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_XDP** 328f3c45326SJoe Stringer * *ctx_in* and *ctx_out* must be NULL. 329f3c45326SJoe Stringer * 330f3c45326SJoe Stringer * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 331f3c45326SJoe Stringer * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 332f3c45326SJoe Stringer * 333f3c45326SJoe Stringer * *ctx_out*, *data_in* and *data_out* must be NULL. 334f3c45326SJoe Stringer * *repeat* must be zero. 335f3c45326SJoe Stringer * 336f67c9cbfSJoe Stringer * Return 337f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 338f67c9cbfSJoe Stringer * is set appropriately. 339f67c9cbfSJoe Stringer * 3402a3fdca4SJoe Stringer * **ENOSPC** 3412a3fdca4SJoe Stringer * Either *data_size_out* or *ctx_size_out* is too small. 3422a3fdca4SJoe Stringer * **ENOTSUPP** 3432a3fdca4SJoe Stringer * This command is not supported by the program type of 3442a3fdca4SJoe Stringer * the program referred to by *prog_fd*. 3452a3fdca4SJoe Stringer * 346f67c9cbfSJoe Stringer * BPF_PROG_GET_NEXT_ID 347f67c9cbfSJoe Stringer * Description 348f67c9cbfSJoe Stringer * Fetch the next eBPF program currently loaded into the kernel. 349f67c9cbfSJoe Stringer * 350f67c9cbfSJoe Stringer * Looks for the eBPF program with an id greater than *start_id* 351f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF programs 352f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 353f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 354f67c9cbfSJoe Stringer * 355f67c9cbfSJoe Stringer * Return 356f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 357f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 358f67c9cbfSJoe Stringer * 359f67c9cbfSJoe Stringer * BPF_MAP_GET_NEXT_ID 360f67c9cbfSJoe Stringer * Description 361f67c9cbfSJoe Stringer * Fetch the next eBPF map currently loaded into the kernel. 362f67c9cbfSJoe Stringer * 363f67c9cbfSJoe Stringer * Looks for the eBPF map with an id greater than *start_id* 364f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF maps 365f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 366f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 367f67c9cbfSJoe Stringer * 368f67c9cbfSJoe Stringer * Return 369f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 370f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 371f67c9cbfSJoe Stringer * 372f67c9cbfSJoe Stringer * BPF_PROG_GET_FD_BY_ID 373f67c9cbfSJoe Stringer * Description 374f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF program corresponding to 375f67c9cbfSJoe Stringer * *prog_id*. 376f67c9cbfSJoe Stringer * 377f67c9cbfSJoe Stringer * Return 378f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 379f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 380f67c9cbfSJoe Stringer * 381f67c9cbfSJoe Stringer * BPF_MAP_GET_FD_BY_ID 382f67c9cbfSJoe Stringer * Description 383f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF map corresponding to 384f67c9cbfSJoe Stringer * *map_id*. 385f67c9cbfSJoe Stringer * 386f67c9cbfSJoe Stringer * Return 387f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 388f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 389f67c9cbfSJoe Stringer * 390f67c9cbfSJoe Stringer * BPF_OBJ_GET_INFO_BY_FD 391f67c9cbfSJoe Stringer * Description 392f67c9cbfSJoe Stringer * Obtain information about the eBPF object corresponding to 393f67c9cbfSJoe Stringer * *bpf_fd*. 394f67c9cbfSJoe Stringer * 395f67c9cbfSJoe Stringer * Populates up to *info_len* bytes of *info*, which will be in 396f67c9cbfSJoe Stringer * one of the following formats depending on the eBPF object type 397f67c9cbfSJoe Stringer * of *bpf_fd*: 398f67c9cbfSJoe Stringer * 399f67c9cbfSJoe Stringer * * **struct bpf_prog_info** 400f67c9cbfSJoe Stringer * * **struct bpf_map_info** 401f67c9cbfSJoe Stringer * * **struct bpf_btf_info** 402f67c9cbfSJoe Stringer * * **struct bpf_link_info** 403f67c9cbfSJoe Stringer * 404f67c9cbfSJoe Stringer * Return 405f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 406f67c9cbfSJoe Stringer * is set appropriately. 407f67c9cbfSJoe Stringer * 408f67c9cbfSJoe Stringer * BPF_PROG_QUERY 409f67c9cbfSJoe Stringer * Description 410f67c9cbfSJoe Stringer * Obtain information about eBPF programs associated with the 411f67c9cbfSJoe Stringer * specified *attach_type* hook. 412f67c9cbfSJoe Stringer * 4135d999994SJoe Stringer * The *target_fd* must be a valid file descriptor for a kernel 4145d999994SJoe Stringer * object which depends on the attach type of *attach_bpf_fd*: 4155d999994SJoe Stringer * 4165d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_DEVICE**, 4175d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SKB**, 4185d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK**, 4195d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 4205d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 4215d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 4225d999994SJoe Stringer * **BPF_PROG_TYPE_SOCK_OPS** 4235d999994SJoe Stringer * 4245d999994SJoe Stringer * Control Group v2 hierarchy with the eBPF controller 4255d999994SJoe Stringer * enabled. Requires the kernel to be compiled with 4265d999994SJoe Stringer * **CONFIG_CGROUP_BPF**. 4275d999994SJoe Stringer * 4285d999994SJoe Stringer * **BPF_PROG_TYPE_FLOW_DISSECTOR** 4295d999994SJoe Stringer * 4305d999994SJoe Stringer * Network namespace (eg /proc/self/ns/net). 4315d999994SJoe Stringer * 4325d999994SJoe Stringer * **BPF_PROG_TYPE_LIRC_MODE2** 4335d999994SJoe Stringer * 4345d999994SJoe Stringer * LIRC device path (eg /dev/lircN). Requires the kernel 4355d999994SJoe Stringer * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 4365d999994SJoe Stringer * 4375d999994SJoe Stringer * **BPF_PROG_QUERY** always fetches the number of programs 4385d999994SJoe Stringer * attached and the *attach_flags* which were used to attach those 4395d999994SJoe Stringer * programs. Additionally, if *prog_ids* is nonzero and the number 4405d999994SJoe Stringer * of attached programs is less than *prog_cnt*, populates 4415d999994SJoe Stringer * *prog_ids* with the eBPF program ids of the programs attached 4425d999994SJoe Stringer * at *target_fd*. 4435d999994SJoe Stringer * 4445d999994SJoe Stringer * The following flags may alter the result: 4455d999994SJoe Stringer * 4465d999994SJoe Stringer * **BPF_F_QUERY_EFFECTIVE** 4475d999994SJoe Stringer * Only return information regarding programs which are 4485d999994SJoe Stringer * currently effective at the specified *target_fd*. 4495d999994SJoe Stringer * 450f67c9cbfSJoe Stringer * Return 451f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 452f67c9cbfSJoe Stringer * is set appropriately. 453f67c9cbfSJoe Stringer * 454f67c9cbfSJoe Stringer * BPF_RAW_TRACEPOINT_OPEN 455f67c9cbfSJoe Stringer * Description 456f67c9cbfSJoe Stringer * Attach an eBPF program to a tracepoint *name* to access kernel 457f67c9cbfSJoe Stringer * internal arguments of the tracepoint in their raw form. 458f67c9cbfSJoe Stringer * 459f67c9cbfSJoe Stringer * The *prog_fd* must be a valid file descriptor associated with 460f67c9cbfSJoe Stringer * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 461f67c9cbfSJoe Stringer * 462f67c9cbfSJoe Stringer * No ABI guarantees are made about the content of tracepoint 463f67c9cbfSJoe Stringer * arguments exposed to the corresponding eBPF program. 464f67c9cbfSJoe Stringer * 465f67c9cbfSJoe Stringer * Applying **close**\ (2) to the file descriptor returned by 466f67c9cbfSJoe Stringer * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 467f67c9cbfSJoe Stringer * 468f67c9cbfSJoe Stringer * Return 469f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 470f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 471f67c9cbfSJoe Stringer * 472f67c9cbfSJoe Stringer * BPF_BTF_LOAD 473f67c9cbfSJoe Stringer * Description 474f67c9cbfSJoe Stringer * Verify and load BPF Type Format (BTF) metadata into the kernel, 475f67c9cbfSJoe Stringer * returning a new file descriptor associated with the metadata. 476f67c9cbfSJoe Stringer * BTF is described in more detail at 477f67c9cbfSJoe Stringer * https://www.kernel.org/doc/html/latest/bpf/btf.html. 478f67c9cbfSJoe Stringer * 479f67c9cbfSJoe Stringer * The *btf* parameter must point to valid memory providing 480f67c9cbfSJoe Stringer * *btf_size* bytes of BTF binary metadata. 481f67c9cbfSJoe Stringer * 482f67c9cbfSJoe Stringer * The returned file descriptor can be passed to other **bpf**\ () 483f67c9cbfSJoe Stringer * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 484f67c9cbfSJoe Stringer * associate the BTF with those objects. 485f67c9cbfSJoe Stringer * 486f67c9cbfSJoe Stringer * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 487f67c9cbfSJoe Stringer * parameters to specify a *btf_log_buf*, *btf_log_size* and 488f67c9cbfSJoe Stringer * *btf_log_level* which allow the kernel to return freeform log 489f67c9cbfSJoe Stringer * output regarding the BTF verification process. 490f67c9cbfSJoe Stringer * 491f67c9cbfSJoe Stringer * Return 492f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 493f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 494f67c9cbfSJoe Stringer * 495f67c9cbfSJoe Stringer * BPF_BTF_GET_FD_BY_ID 496f67c9cbfSJoe Stringer * Description 497f67c9cbfSJoe Stringer * Open a file descriptor for the BPF Type Format (BTF) 498f67c9cbfSJoe Stringer * corresponding to *btf_id*. 