau * writing SYNACK only) 58550813a841SMartin KaFai Lau * 58560813a841SMartin KaFai Lau * sock_ops->skb_data: 58570813a841SMartin KaFai Lau * Referring to the outgoing skb. 58580813a841SMartin KaFai Lau * It covers the TCP header 58590813a841SMartin KaFai Lau * that has already been written 58600813a841SMartin KaFai Lau * by the kernel and the 58610813a841SMartin KaFai Lau * earlier bpf-progs. 58620813a841SMartin KaFai Lau * 58630813a841SMartin KaFai Lau * sock_ops->skb_tcp_flags: 58640813a841SMartin KaFai Lau * The tcp_flags of the outgoing 58650813a841SMartin KaFai Lau * skb. (e.g. SYN, ACK, FIN). 58660813a841SMartin KaFai Lau * 58670813a841SMartin KaFai Lau * bpf_store_hdr_opt() should 58680813a841SMartin KaFai Lau * be used to write the 58690813a841SMartin KaFai Lau * option. 58700813a841SMartin KaFai Lau * 58710813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also 58720813a841SMartin KaFai Lau * be used to search for a 58730813a841SMartin KaFai Lau * particular option that 58740813a841SMartin KaFai Lau * has already been written 58750813a841SMartin KaFai Lau * by the kernel or the 58760813a841SMartin KaFai Lau * earlier bpf-progs. 58770813a841SMartin KaFai Lau */ 5878d4487491SLawrence Brakmo }; 5879d4487491SLawrence Brakmo 5880d4487491SLawrence Brakmo /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 5881d4487491SLawrence Brakmo * changes between the TCP and BPF versions. Ideally this should never happen. 5882d4487491SLawrence Brakmo * If it does, we need to add code to convert them before calling 5883d4487491SLawrence Brakmo * the BPF sock_ops function. 5884d4487491SLawrence Brakmo */ 5885d4487491SLawrence Brakmo enum { 5886d4487491SLawrence Brakmo BPF_TCP_ESTABLISHED = 1, 5887d4487491SLawrence Brakmo BPF_TCP_SYN_SENT, 5888d4487491SLawrence Brakmo BPF_TCP_SYN_RECV, 5889d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT1, 5890d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT2, 5891d4487491SLawrence Brakmo BPF_TCP_TIME_WAIT, 5892d4487491SLawrence Brakmo BPF_TCP_CLOSE, 5893d4487491SLawrence Brakmo BPF_TCP_CLOSE_WAIT, 5894d4487491SLawrence Brakmo BPF_TCP_LAST_ACK, 5895d4487491SLawrence Brakmo BPF_TCP_LISTEN, 5896d4487491SLawrence Brakmo BPF_TCP_CLOSING, /* Now a valid state */ 5897d4487491SLawrence Brakmo BPF_TCP_NEW_SYN_RECV, 5898d4487491SLawrence Brakmo 5899d4487491SLawrence Brakmo BPF_TCP_MAX_STATES /* Leave at the end! */ 590040304b2aSLawrence Brakmo }; 590140304b2aSLawrence Brakmo 59021aae4bddSAndrii Nakryiko enum { 59031aae4bddSAndrii Nakryiko TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 59041aae4bddSAndrii Nakryiko TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 59052b8ee4f0SMartin KaFai Lau TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 5906ca584ba0SMartin KaFai Lau TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 59070813a841SMartin KaFai Lau /* Copy the SYN pkt to optval 59080813a841SMartin KaFai Lau * 59090813a841SMartin KaFai Lau * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 59100813a841SMartin KaFai Lau * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 59110813a841SMartin KaFai Lau * to only getting from the saved_syn. It can either get the 59120813a841SMartin KaFai Lau * syn packet from: 59130813a841SMartin KaFai Lau * 59140813a841SMartin KaFai Lau * 1. the just-received SYN packet (only available when writing the 59150813a841SMartin KaFai Lau * SYNACK). It will be useful when it is not necessary to 59160813a841SMartin KaFai Lau * save the SYN packet for latter use. It is also the only way 59170813a841SMartin KaFai Lau * to get the SYN during syncookie mode because the syn 59180813a841SMartin KaFai Lau * packet cannot be saved during syncookie. 