499f67c9cbfSJoe Stringer * 500f67c9cbfSJoe Stringer * Return 501f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 502f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 503f67c9cbfSJoe Stringer * 504f67c9cbfSJoe Stringer * BPF_TASK_FD_QUERY 505f67c9cbfSJoe Stringer * Description 506f67c9cbfSJoe Stringer * Obtain information about eBPF programs associated with the 507f67c9cbfSJoe Stringer * target process identified by *pid* and *fd*. 508f67c9cbfSJoe Stringer * 509f67c9cbfSJoe Stringer * If the *pid* and *fd* are associated with a tracepoint, kprobe 510f67c9cbfSJoe Stringer * or uprobe perf event, then the *prog_id* and *fd_type* will 511f67c9cbfSJoe Stringer * be populated with the eBPF program id and file descriptor type 512f67c9cbfSJoe Stringer * of type **bpf_task_fd_type**. If associated with a kprobe or 513f67c9cbfSJoe Stringer * uprobe, the *probe_offset* and *probe_addr* will also be 514f67c9cbfSJoe Stringer * populated. Optionally, if *buf* is provided, then up to 515f67c9cbfSJoe Stringer * *buf_len* bytes of *buf* will be populated with the name of 516f67c9cbfSJoe Stringer * the tracepoint, kprobe or uprobe. 517f67c9cbfSJoe Stringer * 518f67c9cbfSJoe Stringer * The resulting *prog_id* may be introspected in deeper detail 519f67c9cbfSJoe Stringer * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 520f67c9cbfSJoe Stringer * 521f67c9cbfSJoe Stringer * Return 522f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 523f67c9cbfSJoe Stringer * is set appropriately. 524f67c9cbfSJoe Stringer * 525f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_AND_DELETE_ELEM 526f67c9cbfSJoe Stringer * Description 527f67c9cbfSJoe Stringer * Look up an element with the given *key* in the map referred to 528f67c9cbfSJoe Stringer * by the file descriptor *fd*, and if found, delete the element. 529f67c9cbfSJoe Stringer * 530f67c9cbfSJoe Stringer * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 531f67c9cbfSJoe Stringer * implement this command as a "pop" operation, deleting the top 532f67c9cbfSJoe Stringer * element rather than one corresponding to *key*. 533f67c9cbfSJoe Stringer * The *key* and *key_len* parameters should be zeroed when 534f67c9cbfSJoe Stringer * issuing this operation for these map types. 535f67c9cbfSJoe Stringer * 536f67c9cbfSJoe Stringer * This command is only valid for the following map types: 537f67c9cbfSJoe Stringer * * **BPF_MAP_TYPE_QUEUE** 538f67c9cbfSJoe Stringer * * **BPF_MAP_TYPE_STACK** 539f67c9cbfSJoe Stringer * 540f67c9cbfSJoe Stringer * Return 541f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 542f67c9cbfSJoe Stringer * is set appropriately. 543f67c9cbfSJoe Stringer * 544f67c9cbfSJoe Stringer * BPF_MAP_FREEZE 545f67c9cbfSJoe Stringer * Description 546f67c9cbfSJoe Stringer * Freeze the permissions of the specified map. 547f67c9cbfSJoe Stringer * 548f67c9cbfSJoe Stringer * Write permissions may be frozen by passing zero *flags*. 549f67c9cbfSJoe Stringer * Upon success, no future syscall invocations may alter the 550f67c9cbfSJoe Stringer * map state of *map_fd*. Write operations from eBPF programs 551f67c9cbfSJoe Stringer * are still possible for a frozen map. 552f67c9cbfSJoe Stringer * 553f67c9cbfSJoe Stringer * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 554f67c9cbfSJoe Stringer * 555f67c9cbfSJoe Stringer * Return 556f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 557f67c9cbfSJoe Stringer * is set appropriately. 558f67c9cbfSJoe Stringer * 559f67c9cbfSJoe Stringer * BPF_BTF_GET_NEXT_ID 560f67c9cbfSJoe Stringer * Description 561f67c9cbfSJoe Stringer * Fetch the next BPF Type Format (BTF) object currently loaded 562f67c9cbfSJoe Stringer * into the kernel. 563f67c9cbfSJoe Stringer * 564f67c9cbfSJoe Stringer * Looks for the BTF object with an id greater than *start_id* 565f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other BTF objects 566f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 567f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 568f67c9cbfSJoe Stringer * 569f67c9cbfSJoe Stringer * Return 570f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 571f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 572f67c9cbfSJoe Stringer * 573f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_BATCH 574f67c9cbfSJoe Stringer * Description 575f67c9cbfSJoe Stringer * Iterate and fetch multiple elements in a map. 576f67c9cbfSJoe Stringer * 5770cb80454SJoe Stringer * Two opaque values are used to manage batch operations, 5780cb80454SJoe Stringer * *in_batch* and *out_batch*. Initially, *in_batch* must be set 5790cb80454SJoe Stringer * to NULL to begin the batched operation. After each subsequent 5800cb80454SJoe Stringer * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 5810cb80454SJoe Stringer * *out_batch* as the *in_batch* for the next operation to 5820cb80454SJoe Stringer * continue iteration from the current point. 5830cb80454SJoe Stringer * 5840cb80454SJoe Stringer * The *keys* and *values* are output parameters which must point 5850cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 5860cb80454SJoe Stringer * and value size of the map *map_fd*. The *keys* buffer must be 5870cb80454SJoe Stringer * of *key_size* * *count*. The *values* buffer must be of 5880cb80454SJoe Stringer * *value_size* * *count*. 5890cb80454SJoe Stringer * 5900cb80454SJoe Stringer * The *elem_flags* argument may be specified as one of the 5910cb80454SJoe Stringer * following: 5920cb80454SJoe Stringer * 5930cb80454SJoe Stringer * **BPF_F_LOCK** 5940cb80454SJoe Stringer * Look up the value of a spin-locked map without 5950cb80454SJoe Stringer * returning the lock. This must be specified if the 5960cb80454SJoe Stringer * elements contain a spinlock. 5970cb80454SJoe Stringer * 5980cb80454SJoe Stringer * On success, *count* elements from the map are copied into the 5990cb80454SJoe Stringer * user buffer, with the keys copied into *keys* and the values 6000cb80454SJoe Stringer * copied into the corresponding indices in *values*. 6010cb80454SJoe Stringer * 6020cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 6030cb80454SJoe Stringer * is set to the number of successfully processed elements. 6040cb80454SJoe Stringer * 605f67c9cbfSJoe Stringer * Return 606f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 607f67c9cbfSJoe Stringer * is set appropriately. 608f67c9cbfSJoe Stringer * 6090cb80454SJoe Stringer * May set *errno* to **ENOSPC** to indicate that *keys* or 6100cb80454SJoe Stringer * *values* is too small to dump an entire bucket during 6110cb80454SJoe Stringer * iteration of a hash-based map type. 6120cb80454SJoe Stringer * 613f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_AND_DELETE_BATCH 614f67c9cbfSJoe Stringer * Description 6150cb80454SJoe Stringer * Iterate and delete all elements in a map. 6160cb80454SJoe Stringer * 6170cb80454SJoe Stringer * This operation has the same behavior as 6180cb80454SJoe Stringer * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 6190cb80454SJoe Stringer * 6200cb80454SJoe Stringer * * Every element that is successfully returned is also deleted 6210cb80454SJoe Stringer * from the map. This is at least *count* elements. Note that 6220cb80454SJoe Stringer * *count* is both an input and an output parameter. 6230cb80454SJoe Stringer * * Upon returning with *errno* set to **EFAULT**, up to 6240cb80454SJoe Stringer * *count* elements may be deleted without returning the keys 6250cb80454SJoe Stringer * and values of the deleted elements. 626f67c9cbfSJoe Stringer * 627f67c9cbfSJoe Stringer * Return 628f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 629f67c9cbfSJoe Stringer * is set appropriately. 630f67c9cbfSJoe Stringer * 631f67c9cbfSJoe Stringer * BPF_MAP_UPDATE_BATCH 632f67c9cbfSJoe Stringer * Description 6330cb80454SJoe Stringer * Update multiple elements in a map by *key*. 6340cb80454SJoe Stringer * 6350cb80454SJoe Stringer * The *keys* and *values* are input parameters which must point 6360cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 6370cb80454SJoe Stringer * and value size of the map *map_fd*. The *keys* buffer must be 6380cb80454SJoe Stringer * of *key_size* * *count*. The *values* buffer must be of 6390cb80454SJoe Stringer * *value_size* * *count*. 6400cb80454SJoe Stringer * 6410cb80454SJoe Stringer * Each element specified in *keys* is sequentially updated to the 6420cb80454SJoe Stringer * value in the corresponding index in *values*. The *in_batch* 6430cb80454SJoe Stringer * and *out_batch* parameters are ignored and should be zeroed. 6440cb80454SJoe Stringer * 6450cb80454SJoe Stringer * The *elem_flags* argument should be specified as one of the 6460cb80454SJoe Stringer * following: 6470cb80454SJoe Stringer * 6480cb80454SJoe Stringer * **BPF_ANY** 6490cb80454SJoe Stringer * Create new elements or update a existing elements. 6500cb80454SJoe Stringer * **BPF_NOEXIST** 6510cb80454SJoe Stringer * Create new elements only if they do not exist. 6520cb80454SJoe Stringer * **BPF_EXIST** 6530cb80454SJoe Stringer * Update existing elements. 6540cb80454SJoe Stringer * **BPF_F_LOCK** 6550cb80454SJoe Stringer * Update spin_lock-ed map elements. This must be 6560cb80454SJoe Stringer * specified if the map value contains a spinlock. 6570cb80454SJoe Stringer * 6580cb80454SJoe Stringer * On success, *count* elements from the map are updated. 6590cb80454SJoe Stringer * 6600cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 6610cb80454SJoe Stringer * is set to the number of successfully processed elements. 662f67c9cbfSJoe Stringer * 663f67c9cbfSJoe Stringer * Return 664f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 665f67c9cbfSJoe Stringer * is set appropriately. 666f67c9cbfSJoe Stringer * 6670cb80454SJoe Stringer * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 6680cb80454SJoe Stringer * **E2BIG**. **E2BIG** indicates that the number of elements in 6690cb80454SJoe Stringer * the map reached the *max_entries* limit specified at map 6700cb80454SJoe Stringer * creation time. 6710cb80454SJoe Stringer * 6720cb80454SJoe Stringer * May set *errno* to one of the following error codes under 6730cb80454SJoe Stringer * specific circumstances: 6740cb80454SJoe Stringer * 6750cb80454SJoe Stringer * **EEXIST** 6760cb80454SJoe Stringer * If *flags* specifies **BPF_NOEXIST** and the element 6770cb80454SJoe Stringer * with *key* already exists in the map. 6780cb80454SJoe Stringer * **ENOENT** 6790cb80454SJoe Stringer * If *flags* specifies **BPF_EXIST** and the element with 6800cb80454SJoe Stringer * *key* does not exist in the map. 6810cb80454SJoe Stringer * 682f67c9cbfSJoe Stringer * BPF_MAP_DELETE_BATCH 683f67c9cbfSJoe Stringer * Description 6840cb80454SJoe Stringer * Delete multiple elements in a map by *key*. 6850cb80454SJoe Stringer * 6860cb80454SJoe Stringer * The *keys* parameter is an input parameter which must point 6870cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 6880cb80454SJoe Stringer * size of the map *map_fd*, that is, *key_size* * *count*. 6890cb80454SJoe Stringer * 6900cb80454SJoe Stringer * Each element specified in *keys* is sequentially deleted. The 6910cb80454SJoe Stringer * *in_batch*, *out_batch*, and *values* parameters are ignored 6920cb80454SJoe Stringer * and should be zeroed. 6930cb80454SJoe Stringer * 6940cb80454SJoe Stringer * The *elem_flags* argument may be specified as one of the 6950cb80454SJoe Stringer * following: 6960cb80454SJoe Stringer * 6970cb80454SJoe Stringer * **BPF_F_LOCK** 6980cb80454SJoe Stringer * Look up the value of a spin-locked map without 6990cb80454SJoe Stringer * returning the lock. This must be specified if the 7000cb80454SJoe Stringer * elements contain a spinlock. 7010cb80454SJoe Stringer * 7020cb80454SJoe Stringer * On success, *count* elements from the map are updated. 7030cb80454SJoe Stringer * 7040cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 7050cb80454SJoe Stringer * is set to the number of successfully processed elements. If 7060cb80454SJoe Stringer * *errno* is **EFAULT**, up to *count* elements may be been 7070cb80454SJoe Stringer * deleted. 708f67c9cbfSJoe Stringer * 709f67c9cbfSJoe Stringer * Return 710f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 711f67c9cbfSJoe Stringer * is set appropriately. 712f67c9cbfSJoe Stringer * 713f67c9cbfSJoe Stringer * BPF_LINK_CREATE 714f67c9cbfSJoe Stringer * Description 715f67c9cbfSJoe Stringer * Attach an eBPF program to a *target_fd* at the specified 716f67c9cbfSJoe Stringer * *attach_type* hook and return a file descriptor handle for 717f67c9cbfSJoe Stringer * managing the link. 718f67c9cbfSJoe Stringer * 719f67c9cbfSJoe Stringer * Return 720f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 721f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 722f67c9cbfSJoe Stringer * 723f67c9cbfSJoe Stringer * BPF_LINK_UPDATE 724f67c9cbfSJoe Stringer * Description 725f67c9cbfSJoe Stringer * Update the eBPF program in the specified *link_fd* to 726f67c9cbfSJoe Stringer * *new_prog_fd*. 727f67c9cbfSJoe Stringer * 728f67c9cbfSJoe Stringer * Return 729f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 730f67c9cbfSJoe Stringer * is set appropriately. 731f67c9cbfSJoe Stringer * 732f67c9cbfSJoe Stringer * BPF_LINK_GET_FD_BY_ID 733f67c9cbfSJoe Stringer * Description 734f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF Link corresponding to 735f67c9cbfSJoe Stringer * *link_id*. 736f67c9cbfSJoe Stringer * 737f67c9cbfSJoe Stringer * Return 738f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 739f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 740f67c9cbfSJoe Stringer * 741f67c9cbfSJoe Stringer * BPF_LINK_GET_NEXT_ID 742f67c9cbfSJoe Stringer * Description 743f67c9cbfSJoe Stringer * Fetch the next eBPF link currently loaded into the kernel. 744f67c9cbfSJoe Stringer * 745f67c9cbfSJoe Stringer * Looks for the eBPF link with an id greater than *start_id* 746f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF links 747f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 748f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 749f67c9cbfSJoe Stringer * 750f67c9cbfSJoe Stringer * Return 751f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 752f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 753f67c9cbfSJoe Stringer * 754f67c9cbfSJoe Stringer * BPF_ENABLE_STATS 755f67c9cbfSJoe Stringer * Description 756f67c9cbfSJoe Stringer * Enable eBPF runtime statistics gathering. 757f67c9cbfSJoe Stringer * 758f67c9cbfSJoe Stringer * Runtime statistics gathering for the eBPF runtime is disabled 759f67c9cbfSJoe Stringer * by default to minimize the corresponding performance overhead. 760f67c9cbfSJoe Stringer * This command enables statistics globally. 761f67c9cbfSJoe Stringer * 762f67c9cbfSJoe Stringer * Multiple programs may independently enable statistics. 763f67c9cbfSJoe Stringer * After gathering the desired statistics, eBPF runtime statistics 764f67c9cbfSJoe Stringer * may be disabled again by calling **close**\ (2) for the file 765f67c9cbfSJoe Stringer * descriptor returned by this function. Statistics will only be 766f67c9cbfSJoe Stringer * disabled system-wide when all outstanding file descriptors 767f67c9cbfSJoe Stringer * returned by prior calls for this subcommand are closed. 768f67c9cbfSJoe Stringer * 769f67c9cbfSJoe Stringer * Return 770f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 771f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 772f67c9cbfSJoe Stringer * 773f67c9cbfSJoe Stringer * BPF_ITER_CREATE 774f67c9cbfSJoe Stringer * Description 775f67c9cbfSJoe Stringer * Create an iterator on top of the specified *link_fd* (as 776f67c9cbfSJoe Stringer * previously created using **BPF_LINK_CREATE**) and return a 777f67c9cbfSJoe Stringer * file descriptor that can be used to trigger the iteration. 778f67c9cbfSJoe Stringer * 779f67c9cbfSJoe Stringer * If the resulting file descriptor is pinned to the filesystem 780f67c9cbfSJoe Stringer * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 781f67c9cbfSJoe Stringer * for that path will trigger the iterator to read kernel state 782f67c9cbfSJoe Stringer * using the eBPF program attached to *link_fd*. 783f67c9cbfSJoe Stringer * 784f67c9cbfSJoe Stringer * Return 785f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 786f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 787f67c9cbfSJoe Stringer * 788f67c9cbfSJoe Stringer * BPF_LINK_DETACH 789f67c9cbfSJoe Stringer * Description 790f67c9cbfSJoe Stringer * Forcefully detach the specified *link_fd* from its 791f67c9cbfSJoe Stringer * corresponding attachment point. 792f67c9cbfSJoe Stringer * 793f67c9cbfSJoe Stringer * Return 794f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 795f67c9cbfSJoe Stringer * is set appropriately. 796f67c9cbfSJoe Stringer * 797f67c9cbfSJoe Stringer * BPF_PROG_BIND_MAP 798f67c9cbfSJoe Stringer * Description 799f67c9cbfSJoe Stringer * Bind a map to the lifetime of an eBPF program. 800f67c9cbfSJoe Stringer * 801f67c9cbfSJoe Stringer * The map identified by *map_fd* is bound to the program 802f67c9cbfSJoe Stringer * identified by *prog_fd* and only released when *prog_fd* is 803f67c9cbfSJoe Stringer * released. This may be used in cases where metadata should be 804f67c9cbfSJoe Stringer * associated with a program which otherwise does not contain any 805f67c9cbfSJoe Stringer * references to the map (for example, embedded in the eBPF 806f67c9cbfSJoe Stringer * program instructions). 807f67c9cbfSJoe Stringer * 808f67c9cbfSJoe Stringer * Return 809f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 810f67c9cbfSJoe Stringer * is set appropriately. 811f67c9cbfSJoe Stringer * 8127799e4d9SJoe Stringer * NOTES 8137799e4d9SJoe Stringer * eBPF objects (maps and programs) can be shared between processes. 8148aacb3c8SJoe Stringer * 8158aacb3c8SJoe Stringer * * After **fork**\ (2), the child inherits file descriptors 8168aacb3c8SJoe Stringer * referring to the same eBPF objects. 8178aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be transferred over 8188aacb3c8SJoe Stringer * **unix**\ (7) domain sockets. 8198aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be duplicated in the 8208aacb3c8SJoe Stringer * usual way, using **dup**\ (2) and similar calls. 