59190813a841SMartin KaFai Lau * 59200813a841SMartin KaFai Lau * OR 59210813a841SMartin KaFai Lau * 59220813a841SMartin KaFai Lau * 2. the earlier saved syn which was done by 59230813a841SMartin KaFai Lau * bpf_setsockopt(TCP_SAVE_SYN). 59240813a841SMartin KaFai Lau * 59250813a841SMartin KaFai Lau * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 59260813a841SMartin KaFai Lau * SYN packet is obtained. 59270813a841SMartin KaFai Lau * 59280813a841SMartin KaFai Lau * If the bpf-prog does not need the IP[46] header, the 59290813a841SMartin KaFai Lau * bpf-prog can avoid parsing the IP header by using 59300813a841SMartin KaFai Lau * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 59310813a841SMartin KaFai Lau * IP[46] and TCP header by using TCP_BPF_SYN_IP. 59320813a841SMartin KaFai Lau * 59330813a841SMartin KaFai Lau * >0: Total number of bytes copied 59340813a841SMartin KaFai Lau * -ENOSPC: Not enough space in optval. Only optlen number of 59350813a841SMartin KaFai Lau * bytes is copied. 59360813a841SMartin KaFai Lau * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 59370813a841SMartin KaFai Lau * is not saved by setsockopt(TCP_SAVE_SYN). 59380813a841SMartin KaFai Lau */ 59390813a841SMartin KaFai Lau TCP_BPF_SYN = 1005, /* Copy the TCP header */ 59400813a841SMartin KaFai Lau TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 5941267cf9faSMartin KaFai Lau TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 59420813a841SMartin KaFai Lau }; 59430813a841SMartin KaFai Lau 59440813a841SMartin KaFai Lau enum { 59450813a841SMartin KaFai Lau BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 59460813a841SMartin KaFai Lau }; 59470813a841SMartin KaFai Lau 59480813a841SMartin KaFai Lau /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 59490813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 59500813a841SMartin KaFai Lau */ 59510813a841SMartin KaFai Lau enum { 59520813a841SMartin KaFai Lau BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 59530813a841SMartin KaFai Lau * total option spaces 59540813a841SMartin KaFai Lau * required for an established 59550813a841SMartin KaFai Lau * sk in order to calculate the 59560813a841SMartin KaFai Lau * MSS. No skb is actually 59570813a841SMartin KaFai Lau * sent. 59580813a841SMartin KaFai Lau */ 59590813a841SMartin KaFai Lau BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 59600813a841SMartin KaFai Lau * when sending a SYN. 59610813a841SMartin KaFai Lau */ 59621aae4bddSAndrii Nakryiko }; 5963fc747810SLawrence Brakmo 5964908432caSYonghong Song struct bpf_perf_event_value { 5965908432caSYonghong Song __u64 counter; 5966908432caSYonghong Song __u64 enabled; 5967908432caSYonghong Song __u64 running; 5968908432caSYonghong Song }; 5969908432caSYonghong Song 59701aae4bddSAndrii