8218aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be pinned to the 8228aacb3c8SJoe Stringer * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 8238aacb3c8SJoe Stringer * 8248aacb3c8SJoe Stringer * An eBPF object is deallocated only after all file descriptors referring 8258aacb3c8SJoe Stringer * to the object have been closed and no references remain pinned to the 8268aacb3c8SJoe Stringer * filesystem or attached (for example, bound to a program or device). 8277799e4d9SJoe Stringer */ 82899c55f7dSAlexei Starovoitov enum bpf_cmd { 82999c55f7dSAlexei Starovoitov BPF_MAP_CREATE, 830db20fd2bSAlexei Starovoitov BPF_MAP_LOOKUP_ELEM, 831db20fd2bSAlexei Starovoitov BPF_MAP_UPDATE_ELEM, 832db20fd2bSAlexei Starovoitov BPF_MAP_DELETE_ELEM, 833db20fd2bSAlexei Starovoitov BPF_MAP_GET_NEXT_KEY, 83409756af4SAlexei Starovoitov BPF_PROG_LOAD, 835b2197755SDaniel Borkmann BPF_OBJ_PIN, 836b2197755SDaniel Borkmann BPF_OBJ_GET, 837f4324551SDaniel Mack BPF_PROG_ATTACH, 838f4324551SDaniel Mack BPF_PROG_DETACH, 8391cf1cae9SAlexei Starovoitov BPF_PROG_TEST_RUN, 84034ad5580SMartin KaFai Lau BPF_PROG_GET_NEXT_ID, 84134ad5580SMartin KaFai Lau BPF_MAP_GET_NEXT_ID, 842b16d9aa4SMartin KaFai Lau BPF_PROG_GET_FD_BY_ID, 843bd5f5f4eSMartin KaFai Lau BPF_MAP_GET_FD_BY_ID, 8441e270976SMartin KaFai Lau BPF_OBJ_GET_INFO_BY_FD, 845468e2f64SAlexei Starovoitov BPF_PROG_QUERY, 846c4f6699dSAlexei Starovoitov BPF_RAW_TRACEPOINT_OPEN, 847f56a653cSMartin KaFai Lau BPF_BTF_LOAD, 84878958fcaSMartin KaFai Lau BPF_BTF_GET_FD_BY_ID, 84941bdc4b4SYonghong Song BPF_TASK_FD_QUERY, 850bd513cd0SMauricio Vasquez B BPF_MAP_LOOKUP_AND_DELETE_ELEM, 85187df15deSDaniel Borkmann BPF_MAP_FREEZE, 8521b9ed84eSQuentin Monnet BPF_BTF_GET_NEXT_ID, 853cb4d03abSBrian Vazquez BPF_MAP_LOOKUP_BATCH, 85405799638SYonghong Song BPF_MAP_LOOKUP_AND_DELETE_BATCH, 855aa2e93b8SBrian Vazquez BPF_MAP_UPDATE_BATCH, 856aa2e93b8SBrian Vazquez BPF_MAP_DELETE_BATCH, 857af6eea57SAndrii Nakryiko BPF_LINK_CREATE, 8580c991ebcSAndrii Nakryiko BPF_LINK_UPDATE, 8592d602c8cSAndrii Nakryiko BPF_LINK_GET_FD_BY_ID, 8602d602c8cSAndrii Nakryiko BPF_LINK_GET_NEXT_ID, 861d46edd67SSong Liu BPF_ENABLE_STATS, 862ac51d99bSYonghong Song BPF_ITER_CREATE, 86373b11c2aSAndrii Nakryiko BPF_LINK_DETACH, 864ef15314aSYiFei Zhu BPF_PROG_BIND_MAP, 86599c55f7dSAlexei Starovoitov }; 86699c55f7dSAlexei Starovoitov 86799c55f7dSAlexei Starovoitov enum bpf_map_type { 86899c55f7dSAlexei Starovoitov BPF_MAP_TYPE_UNSPEC, 8690f8e4bd8SAlexei Starovoitov BPF_MAP_TYPE_HASH, 87028fbcfa0SAlexei Starovoitov BPF_MAP_TYPE_ARRAY, 87104fd61abSAlexei Starovoitov BPF_MAP_TYPE_PROG_ARRAY, 872ea317b26SKaixu Xia BPF_MAP_TYPE_PERF_EVENT_ARRAY, 873824bd0ceSAlexei Starovoitov BPF_MAP_TYPE_PERCPU_HASH, 874a10423b8SAlexei Starovoitov BPF_MAP_TYPE_PERCPU_ARRAY, 875d5a3b1f6SAlexei Starovoitov BPF_MAP_TYPE_STACK_TRACE, 8764ed8ec52SMartin KaFai Lau BPF_MAP_TYPE_CGROUP_ARRAY, 87729ba732aSMartin KaFai Lau BPF_MAP_TYPE_LRU_HASH, 8788f844938SMartin KaFai Lau BPF_MAP_TYPE_LRU_PERCPU_HASH, 879b95a5c4dSDaniel Mack BPF_MAP_TYPE_LPM_TRIE, 88056f668dfSMartin KaFai Lau BPF_MAP_TYPE_ARRAY_OF_MAPS, 881bcc6b1b7SMartin KaFai Lau BPF_MAP_TYPE_HASH_OF_MAPS, 882546ac1ffSJohn Fastabend BPF_MAP_TYPE_DEVMAP, 883174a79ffSJohn Fastabend BPF_MAP_TYPE_SOCKMAP, 8846710e112SJesper Dangaard Brouer BPF_MAP_TYPE_CPUMAP, 885fbfc504aSBjörn Töpel BPF_MAP_TYPE_XSKMAP, 88681110384SJohn Fastabend BPF_MAP_TYPE_SOCKHASH, 887de9cbbaaSRoman Gushchin BPF_MAP_TYPE_CGROUP_STORAGE, 8885dc4c4b7SMartin KaFai Lau BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 889b741f163SRoman Gushchin BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 890f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_QUEUE, 891f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_STACK, 8926ac99e8fSMartin KaFai Lau BPF_MAP_TYPE_SK_STORAGE, 8936f9d451aSToke Høiland-Jørgensen BPF_MAP_TYPE_DEVMAP_HASH, 89485d33df3SMartin KaFai Lau BPF_MAP_TYPE_STRUCT_OPS, 895457f4436SAndrii Nakryiko BPF_MAP_TYPE_RINGBUF, 8968ea63684SKP Singh BPF_MAP_TYPE_INODE_STORAGE, 8974cf1bc1fSKP Singh BPF_MAP_TYPE_TASK_STORAGE, 89899c55f7dSAlexei Starovoitov }; 89999c55f7dSAlexei Starovoitov 9006c4fc209SDaniel Borkmann /* Note that tracing related programs such as 9016c4fc209SDaniel Borkmann * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 9026c4fc209SDaniel Borkmann * are not subject to a stable API since kernel internal data 9036c4fc209SDaniel Borkmann * structures can change from release to release and may 9046c4fc209SDaniel Borkmann * therefore break existing tracing BPF programs. Tracing BPF 9056c4fc209SDaniel Borkmann * programs correspond to /a/ specific kernel which is to be 9066c4fc209SDaniel Borkmann * analyzed, and not /a/ specific kernel /and/ all future ones. 9076c4fc209SDaniel Borkmann */ 90809756af4SAlexei Starovoitov enum bpf_prog_type { 90909756af4SAlexei Starovoitov BPF_PROG_TYPE_UNSPEC, 910ddd872bcSAlexei Starovoitov BPF_PROG_TYPE_SOCKET_FILTER, 9112541517cSAlexei Starovoitov BPF_PROG_TYPE_KPROBE, 91296be4325SDaniel Borkmann BPF_PROG_TYPE_SCHED_CLS, 91394caee8cSDaniel Borkmann BPF_PROG_TYPE_SCHED_ACT, 91498b5c2c6SAlexei Starovoitov BPF_PROG_TYPE_TRACEPOINT, 9156a773a15SBrenden Blanco BPF_PROG_TYPE_XDP, 9160515e599SAlexei Starovoitov BPF_PROG_TYPE_PERF_EVENT, 9170e33661dSDaniel Mack BPF_PROG_TYPE_CGROUP_SKB, 91861023658SDavid Ahern BPF_PROG_TYPE_CGROUP_SOCK, 9193a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_IN, 9203a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_OUT, 9213a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_XMIT, 92240304b2aSLawrence Brakmo BPF_PROG_TYPE_SOCK_OPS, 923b005fd18SJohn Fastabend BPF_PROG_TYPE_SK_SKB, 924ebc614f6SRoman Gushchin BPF_PROG_TYPE_CGROUP_DEVICE, 9254f738adbSJohn Fastabend BPF_PROG_TYPE_SK_MSG, 926c4f6699dSAlexei Starovoitov BPF_PROG_TYPE_RAW_TRACEPOINT, 9274fbac77dSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 928004d4b27SMathieu Xhonneux BPF_PROG_TYPE_LWT_SEG6LOCAL, 929f4364dcfSSean Young BPF_PROG_TYPE_LIRC_MODE2, 9302dbb9b9eSMartin KaFai Lau BPF_PROG_TYPE_SK_REUSEPORT, 931d58e468bSPetar Penkov BPF_PROG_TYPE_FLOW_DISSECTOR, 9327b146cebSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SYSCTL, 9339df1c28bSMatt Mullins BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 9340d01da6aSStanislav Fomichev BPF_PROG_TYPE_CGROUP_SOCKOPT, 935f1b9509cSAlexei Starovoitov BPF_PROG_TYPE_TRACING, 93627ae7997SMartin KaFai Lau BPF_PROG_TYPE_STRUCT_OPS, 937be8704ffSAlexei Starovoitov BPF_PROG_TYPE_EXT, 938fc611f47SKP Singh BPF_PROG_TYPE_LSM, 939e9ddbb77SJakub Sitnicki BPF_PROG_TYPE_SK_LOOKUP, 94009756af4SAlexei Starovoitov }; 94109756af4SAlexei Starovoitov 9420e33661dSDaniel Mack enum bpf_attach_type { 9430e33661dSDaniel Mack BPF_CGROUP_INET_INGRESS, 9440e33661dSDaniel Mack BPF_CGROUP_INET_EGRESS, 94561023658SDavid Ahern BPF_CGROUP_INET_SOCK_CREATE, 94640304b2aSLawrence Brakmo BPF_CGROUP_SOCK_OPS, 947464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_PARSER, 948464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_VERDICT, 949ebc614f6SRoman Gushchin BPF_CGROUP_DEVICE, 9504f738adbSJohn Fastabend BPF_SK_MSG_VERDICT, 9514fbac77dSAndrey Ignatov BPF_CGROUP_INET4_BIND, 9524fbac77dSAndrey Ignatov BPF_CGROUP_INET6_BIND, 953d74bad4eSAndrey Ignatov BPF_CGROUP_INET4_CONNECT, 954d74bad4eSAndrey Ignatov BPF_CGROUP_INET6_CONNECT, 955aac3fc32SAndrey Ignatov BPF_CGROUP_INET4_POST_BIND, 956aac3fc32SAndrey Ignatov BPF_CGROUP_INET6_POST_BIND, 9571cedee13SAndrey Ignatov BPF_CGROUP_UDP4_SENDMSG, 9581cedee13SAndrey Ignatov BPF_CGROUP_UDP6_SENDMSG, 959f4364dcfSSean Young BPF_LIRC_MODE2, 960d58e468bSPetar Penkov BPF_FLOW_DISSECTOR, 9617b146cebSAndrey Ignatov BPF_CGROUP_SYSCTL, 962983695faSDaniel Borkmann BPF_CGROUP_UDP4_RECVMSG, 963983695faSDaniel Borkmann BPF_CGROUP_UDP6_RECVMSG, 9640d01da6aSStanislav Fomichev BPF_CGROUP_GETSOCKOPT, 9650d01da6aSStanislav Fomichev BPF_CGROUP_SETSOCKOPT, 966f1b9509cSAlexei Starovoitov BPF_TRACE_RAW_TP, 967fec56f58SAlexei Starovoitov BPF_TRACE_FENTRY, 968fec56f58SAlexei Starovoitov BPF_TRACE_FEXIT, 969ae240823SKP Singh BPF_MODIFY_RETURN, 970fc611f47SKP Singh BPF_LSM_MAC, 97115d83c4dSYonghong Song BPF_TRACE_ITER, 9721b66d253SDaniel Borkmann BPF_CGROUP_INET4_GETPEERNAME, 9731b66d253SDaniel Borkmann BPF_CGROUP_INET6_GETPEERNAME, 9741b66d253SDaniel Borkmann BPF_CGROUP_INET4_GETSOCKNAME, 9751b66d253SDaniel Borkmann BPF_CGROUP_INET6_GETSOCKNAME, 976fbee97feSDavid Ahern BPF_XDP_DEVMAP, 977f5836749SStanislav Fomichev BPF_CGROUP_INET_SOCK_RELEASE, 97892164774SLorenzo Bianconi BPF_XDP_CPUMAP, 979e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP, 980aa8d3a71SAndrii Nakryiko BPF_XDP, 981a7ba4558SCong Wang BPF_SK_SKB_VERDICT, 9820e33661dSDaniel Mack __MAX_BPF_ATTACH_TYPE 9830e33661dSDaniel Mack }; 9840e33661dSDaniel Mack 9850e33661dSDaniel Mack #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 9860e33661dSDaniel Mack 987f2e10bffSAndrii Nakryiko enum bpf_link_type { 988f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_UNSPEC = 0, 989f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 990f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_TRACING = 2, 991f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_CGROUP = 3, 992de4e05caSYonghong Song BPF_LINK_TYPE_ITER = 4, 9937f045a49SJakub Sitnicki BPF_LINK_TYPE_NETNS = 5, 994aa8d3a71SAndrii Nakryiko BPF_LINK_TYPE_XDP = 6, 995f2e10bffSAndrii Nakryiko 996f2e10bffSAndrii Nakryiko MAX_BPF_LINK_TYPE, 997f2e10bffSAndrii Nakryiko }; 998f2e10bffSAndrii Nakryiko 999324bda9eSAlexei Starovoitov /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1000324bda9eSAlexei Starovoitov * 1001324bda9eSAlexei Starovoitov * NONE(default): No further bpf programs allowed in the subtree. 1002324bda9eSAlexei Starovoitov * 1003324bda9eSAlexei Starovoitov * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1004324bda9eSAlexei Starovoitov * the program in this cgroup yields to sub-cgroup program. 1005324bda9eSAlexei Starovoitov * 1006324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1007324bda9eSAlexei Starovoitov * that cgroup program gets run in addition to the program in this cgroup. 1008324bda9eSAlexei Starovoitov * 1009324bda9eSAlexei Starovoitov * Only one program is allowed to be attached to a cgroup with 1010324bda9eSAlexei Starovoitov * NONE or BPF_F_ALLOW_OVERRIDE flag. 1011324bda9eSAlexei Starovoitov * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1012324bda9eSAlexei Starovoitov * release old program and attach the new one. Attach flags has to match. 1013324bda9eSAlexei Starovoitov * 1014324bda9eSAlexei Starovoitov * Multiple programs are allowed to be attached to a cgroup with 1015324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1016324bda9eSAlexei Starovoitov * (those that were attached first, run first) 1017324bda9eSAlexei Starovoitov * The programs of sub-cgroup are executed first, then programs of 1018324bda9eSAlexei Starovoitov * this cgroup and then programs of parent cgroup. 1019324bda9eSAlexei Starovoitov * When children program makes decision (like picking TCP CA or sock bind) 1020324bda9eSAlexei Starovoitov * parent program has a chance to override it. 1021324bda9eSAlexei Starovoitov * 10227dd68b32SAndrey Ignatov * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 10237dd68b32SAndrey Ignatov * programs for a cgroup. Though it's possible to replace an old program at 10247dd68b32SAndrey Ignatov * any position by also specifying BPF_F_REPLACE flag and position itself in 10257dd68b32SAndrey Ignatov * replace_bpf_fd attribute. Old program at this position will be released. 10267dd68b32SAndrey Ignatov * 1027324bda9eSAlexei Starovoitov * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1028324bda9eSAlexei Starovoitov * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1029324bda9eSAlexei Starovoitov * Ex1: 1030324bda9eSAlexei Starovoitov * cgrp1 (MULTI progs A, B) -> 1031324bda9eSAlexei Starovoitov * cgrp2 (OVERRIDE prog C) -> 1032324bda9eSAlexei Starovoitov * cgrp3 (MULTI prog D) -> 1033324bda9eSAlexei Starovoitov * cgrp4 (OVERRIDE prog E) -> 1034324bda9eSAlexei Starovoitov * cgrp5 (NONE prog F) 1035324bda9eSAlexei Starovoitov * the event in cgrp5 triggers execution of F,D,A,B in that order. 1036324bda9eSAlexei Starovoitov * if prog F is detached, the execution is E,D,A,B 1037324bda9eSAlexei Starovoitov * if prog F and D are detached, the execution is E,A,B 1038324bda9eSAlexei Starovoitov * if prog F, E and D are detached, the execution is C,A,B 1039324bda9eSAlexei Starovoitov * 1040324bda9eSAlexei Starovoitov * All eligible programs are executed regardless of return code from 1041324bda9eSAlexei Starovoitov * earlier programs. 10427f677633SAlexei Starovoitov */ 10437f677633SAlexei Starovoitov #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1044324bda9eSAlexei Starovoitov #define BPF_F_ALLOW_MULTI (1U << 1) 10457dd68b32SAndrey Ignatov #define BPF_F_REPLACE (1U << 2) 10467f677633SAlexei Starovoitov 1047e07b98d9SDavid S. Miller /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1048e07b98d9SDavid S. Miller * verifier will perform strict alignment checking as if the kernel 1049e07b98d9SDavid S. Miller * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1050e07b98d9SDavid S. Miller * and NET_IP_ALIGN defined to 2. 1051e07b98d9SDavid S. Miller */ 1052e07b98d9SDavid S. Miller #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1053e07b98d9SDavid S. Miller 1054e9ee9efcSDavid Miller /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 1055e9ee9efcSDavid Miller * verifier will allow any alignment whatsoever. On platforms 1056e9ee9efcSDavid Miller * with strict alignment requirements for loads ands stores (such 1057e9ee9efcSDavid Miller * as sparc and mips) the verifier validates that all loads and 1058e9ee9efcSDavid Miller * stores provably follow this requirement. This flag turns that 1059e9ee9efcSDavid Miller * checking and enforcement off. 1060e9ee9efcSDavid Miller * 1061e9ee9efcSDavid Miller * It is mostly used for testing when we want to validate the 1062e9ee9efcSDavid Miller * context and memory access aspects of the verifier, but because 1063e9ee9efcSDavid Miller * of an unaligned access the alignment check would trigger before 1064e9ee9efcSDavid Miller * the one we are interested in. 1065e9ee9efcSDavid Miller */ 1066e9ee9efcSDavid Miller #define BPF_F_ANY_ALIGNMENT (1U << 1) 1067e9ee9efcSDavid Miller 1068c240eff6SJiong Wang /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1069c240eff6SJiong Wang * Verifier does sub-register def/use analysis and identifies instructions whose 1070c240eff6SJiong Wang * def only matters for low 32-bit, high 32-bit is never referenced later 1071c240eff6SJiong Wang * through implicit zero extension. Therefore verifier notifies JIT back-ends 1072c240eff6SJiong Wang * that it is safe to ignore clearing high 32-bit for these instructions. This 1073c240eff6SJiong Wang * saves some back-ends a lot of code-gen. However such optimization is not 1074c240eff6SJiong Wang * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1075c240eff6SJiong Wang * hence hasn't used verifier's analysis result. But, we really want to have a 1076c240eff6SJiong Wang * way to be able to verify the correctness of the described optimization on 1077c240eff6SJiong Wang * x86_64 on which testsuites are frequently exercised. 1078c240eff6SJiong Wang * 1079c240eff6SJiong Wang * So, this flag is introduced. Once it is set, verifier will randomize high 1080c240eff6SJiong Wang * 32-bit for those instructions who has been identified as safe to ignore them. 1081c240eff6SJiong Wang * Then, if verifier is not doing correct analysis, such randomization will 1082c240eff6SJiong Wang * regress tests to expose bugs. 1083c240eff6SJiong Wang */ 1084c240eff6SJiong Wang #define BPF_F_TEST_RND_HI32 (1U << 2) 1085c240eff6SJiong Wang 108610d274e8SAlexei Starovoitov /* The verifier internal test flag. Behavior is undefined */ 108710d274e8SAlexei Starovoitov #define BPF_F_TEST_STATE_FREQ (1U << 3) 108810d274e8SAlexei Starovoitov 10891e6c62a8SAlexei Starovoitov /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 10901e6c62a8SAlexei Starovoitov * restrict map and helper usage for such programs. Sleepable BPF programs can 10911e6c62a8SAlexei Starovoitov * only be attached to hooks where kernel execution context allows sleeping. 10921e6c62a8SAlexei Starovoitov * Such programs are allowed to use helpers that may sleep like 10931e6c62a8SAlexei Starovoitov * bpf_copy_from_user(). 10941e6c62a8SAlexei Starovoitov */ 10951e6c62a8SAlexei Starovoitov #define BPF_F_SLEEPABLE (1U << 4) 10961e6c62a8SAlexei Starovoitov 1097d8eca5bbSDaniel Borkmann /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 10984976b718SHao Luo * the following extensions: 1099d8eca5bbSDaniel Borkmann * 11004976b718SHao Luo * insn[0].src_reg: BPF_PSEUDO_MAP_FD 11014976b718SHao Luo * insn[0].imm: map fd 11024976b718SHao Luo * insn[1].imm: 0 11034976b718SHao Luo * insn[0].off: 0 11044976b718SHao Luo * insn[1].off: 0 11054976b718SHao Luo * ldimm64 rewrite: address of map 11064976b718SHao Luo * verifier type: CONST_PTR_TO_MAP 1107d8eca5bbSDaniel Borkmann */ 1108f1a66f85SDaniel Borkmann #define BPF_PSEUDO_MAP_FD 1 11094976b718SHao Luo /* insn[0].src_reg: BPF_PSEUDO_MAP_VALUE 11104976b718SHao Luo * insn[0].imm: map fd 11114976b718SHao Luo * insn[1].imm: offset into value 11124976b718SHao Luo * insn[0].off: 0 11134976b718SHao Luo * insn[1].off: 0 11144976b718SHao Luo * ldimm64 rewrite: address of map[0]+offset 11154976b718SHao Luo * verifier type: PTR_TO_MAP_VALUE 11164976b718SHao Luo */ 1117d8eca5bbSDaniel Borkmann #define BPF_PSEUDO_MAP_VALUE 2 11184976b718SHao Luo /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 11194976b718SHao Luo * insn[0].imm: kernel btd id of VAR 11204976b718SHao Luo * insn[1].imm: 0 11214976b718SHao Luo * insn[0].off: 0 11224976b718SHao Luo * insn[1].off: 0 11234976b718SHao Luo * ldimm64 rewrite: address of the kernel variable 11244976b718SHao Luo * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 11254976b718SHao Luo * is struct/union. 