Nakryiko enum { 59711aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 59721aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_READ = (1ULL << 1), 59731aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_WRITE = (1ULL << 2), 59741aae4bddSAndrii Nakryiko }; 5975ebc614f6SRoman Gushchin 59761aae4bddSAndrii Nakryiko enum { 59771aae4bddSAndrii Nakryiko BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 59781aae4bddSAndrii Nakryiko BPF_DEVCG_DEV_CHAR = (1ULL << 1), 59791aae4bddSAndrii Nakryiko }; 5980ebc614f6SRoman Gushchin 5981ebc614f6SRoman Gushchin struct bpf_cgroup_dev_ctx { 598206ef0ccbSYonghong Song /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 598306ef0ccbSYonghong Song __u32 access_type; 5984ebc614f6SRoman Gushchin __u32 major; 5985ebc614f6SRoman Gushchin __u32 minor; 5986ebc614f6SRoman Gushchin }; 5987ebc614f6SRoman Gushchin 5988c4f6699dSAlexei Starovoitov struct bpf_raw_tracepoint_args { 5989c4f6699dSAlexei Starovoitov __u64 args[0]; 5990c4f6699dSAlexei Starovoitov }; 5991c4f6699dSAlexei Starovoitov 599287f5fc7eSDavid Ahern /* DIRECT: Skip the FIB rules and go to FIB table associated with device 599387f5fc7eSDavid Ahern * OUTPUT: Do lookup from egress perspective; default is ingress 599487f5fc7eSDavid Ahern */ 59951aae4bddSAndrii Nakryiko enum { 59961aae4bddSAndrii Nakryiko BPF_FIB_LOOKUP_DIRECT = (1U << 0), 59971aae4bddSAndrii Nakryiko BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 59981aae4bddSAndrii Nakryiko }; 599987f5fc7eSDavid Ahern 60004c79579bSDavid Ahern enum { 60014c79579bSDavid Ahern BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 60024c79579bSDavid Ahern BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 60034c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 60044c79579bSDavid Ahern BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 60054c79579bSDavid Ahern BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 60064c79579bSDavid Ahern BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 60074c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 60084c79579bSDavid Ahern BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 60094c79579bSDavid Ahern BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 60104c79579bSDavid Ahern }; 60114c79579bSDavid Ahern 601287f5fc7eSDavid Ahern struct bpf_fib_lookup { 6013fa898d76SDavid Ahern /* input: network family for lookup (AF_INET, AF_INET6) 6014fa898d76SDavid Ahern * output: network family of egress nexthop 6015fa898d76SDavid Ahern */ 6016fa898d76SDavid Ahern __u8 family; 601787f5fc7eSDavid Ahern 601887f5fc7eSDavid Ahern /* set if lookup is to consider L4 data - e.g., FIB rules */ 601987f5fc7eSDavid Ahern __u8 l4_protocol; 602087f5fc7eSDavid Ahern __be16 sport; 602187f5fc7eSDavid Ahern __be16 dport; 602287f5fc7eSDavid Ahern 6023e1850ea9SJesper Dangaard Brouer union { /* used for MTU check */ 6024e1850ea9SJesper Dangaard Brouer /* input to lookup */ 6025e1850ea9SJesper Dangaard Brouer __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 60264c79579bSDavid Ahern 6027e1850ea9SJesper Dangaard Brouer /* output: MTU value */ 6028e1850ea9SJesper Dangaard Brouer __u16 mtu_result; 6029e1850ea9SJesper Dangaard Brouer }; 60304c79579bSDavid Ahern /* input: L3 device index