11264976b718SHao Luo */ 11274976b718SHao Luo #define BPF_PSEUDO_BTF_ID 3 112869c087baSYonghong Song /* insn[0].src_reg: BPF_PSEUDO_FUNC 112969c087baSYonghong Song * insn[0].imm: insn offset to the func 113069c087baSYonghong Song * insn[1].imm: 0 113169c087baSYonghong Song * insn[0].off: 0 113269c087baSYonghong Song * insn[1].off: 0 113369c087baSYonghong Song * ldimm64 rewrite: address of the function 113469c087baSYonghong Song * verifier type: PTR_TO_FUNC. 113569c087baSYonghong Song */ 113669c087baSYonghong Song #define BPF_PSEUDO_FUNC 4 1137f1a66f85SDaniel Borkmann 1138cc8b0b92SAlexei Starovoitov /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1139cc8b0b92SAlexei Starovoitov * offset to another bpf function 1140cc8b0b92SAlexei Starovoitov */ 1141cc8b0b92SAlexei Starovoitov #define BPF_PSEUDO_CALL 1 1142e6ac2450SMartin KaFai Lau /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1143e6ac2450SMartin KaFai Lau * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1144e6ac2450SMartin KaFai Lau */ 1145e6ac2450SMartin KaFai Lau #define BPF_PSEUDO_KFUNC_CALL 2 1146cc8b0b92SAlexei Starovoitov 11473274f520SAlexei Starovoitov /* flags for BPF_MAP_UPDATE_ELEM command */ 11481aae4bddSAndrii Nakryiko enum { 11491aae4bddSAndrii Nakryiko BPF_ANY = 0, /* create new element or update existing */ 11501aae4bddSAndrii Nakryiko BPF_NOEXIST = 1, /* create new element if it didn't exist */ 11511aae4bddSAndrii Nakryiko BPF_EXIST = 2, /* update existing element */ 11521aae4bddSAndrii Nakryiko BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 11531aae4bddSAndrii Nakryiko }; 11543274f520SAlexei Starovoitov 115596eabe7aSMartin KaFai Lau /* flags for BPF_MAP_CREATE command */ 11561aae4bddSAndrii Nakryiko enum { 11571aae4bddSAndrii Nakryiko BPF_F_NO_PREALLOC = (1U << 0), 115829ba732aSMartin KaFai Lau /* Instead of having one common LRU list in the 11598f844938SMartin KaFai Lau * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 116029ba732aSMartin KaFai Lau * which can scale and perform better. 116129ba732aSMartin KaFai Lau * Note, the LRU nodes (including free nodes) cannot be moved 116229ba732aSMartin KaFai Lau * across different LRU lists. 116329ba732aSMartin KaFai Lau */ 11641aae4bddSAndrii Nakryiko BPF_F_NO_COMMON_LRU = (1U << 1), 116596eabe7aSMartin KaFai Lau /* Specify numa node during map creation */ 11661aae4bddSAndrii Nakryiko BPF_F_NUMA_NODE = (1U << 2), 1167cb4d2b3fSMartin KaFai Lau 1168591fe988SDaniel Borkmann /* Flags for accessing BPF object from syscall side. */ 11691aae4bddSAndrii Nakryiko BPF_F_RDONLY = (1U << 3), 11701aae4bddSAndrii Nakryiko BPF_F_WRONLY = (1U << 4), 11716e71b04aSChenbo Feng 1172615755a7SSong Liu /* Flag for stack_map, store build_id+offset instead of pointer */ 11731aae4bddSAndrii Nakryiko BPF_F_STACK_BUILD_ID = (1U << 5), 1174615755a7SSong Liu 117596b3b6c9SLorenz Bauer /* Zero-initialize hash function seed. This should only be used for testing. */ 11761aae4bddSAndrii Nakryiko BPF_F_ZERO_SEED = (1U << 6), 117796b3b6c9SLorenz Bauer 1178591fe988SDaniel Borkmann /* Flags for accessing BPF object from program side. */ 11791aae4bddSAndrii Nakryiko BPF_F_RDONLY_PROG = (1U << 7), 11801aae4bddSAndrii Nakryiko BPF_F_WRONLY_PROG = (1U << 8), 1181591fe988SDaniel Borkmann 11828f51dfc7SStanislav Fomichev /* Clone map from listener for newly accepted socket */ 11831aae4bddSAndrii Nakryiko BPF_F_CLONE = (1U << 9), 11848f51dfc7SStanislav Fomichev 1185fc970227SAndrii Nakryiko /* Enable memory-mapping BPF map */ 11861aae4bddSAndrii Nakryiko BPF_F_MMAPABLE = (1U << 10), 1187792cacccSSong Liu 1188792cacccSSong Liu /* Share perf_event among processes */ 1189792cacccSSong Liu BPF_F_PRESERVE_ELEMS = (1U << 11), 11904a8f87e6SDaniel Borkmann 11914a8f87e6SDaniel Borkmann /* Create a map that is suitable to be an inner map with dynamic max entries */ 11924a8f87e6SDaniel Borkmann BPF_F_INNER_MAP = (1U << 12), 11931aae4bddSAndrii Nakryiko }; 1194fc970227SAndrii Nakryiko 1195f5bfcd95SAndrey Ignatov /* Flags for BPF_PROG_QUERY. */ 1196f5bfcd95SAndrey Ignatov 1197f5bfcd95SAndrey Ignatov /* Query effective (directly attached + inherited from ancestor cgroups) 1198f5bfcd95SAndrey Ignatov * programs that will be executed for events within a cgroup. 1199f5bfcd95SAndrey Ignatov * attach_flags with this flag are returned only for directly attached programs. 1200f5bfcd95SAndrey Ignatov */ 12012f183360SLorenz Bauer #define BPF_F_QUERY_EFFECTIVE (1U << 0) 12022f183360SLorenz Bauer 12031b4d60ecSSong Liu /* Flags for BPF_PROG_TEST_RUN */ 12041b4d60ecSSong Liu 12051b4d60ecSSong Liu /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 12061b4d60ecSSong Liu #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 12071b4d60ecSSong Liu 1208d46edd67SSong Liu /* type for BPF_ENABLE_STATS */ 1209d46edd67SSong Liu enum bpf_stats_type { 1210d46edd67SSong Liu /* enabled run_time_ns and run_cnt */ 1211d46edd67SSong Liu BPF_STATS_RUN_TIME = 0, 1212d46edd67SSong Liu }; 1213d46edd67SSong Liu 1214615755a7SSong Liu enum bpf_stack_build_id_status { 1215615755a7SSong Liu /* user space need an empty entry to identify end of a trace */ 1216615755a7SSong Liu BPF_STACK_BUILD_ID_EMPTY = 0, 1217615755a7SSong Liu /* with valid build_id and offset */ 1218615755a7SSong Liu BPF_STACK_BUILD_ID_VALID = 1, 1219615755a7SSong Liu /* couldn't get build_id, fallback to ip */ 1220615755a7SSong Liu BPF_STACK_BUILD_ID_IP = 2, 1221615755a7SSong Liu }; 1222615755a7SSong Liu 1223615755a7SSong Liu #define BPF_BUILD_ID_SIZE 20 1224615755a7SSong Liu struct bpf_stack_build_id { 1225615755a7SSong Liu __s32 status; 1226615755a7SSong Liu unsigned char build_id[BPF_BUILD_ID_SIZE]; 1227615755a7SSong Liu union { 1228615755a7SSong Liu __u64 offset; 1229615755a7SSong Liu __u64 ip; 1230615755a7SSong Liu }; 1231615755a7SSong Liu }; 1232615755a7SSong Liu 12331aae4bddSAndrii Nakryiko #define BPF_OBJ_NAME_LEN 16U 12341aae4bddSAndrii Nakryiko 123599c55f7dSAlexei Starovoitov union bpf_attr { 123699c55f7dSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_CREATE command */ 123799c55f7dSAlexei Starovoitov __u32 map_type; /* one of enum bpf_map_type */ 123899c55f7dSAlexei Starovoitov __u32 key_size; /* size of key in bytes */ 123999c55f7dSAlexei Starovoitov __u32 value_size; /* size of value in bytes */ 124099c55f7dSAlexei Starovoitov __u32 max_entries; /* max number of entries in a map */ 124196eabe7aSMartin KaFai Lau __u32 map_flags; /* BPF_MAP_CREATE related 124296eabe7aSMartin KaFai Lau * flags defined above. 124396eabe7aSMartin KaFai Lau */ 124456f668dfSMartin KaFai Lau __u32 inner_map_fd; /* fd pointing to the inner map */ 124596eabe7aSMartin KaFai Lau __u32 numa_node; /* numa node (effective only if 124696eabe7aSMartin KaFai Lau * BPF_F_NUMA_NODE is set). 124796eabe7aSMartin KaFai Lau */ 1248067cae47SMartin KaFai Lau char map_name[BPF_OBJ_NAME_LEN]; 1249a3884572SJakub Kicinski __u32 map_ifindex; /* ifindex of netdev to create on */ 1250a26ca7c9SMartin KaFai Lau __u32 btf_fd; /* fd pointing to a BTF type data */ 12519b2cf328SMartin KaFai Lau __u32 btf_key_type_id; /* BTF type_id of the key */ 12529b2cf328SMartin KaFai Lau __u32 btf_value_type_id; /* BTF type_id of the value */ 125385d33df3SMartin KaFai Lau __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 125485d33df3SMartin KaFai Lau * struct stored as the 125585d33df3SMartin KaFai Lau * map value 125685d33df3SMartin KaFai Lau */ 125799c55f7dSAlexei Starovoitov }; 1258db20fd2bSAlexei Starovoitov 1259db20fd2bSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1260db20fd2bSAlexei Starovoitov __u32 map_fd; 1261db20fd2bSAlexei Starovoitov __aligned_u64 key; 1262db20fd2bSAlexei Starovoitov union { 1263db20fd2bSAlexei Starovoitov __aligned_u64 value; 1264db20fd2bSAlexei Starovoitov __aligned_u64 next_key; 1265db20fd2bSAlexei Starovoitov }; 12663274f520SAlexei Starovoitov __u64 flags; 1267db20fd2bSAlexei Starovoitov }; 126809756af4SAlexei Starovoitov 1269cb4d03abSBrian Vazquez struct { /* struct used by BPF_MAP_*_BATCH commands */ 1270cb4d03abSBrian Vazquez __aligned_u64 in_batch; /* start batch, 1271cb4d03abSBrian Vazquez * NULL to start from beginning 1272cb4d03abSBrian Vazquez */ 1273cb4d03abSBrian Vazquez __aligned_u64 out_batch; /* output: next start batch */ 1274cb4d03abSBrian Vazquez __aligned_u64 keys; 1275cb4d03abSBrian Vazquez __aligned_u64 values; 1276cb4d03abSBrian Vazquez __u32 count; /* input/output: 1277cb4d03abSBrian Vazquez * input: # of key/value 1278cb4d03abSBrian Vazquez * elements 1279cb4d03abSBrian Vazquez * output: # of filled elements 1280cb4d03abSBrian Vazquez */ 1281cb4d03abSBrian Vazquez __u32 map_fd; 1282cb4d03abSBrian Vazquez __u64 elem_flags; 1283cb4d03abSBrian Vazquez __u64 flags; 1284cb4d03abSBrian Vazquez } batch; 1285cb4d03abSBrian Vazquez 128609756af4SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_LOAD command */ 128709756af4SAlexei Starovoitov __u32 prog_type; /* one of enum bpf_prog_type */ 128809756af4SAlexei Starovoitov __u32 insn_cnt; 128909756af4SAlexei Starovoitov __aligned_u64 insns; 129009756af4SAlexei Starovoitov __aligned_u64 license; 1291cbd35700SAlexei Starovoitov __u32 log_level; /* verbosity level of verifier */ 1292cbd35700SAlexei Starovoitov __u32 log_size; /* size of user buffer */ 1293cbd35700SAlexei Starovoitov __aligned_u64 log_buf; /* user supplied buffer */ 12946c4fc209SDaniel Borkmann __u32 kern_version; /* not used */ 1295e07b98d9SDavid S. Miller __u32 prog_flags; 1296067cae47SMartin KaFai Lau char prog_name[BPF_OBJ_NAME_LEN]; 12971f6f4cb7SJakub Kicinski __u32 prog_ifindex; /* ifindex of netdev to prep for */ 12985e43f899SAndrey Ignatov /* For some prog types expected attach type must be known at 12995e43f899SAndrey Ignatov * load time to verify attach type specific parts of prog 13005e43f899SAndrey Ignatov * (context accesses, allowed helpers, etc). 13015e43f899SAndrey Ignatov */ 13025e43f899SAndrey Ignatov __u32 expected_attach_type; 1303838e9690SYonghong Song __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1304838e9690SYonghong Song __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1305838e9690SYonghong Song __aligned_u64 func_info; /* func info */ 1306838e9690SYonghong Song __u32 func_info_cnt; /* number of bpf_func_info records */ 1307c454a46bSMartin KaFai Lau __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1308c454a46bSMartin KaFai Lau __aligned_u64 line_info; /* line info */ 1309c454a46bSMartin KaFai Lau __u32 line_info_cnt; /* number of bpf_line_info records */ 1310ccfe29ebSAlexei Starovoitov __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1311290248a5SAndrii Nakryiko union { 1312290248a5SAndrii Nakryiko /* valid prog_fd to attach to bpf prog */ 1313290248a5SAndrii Nakryiko __u32 attach_prog_fd; 1314290248a5SAndrii Nakryiko /* or valid module BTF object fd or 0 to attach to vmlinux */ 1315290248a5SAndrii Nakryiko __u32 attach_btf_obj_fd; 1316290248a5SAndrii Nakryiko }; 131709756af4SAlexei Starovoitov }; 1318b2197755SDaniel Borkmann 1319b2197755SDaniel Borkmann struct { /* anonymous struct used by BPF_OBJ_* commands */ 1320b2197755SDaniel Borkmann __aligned_u64 pathname; 1321b2197755SDaniel Borkmann __u32 bpf_fd; 13226e71b04aSChenbo Feng __u32 file_flags; 1323b2197755SDaniel Borkmann }; 1324f4324551SDaniel Mack 1325f4324551SDaniel Mack struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1326f4324551SDaniel Mack __u32 target_fd; /* container object to attach to */ 1327f4324551SDaniel Mack __u32 attach_bpf_fd; /* eBPF program to attach */ 1328f4324551SDaniel Mack __u32 attach_type; 13297f677633SAlexei Starovoitov __u32 attach_flags; 13307dd68b32SAndrey Ignatov __u32 replace_bpf_fd; /* previously attached eBPF 13317dd68b32SAndrey Ignatov * program to replace if 13327dd68b32SAndrey Ignatov * BPF_F_REPLACE is used 13337dd68b32SAndrey Ignatov */ 1334f4324551SDaniel Mack }; 13351cf1cae9SAlexei Starovoitov 13361cf1cae9SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 13371cf1cae9SAlexei Starovoitov __u32 prog_fd; 13381cf1cae9SAlexei Starovoitov __u32 retval; 1339b5a36b1eSLorenz Bauer __u32 data_size_in; /* input: len of data_in */ 1340b5a36b1eSLorenz Bauer __u32 data_size_out; /* input/output: len of data_out 1341b5a36b1eSLorenz Bauer * returns ENOSPC if data_out 1342b5a36b1eSLorenz Bauer * is too small. 1343b5a36b1eSLorenz Bauer */ 13441cf1cae9SAlexei Starovoitov __aligned_u64 data_in; 13451cf1cae9SAlexei Starovoitov __aligned_u64 data_out; 13461cf1cae9SAlexei Starovoitov __u32 repeat; 13471cf1cae9SAlexei Starovoitov __u32 duration; 1348b0b9395dSStanislav Fomichev __u32 ctx_size_in; /* input: len of ctx_in */ 1349b0b9395dSStanislav Fomichev __u32 ctx_size_out; /* input/output: len of ctx_out 1350b0b9395dSStanislav Fomichev * returns ENOSPC if ctx_out 1351b0b9395dSStanislav Fomichev * is too small. 1352b0b9395dSStanislav Fomichev */ 1353b0b9395dSStanislav Fomichev __aligned_u64 ctx_in; 1354b0b9395dSStanislav Fomichev __aligned_u64 ctx_out; 13551b4d60ecSSong Liu __u32 flags; 13561b4d60ecSSong Liu __u32 cpu; 13571cf1cae9SAlexei Starovoitov } test; 135834ad5580SMartin KaFai Lau 1359b16d9aa4SMartin KaFai Lau struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1360b16d9aa4SMartin KaFai Lau union { 136134ad5580SMartin KaFai Lau __u32 start_id; 1362b16d9aa4SMartin KaFai Lau __u32 prog_id; 1363bd5f5f4eSMartin KaFai Lau __u32 map_id; 136478958fcaSMartin KaFai Lau __u32 btf_id; 1365a3b80e10SAndrii Nakryiko __u32 link_id; 1366b16d9aa4SMartin KaFai Lau }; 136734ad5580SMartin KaFai Lau __u32 next_id; 13686e71b04aSChenbo Feng __u32 open_flags; 136934ad5580SMartin KaFai Lau }; 13701e270976SMartin KaFai Lau 13711e270976SMartin KaFai Lau struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 13721e270976SMartin KaFai Lau __u32 bpf_fd; 13731e270976SMartin KaFai Lau __u32 info_len; 13741e270976SMartin KaFai Lau __aligned_u64 info; 13751e270976SMartin KaFai Lau } info; 1376468e2f64SAlexei Starovoitov 1377468e2f64SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1378468e2f64SAlexei Starovoitov __u32 target_fd; /* container object to query */ 1379468e2f64SAlexei Starovoitov __u32 attach_type; 1380468e2f64SAlexei Starovoitov __u32 query_flags; 1381468e2f64SAlexei Starovoitov __u32 attach_flags; 1382468e2f64SAlexei Starovoitov __aligned_u64 prog_ids; 1383468e2f64SAlexei Starovoitov __u32 prog_cnt; 1384468e2f64SAlexei Starovoitov } query; 1385c4f6699dSAlexei Starovoitov 1386af6eea57SAndrii Nakryiko struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1387c4f6699dSAlexei Starovoitov __u64 name; 1388c4f6699dSAlexei Starovoitov __u32 prog_fd; 1389c4f6699dSAlexei Starovoitov } raw_tracepoint; 1390f56a653cSMartin KaFai Lau 1391f56a653cSMartin KaFai Lau struct { /* anonymous struct for BPF_BTF_LOAD */ 1392f56a653cSMartin KaFai Lau __aligned_u64 btf; 1393f56a653cSMartin KaFai Lau __aligned_u64 btf_log_buf; 1394f56a653cSMartin KaFai Lau __u32 btf_size; 1395f56a653cSMartin KaFai Lau __u32 btf_log_size; 1396f56a653cSMartin KaFai Lau __u32 btf_log_level; 1397f56a653cSMartin KaFai Lau }; 139841bdc4b4SYonghong Song 139941bdc4b4SYonghong Song struct { 140041bdc4b4SYonghong Song __u32 pid; /* input: pid */ 140141bdc4b4SYonghong Song __u32 fd; /* input: fd */ 140241bdc4b4SYonghong Song __u32 flags; /* input: flags */ 140341bdc4b4SYonghong Song __u32 buf_len; /* input/output: buf len */ 140441bdc4b4SYonghong Song __aligned_u64 buf; /* input/output: 140541bdc4b4SYonghong Song * tp_name for tracepoint 140641bdc4b4SYonghong Song * symbol for kprobe 140741bdc4b4SYonghong Song * filename for uprobe 140841bdc4b4SYonghong Song */ 140941bdc4b4SYonghong Song __u32 prog_id; /* output: prod_id */ 141041bdc4b4SYonghong Song __u32 fd_type; /* output: BPF_FD_TYPE_* */ 141141bdc4b4SYonghong Song __u64 probe_offset; /* output: probe_offset */ 141241bdc4b4SYonghong Song __u64 probe_addr; /* output: probe_addr */ 141341bdc4b4SYonghong Song } task_fd_query; 1414af6eea57SAndrii Nakryiko 1415af6eea57SAndrii Nakryiko struct { /* struct used by BPF_LINK_CREATE command */ 1416af6eea57SAndrii Nakryiko __u32 prog_fd; /* eBPF program to attach */ 1417aa8d3a71SAndrii Nakryiko union { 1418af6eea57SAndrii Nakryiko __u32 target_fd; /* object to attach to */ 1419aa8d3a71SAndrii Nakryiko __u32 target_ifindex; /* target ifindex */ 1420aa8d3a71SAndrii Nakryiko }; 1421af6eea57SAndrii Nakryiko __u32 attach_type; /* attach type */ 1422af6eea57SAndrii Nakryiko __u32 flags; /* extra flags */ 14234a1e7c0cSToke Høiland-Jørgensen union { 14244a1e7c0cSToke Høiland-Jørgensen __u32 target_btf_id; /* btf_id of target to attach to */ 14254a1e7c0cSToke Høiland-Jørgensen struct { 14265e7b3020SYonghong Song __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 14275e7b3020SYonghong Song __u32 iter_info_len; /* iter_info length */ 14284a1e7c0cSToke Høiland-Jørgensen }; 14294a1e7c0cSToke Høiland-Jørgensen }; 1430af6eea57SAndrii Nakryiko } link_create; 14310c991ebcSAndrii Nakryiko 14320c991ebcSAndrii Nakryiko struct { /* struct used by BPF_LINK_UPDATE command */ 14330c991ebcSAndrii Nakryiko __u32 link_fd; /* link fd */ 14340c991ebcSAndrii Nakryiko /* new program fd to update link with */ 14350c991ebcSAndrii Nakryiko __u32 new_prog_fd; 14360c991ebcSAndrii Nakryiko __u32 flags; /* extra flags */ 14370c991ebcSAndrii Nakryiko /* expected link's program fd; is specified only if 14380c991ebcSAndrii Nakryiko * BPF_F_REPLACE flag is set in flags */ 14390c991ebcSAndrii Nakryiko __u32 old_prog_fd; 14400c991ebcSAndrii Nakryiko } link_update; 14410c991ebcSAndrii Nakryiko 144273b11c2aSAndrii Nakryiko struct { 144373b11c2aSAndrii Nakryiko __u32 link_fd; 144473b11c2aSAndrii Nakryiko } link_detach; 144573b11c2aSAndrii Nakryiko 1446d46edd67SSong Liu struct { /* struct used by BPF_ENABLE_STATS command */ 1447d46edd67SSong Liu __u32 type; 1448d46edd67SSong Liu } enable_stats; 1449d46edd67SSong Liu 1450ac51d99bSYonghong Song struct { /* struct used by BPF_ITER_CREATE command */ 1451ac51d99bSYonghong Song __u32 link_fd; 1452ac51d99bSYonghong Song __u32 flags; 1453ac51d99bSYonghong Song } iter_create; 1454ac51d99bSYonghong Song 1455ef15314aSYiFei Zhu struct { /* struct used by BPF_PROG_BIND_MAP command */ 1456ef15314aSYiFei Zhu __u32 prog_fd; 1457ef15314aSYiFei Zhu __u32 map_fd; 1458ef15314aSYiFei Zhu __u32 flags; /* extra flags */ 1459ef15314aSYiFei Zhu } prog_bind_map; 1460ef15314aSYiFei Zhu 146199c55f7dSAlexei Starovoitov } __attribute__((aligned(8))); 146299c55f7dSAlexei Starovoitov 146356a092c8SQuentin Monnet /* The description below is an attempt at providing documentation to eBPF 146456a092c8SQuentin Monnet * developers about the multiple available eBPF helper functions. It can be 146556a092c8SQuentin Monnet * parsed and used to produce a manual page. The workflow is the following, 146656a092c8SQuentin Monnet * and requires the rst2man utility: 1467ebb676daSThomas Graf * 1468923a932cSJoe Stringer * $ ./scripts/bpf_doc.py \ 146956a092c8SQuentin Monnet * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 147056a092c8SQuentin Monnet * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 147156a092c8SQuentin Monnet * $ man /tmp/bpf-helpers.7 1472ebb676daSThomas Graf * 147356a092c8SQuentin Monnet * Note that in order to produce this external documentation, some RST 147456a092c8SQuentin Monnet * formatting is used in the descriptions to get "bold" and "italics" in 147556a092c8SQuentin Monnet * manual pages. Also note that the few trailing white spaces are 147656a092c8SQuentin Monnet * intentional, removing them would break paragraphs for rst2man. 1477ebb676daSThomas Graf * 147856a092c8SQuentin Monnet * Start of BPF helper function descriptions: 1479ad4a5223SQuentin Monnet * 1480ad4a5223SQuentin Monnet * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1481ad4a5223SQuentin Monnet * Description 1482ad4a5223SQuentin Monnet * Perform a lookup in *map* for an entry associated to *key*. 