for lookup 60314c79579bSDavid Ahern * output: device index from FIB lookup 60324c79579bSDavid Ahern */ 60334c79579bSDavid Ahern __u32 ifindex; 603487f5fc7eSDavid Ahern 603587f5fc7eSDavid Ahern union { 603687f5fc7eSDavid Ahern /* inputs to lookup */ 603787f5fc7eSDavid Ahern __u8 tos; /* AF_INET */ 6038bd3a08aaSDavid Ahern __be32 flowinfo; /* AF_INET6, flow_label + priority */ 603987f5fc7eSDavid Ahern 6040fa898d76SDavid Ahern /* output: metric of fib result (IPv4/IPv6 only) */ 604187f5fc7eSDavid Ahern __u32 rt_metric; 604287f5fc7eSDavid Ahern }; 604387f5fc7eSDavid Ahern 604487f5fc7eSDavid Ahern union { 604587f5fc7eSDavid Ahern __be32 ipv4_src; 604687f5fc7eSDavid Ahern __u32 ipv6_src[4]; /* in6_addr; network order */ 604787f5fc7eSDavid Ahern }; 604887f5fc7eSDavid Ahern 6049fa898d76SDavid Ahern /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 6050fa898d76SDavid Ahern * network header. output: bpf_fib_lookup sets to gateway address 6051fa898d76SDavid Ahern * if FIB lookup returns gateway route 605287f5fc7eSDavid Ahern */ 605387f5fc7eSDavid Ahern union { 605487f5fc7eSDavid Ahern __be32 ipv4_dst; 605587f5fc7eSDavid Ahern __u32 ipv6_dst[4]; /* in6_addr; network order */ 605687f5fc7eSDavid Ahern }; 605787f5fc7eSDavid Ahern 605887f5fc7eSDavid Ahern /* output */ 605987f5fc7eSDavid Ahern __be16 h_vlan_proto; 606087f5fc7eSDavid Ahern __be16 h_vlan_TCI; 606187f5fc7eSDavid Ahern __u8 smac[6]; /* ETH_ALEN */ 606287f5fc7eSDavid Ahern __u8 dmac[6]; /* ETH_ALEN */ 606387f5fc7eSDavid Ahern }; 606487f5fc7eSDavid Ahern 6065ba452c9eSToke Høiland-Jørgensen struct bpf_redir_neigh { 6066ba452c9eSToke Høiland-Jørgensen /* network family for lookup (AF_INET, AF_INET6) */ 6067ba452c9eSToke Høiland-Jørgensen __u32 nh_family; 6068ba452c9eSToke Høiland-Jørgensen /* network address of nexthop; skips fib lookup to find gateway */ 6069ba452c9eSToke Høiland-Jørgensen union { 6070ba452c9eSToke Høiland-Jørgensen __be32 ipv4_nh; 6071ba452c9eSToke Høiland-Jørgensen __u32 ipv6_nh[4]; /* in6_addr; network order */ 6072ba452c9eSToke Høiland-Jørgensen }; 6073ba452c9eSToke Høiland-Jørgensen }; 6074ba452c9eSToke Høiland-Jørgensen 607534b2021cSJesper Dangaard Brouer /* bpf_check_mtu flags*/ 607634b2021cSJesper Dangaard Brouer enum bpf_check_mtu_flags { 607734b2021cSJesper Dangaard Brouer BPF_MTU_CHK_SEGS = (1U << 0), 607834b2021cSJesper Dangaard Brouer }; 607934b2021cSJesper Dangaard Brouer 608034b2021cSJesper Dangaard Brouer enum bpf_check_mtu_ret { 608134b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 608234b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 608334b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 608434b2021cSJesper Dangaard Brouer }; 608534b2021cSJesper Dangaard Brouer 608641bdc4b4SYonghong Song enum bpf_task_fd_type { 608741bdc4b4SYonghong Song BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 608841bdc4b4SYonghong Song BPF_FD_TYPE_TRACEPOINT, /* tp name */ 608941bdc4b4SYonghong Song BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 609041bdc4b4SYonghong Song BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 609141bdc4b4SYonghong Song BPF_FD_TYPE_UPROBE, /* filename + offset */ 609241bdc4b4SYonghong Song BPF_FD_TYPE_URETPROBE, /* filename + offset */ 609341bdc4b4SYonghong Song }; 609441bdc4b4SYonghong Song 60951aae4bddSAndrii Nakryiko enum { 60961aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 60971aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 60981aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 60991aae4bddSAndrii Nakryiko }; 6100086f9568SStanislav Fomichev 6101d58e468bSPetar Penkov struct bpf_flow_keys { 6102d58e468bSPetar Penkov __u16 nhoff; 6103d58e468bSPetar Penkov __u16 thoff; 6104d58e468bSPetar Penkov __u16 addr_proto; /* ETH_P_* of valid addrs */ 6105d58e468bSPetar Penkov __u8 is_frag; 6106d58e468bSPetar Penkov __u8 is_first_frag; 6107d58e468bSPetar Penkov __u8 is_encap; 6108d58e468bSPetar Penkov __u8 ip_proto; 6109d58e468bSPetar Penkov __be16 n_proto; 6110d58e468bSPetar Penkov __be16 sport; 6111d58e468bSPetar Penkov __be16 dport; 6112d58e468bSPetar Penkov union { 6113d58e468bSPetar Penkov struct { 6114d58e468bSPetar Penkov __be32 ipv4_src; 6115d58e468bSPetar Penkov __be32 ipv4_dst; 6116d58e468bSPetar Penkov }; 6117d58e468bSPetar Penkov struct { 6118d58e468bSPetar Penkov __u32 ipv6_src[4]; /* in6_addr; network order */ 6119d58e468bSPetar Penkov __u32 ipv6_dst[4]; /* in6_addr; network order */ 6120d58e468bSPetar Penkov }; 6121d58e468bSPetar Penkov }; 6122086f9568SStanislav Fomichev __u32 flags; 612371c99e32SStanislav Fomichev __be32 flow_label; 6124d58e468bSPetar Penkov }; 6125d58e468bSPetar Penkov 6126838e9690SYonghong Song struct bpf_func_info { 6127d30d42e0SMartin KaFai Lau __u32 insn_off; 6128838e9690SYonghong Song __u32 type_id; 6129838e9690SYonghong Song }; 6130838e9690SYonghong Song 6131c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 6132c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 6133c454a46bSMartin KaFai Lau 6134c454a46bSMartin KaFai Lau struct bpf_line_info { 6135c454a46bSMartin KaFai Lau __u32 insn_off; 6136c454a46bSMartin KaFai Lau __u32 file_name_off; 6137c454a46bSMartin KaFai Lau __u32 line_off; 6138c454a46bSMartin KaFai Lau __u32 line_col; 6139c454a46bSMartin KaFai Lau }; 6140c454a46bSMartin KaFai Lau 6141d83525caSAlexei Starovoitov struct bpf_spin_lock { 6142d83525caSAlexei Starovoitov __u32 val; 6143d83525caSAlexei Starovoitov }; 61447b146cebSAndrey Ignatov 6145*b00628b1SAlexei Starovoitov struct bpf_timer { 6146*b00628b1SAlexei Starovoitov __u64 :64; 6147*b00628b1SAlexei Starovoitov __u64 :64; 6148*b00628b1SAlexei Starovoitov } __attribute__((aligned(8))); 6149*b00628b1SAlexei Starovoitov 61507b146cebSAndrey Ignatov struct bpf_sysctl { 61517b146cebSAndrey Ignatov __u32 write; /* Sysctl is being read (= 0) or written (= 1). 61527b146cebSAndrey Ignatov * Allows 1,2,4-byte read, but no write. 61537b146cebSAndrey Ignatov */ 6154e1550bfeSAndrey Ignatov __u32 file_pos; /* Sysctl file position to read from, write to. 6155e1550bfeSAndrey Ignatov * Allows 1,2,4-byte read an 4-byte write. 6156e1550bfeSAndrey Ignatov */ 61577b146cebSAndrey Ignatov }; 61587b146cebSAndrey Ignatov 61590d01da6aSStanislav Fomichev struct bpf_sockopt { 61600d01da6aSStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 61610d01da6aSStanislav Fomichev __bpf_md_ptr(void *, optval); 61620d01da6aSStanislav Fomichev __bpf_md_ptr(void *, optval_end); 61630d01da6aSStanislav Fomichev 61640d01da6aSStanislav Fomichev __s32 level; 61650d01da6aSStanislav Fomichev __s32 optname; 61660d01da6aSStanislav Fomichev __s32 optlen; 61670d01da6aSStanislav Fomichev __s32 retval; 61680d01da6aSStanislav Fomichev }; 61690d01da6aSStanislav Fomichev 6170b4490c5cSCarlos Neira struct bpf_pidns_info { 6171b4490c5cSCarlos Neira __u32 pid; 6172b4490c5cSCarlos Neira __u32 tgid; 6173b4490c5cSCarlos Neira }; 6174e9ddbb77SJakub Sitnicki 6175e9ddbb77SJakub Sitnicki /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 6176e9ddbb77SJakub Sitnicki struct bpf_sk_lookup { 61777c32e8f8SLorenz Bauer union { 6178e9ddbb77SJakub Sitnicki __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 61797c32e8f8SLorenz Bauer __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 61807c32e8f8SLorenz Bauer }; 6181e9ddbb77SJakub Sitnicki 6182e9ddbb77SJakub Sitnicki __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 6183e9ddbb77SJakub Sitnicki __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 6184e9ddbb77SJakub Sitnicki __u32 remote_ip4; /* Network byte order */ 6185e9ddbb77SJakub Sitnicki __u32 remote_ip6[4]; /* Network byte order */ 6186e9ddbb77SJakub Sitnicki __u32 remote_port; /* Network byte order */ 6187e9ddbb77SJakub Sitnicki __u32 local_ip4; /* Network byte order */ 6188e9ddbb77SJakub Sitnicki __u32 local_ip6[4]; /* Network byte order */ 6189e9ddbb77SJakub Sitnicki __u32 local_port; /* Host byte order */ 6190e9ddbb77SJakub Sitnicki }; 6191e9ddbb77SJakub Sitnicki 6192c4d0bfb4SAlan Maguire /* 6193c4d0bfb4SAlan Maguire * struct btf_ptr is used for typed pointer representation; the 6194c4d0bfb4SAlan Maguire * type id is used to render the pointer data as the appropriate type 6195c4d0bfb4SAlan Maguire * via the bpf_snprintf_btf() helper described above. A flags field - 6196c4d0bfb4SAlan Maguire * potentially to specify additional details about the BTF pointer 6197c4d0bfb4SAlan Maguire * (rather than its mode of display) - is included for future use. 6198c4d0bfb4SAlan Maguire * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 6199c4d0bfb4SAlan Maguire */ 6200c4d0bfb4SAlan Maguire struct btf_ptr { 6201c4d0bfb4SAlan Maguire void *ptr; 6202c4d0bfb4SAlan Maguire __u32 type_id; 6203c4d0bfb4SAlan Maguire __u32 flags; /* BTF ptr flags; unused at present. */ 6204c4d0bfb4SAlan Maguire }; 6205c4d0bfb4SAlan Maguire 6206c4d0bfb4SAlan Maguire /* 6207c4d0bfb4SAlan Maguire * Flags to control bpf_snprintf_btf() behaviour. 6208c4d0bfb4SAlan Maguire * - BTF_F_COMPACT: no formatting around type information 6209c4d0bfb4SAlan Maguire * - BTF_F_NONAME: no struct/union member names/types 6210c4d0bfb4SAlan Maguire * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 6211c4d0bfb4SAlan Maguire * equivalent to %px. 6212c4d0bfb4SAlan Maguire * - BTF_F_ZERO: show zero-valued struct/union members; they 6213c4d0bfb4SAlan Maguire * are not displayed by default 6214c4d0bfb4SAlan Maguire */ 6215c4d0bfb4SAlan Maguire enum { 6216c4d0bfb4SAlan Maguire BTF_F_COMPACT = (1ULL << 0), 6217c4d0bfb4SAlan Maguire BTF_F_NONAME = (1ULL << 1), 6218c4d0bfb4SAlan Maguire BTF_F_PTR_RAW = (1ULL << 2), 6219c4d0bfb4SAlan Maguire BTF_F_ZERO = (1ULL << 3), 6220c4d0bfb4SAlan Maguire }; 6221c4d0bfb4SAlan Maguire 6222daedfb22SAlexei Starovoitov #endif /* _UAPI__LINUX_BPF_H__ */ 6223