1483ad4a5223SQuentin Monnet * Return 1484ad4a5223SQuentin Monnet * Map value associated to *key*, or **NULL** if no entry was 1485ad4a5223SQuentin Monnet * found. 1486ad4a5223SQuentin Monnet * 1487bdb7b79bSAndrii Nakryiko * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1488ad4a5223SQuentin Monnet * Description 1489ad4a5223SQuentin Monnet * Add or update the value of the entry associated to *key* in 1490ad4a5223SQuentin Monnet * *map* with *value*. *flags* is one of: 1491ad4a5223SQuentin Monnet * 1492ad4a5223SQuentin Monnet * **BPF_NOEXIST** 1493ad4a5223SQuentin Monnet * The entry for *key* must not exist in the map. 1494ad4a5223SQuentin Monnet * **BPF_EXIST** 1495ad4a5223SQuentin Monnet * The entry for *key* must already exist in the map. 1496ad4a5223SQuentin Monnet * **BPF_ANY** 1497ad4a5223SQuentin Monnet * No condition on the existence of the entry for *key*. 1498ad4a5223SQuentin Monnet * 1499ad4a5223SQuentin Monnet * Flag value **BPF_NOEXIST** cannot be used for maps of types 1500ad4a5223SQuentin Monnet * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1501ad4a5223SQuentin Monnet * elements always exist), the helper would return an error. 1502ad4a5223SQuentin Monnet * Return 1503ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1504ad4a5223SQuentin Monnet * 1505bdb7b79bSAndrii Nakryiko * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1506ad4a5223SQuentin Monnet * Description 1507ad4a5223SQuentin Monnet * Delete entry with *key* from *map*. 1508ad4a5223SQuentin Monnet * Return 1509ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1510ad4a5223SQuentin Monnet * 1511bdb7b79bSAndrii Nakryiko * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1512ad4a5223SQuentin Monnet * Description 1513ad4a5223SQuentin Monnet * For tracing programs, safely attempt to read *size* bytes from 15146ae08ae3SDaniel Borkmann * kernel space address *unsafe_ptr* and store the data in *dst*. 15156ae08ae3SDaniel Borkmann * 1516ab8d7809SQuentin Monnet * Generally, use **bpf_probe_read_user**\ () or 1517ab8d7809SQuentin Monnet * **bpf_probe_read_kernel**\ () instead. 1518ad4a5223SQuentin Monnet * Return 1519ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1520ad4a5223SQuentin Monnet * 1521ad4a5223SQuentin Monnet * u64 bpf_ktime_get_ns(void) 1522ad4a5223SQuentin Monnet * Description 1523ad4a5223SQuentin Monnet * Return the time elapsed since system boot, in nanoseconds. 152471d19214SMaciej Żenczykowski * Does not include time the system was suspended. 1525ab8d7809SQuentin Monnet * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1526ad4a5223SQuentin Monnet * Return 1527ad4a5223SQuentin Monnet * Current *ktime*. 1528ad4a5223SQuentin Monnet * 1529bdb7b79bSAndrii Nakryiko * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1530ad4a5223SQuentin Monnet * Description 1531ad4a5223SQuentin Monnet * This helper is a "printk()-like" facility for debugging. It 1532ad4a5223SQuentin Monnet * prints a message defined by format *fmt* (of size *fmt_size*) 1533ad4a5223SQuentin Monnet * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 1534ad4a5223SQuentin Monnet * available. It can take up to three additional **u64** 1535ad4a5223SQuentin Monnet * arguments (as an eBPF helpers, the total number of arguments is 1536ad4a5223SQuentin Monnet * limited to five). 1537ad4a5223SQuentin Monnet * 1538ad4a5223SQuentin Monnet * Each time the helper is called, it appends a line to the trace. 153955c33dfbSPeter Wu * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 154055c33dfbSPeter Wu * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 1541ad4a5223SQuentin Monnet * The format of the trace is customizable, and the exact output 1542ad4a5223SQuentin Monnet * one will get depends on the options set in 1543ad4a5223SQuentin Monnet * *\/sys/kernel/debug/tracing/trace_options* (see also the 1544ad4a5223SQuentin Monnet * *README* file under the same directory). However, it usually 1545ad4a5223SQuentin Monnet * defaults to something like: 1546ad4a5223SQuentin Monnet * 1547ad4a5223SQuentin Monnet * :: 1548ad4a5223SQuentin Monnet * 1549ad4a5223SQuentin Monnet * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1550ad4a5223SQuentin Monnet * 1551ad4a5223SQuentin Monnet * In the above: 1552ad4a5223SQuentin Monnet * 1553ad4a5223SQuentin Monnet * * ``telnet`` is the name of the current task. 1554ad4a5223SQuentin Monnet * * ``470`` is the PID of the current task. 1555ad4a5223SQuentin Monnet * * ``001`` is the CPU number on which the task is 1556ad4a5223SQuentin Monnet * running. 1557ad4a5223SQuentin Monnet * * In ``.N..``, each character refers to a set of 1558ad4a5223SQuentin Monnet * options (whether irqs are enabled, scheduling 1559ad4a5223SQuentin Monnet * options, whether hard/softirqs are running, level of 1560ad4a5223SQuentin Monnet * preempt_disabled respectively). **N** means that 1561ad4a5223SQuentin Monnet * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1562ad4a5223SQuentin Monnet * are set. 1563ad4a5223SQuentin Monnet * * ``419421.045894`` is a timestamp. 1564ad4a5223SQuentin Monnet * * ``0x00000001`` is a fake value used by BPF for the 1565ad4a5223SQuentin Monnet * instruction pointer register. 1566ad4a5223SQuentin Monnet * * ``<formatted msg>`` is the message formatted with 1567ad4a5223SQuentin Monnet * *fmt*. 1568ad4a5223SQuentin Monnet * 1569ad4a5223SQuentin Monnet * The conversion specifiers supported by *fmt* are similar, but 1570ad4a5223SQuentin Monnet * more limited than for printk(). They are **%d**, **%i**, 1571ad4a5223SQuentin Monnet * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1572ad4a5223SQuentin Monnet * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1573ad4a5223SQuentin Monnet * of field, padding with zeroes, etc.) is available, and the 1574ad4a5223SQuentin Monnet * helper will return **-EINVAL** (but print nothing) if it 1575ad4a5223SQuentin Monnet * encounters an unknown specifier. 1576ad4a5223SQuentin Monnet * 1577ad4a5223SQuentin Monnet * Also, note that **bpf_trace_printk**\ () is slow, and should 1578ad4a5223SQuentin Monnet * only be used for debugging purposes. For this reason, a notice 1579b16fc097STobias Klauser * block (spanning several lines) is printed to kernel logs and 1580ad4a5223SQuentin Monnet * states that the helper should not be used "for production use" 1581ad4a5223SQuentin Monnet * the first time this helper is used (or more precisely, when 1582ad4a5223SQuentin Monnet * **trace_printk**\ () buffers are allocated). For passing values 1583ad4a5223SQuentin Monnet * to user space, perf events should be preferred. 1584ad4a5223SQuentin Monnet * Return 1585ad4a5223SQuentin Monnet * The number of bytes written to the buffer, or a negative error 1586ad4a5223SQuentin Monnet * in case of failure. 1587ad4a5223SQuentin Monnet * 15881fdd08beSQuentin Monnet * u32 bpf_get_prandom_u32(void) 15891fdd08beSQuentin Monnet * Description 15901fdd08beSQuentin Monnet * Get a pseudo-random number. 15911fdd08beSQuentin Monnet * 15921fdd08beSQuentin Monnet * From a security point of view, this helper uses its own 15931fdd08beSQuentin Monnet * pseudo-random internal state, and cannot be used to infer the 15941fdd08beSQuentin Monnet * seed of other random functions in the kernel. However, it is 15951fdd08beSQuentin Monnet * essential to note that the generator used by the helper is not 15961fdd08beSQuentin Monnet * cryptographically secure. 15971fdd08beSQuentin Monnet * Return 15981fdd08beSQuentin Monnet * A random 32-bit unsigned value. 15991fdd08beSQuentin Monnet * 16001fdd08beSQuentin Monnet * u32 bpf_get_smp_processor_id(void) 16011fdd08beSQuentin Monnet * Description 16021fdd08beSQuentin Monnet * Get the SMP (symmetric multiprocessing) processor id. Note that 16031fdd08beSQuentin Monnet * all programs run with preemption disabled, which means that the 16041fdd08beSQuentin Monnet * SMP processor id is stable during all the execution of the 16051fdd08beSQuentin Monnet * program. 16061fdd08beSQuentin Monnet * Return 16071fdd08beSQuentin Monnet * The SMP id of the processor running the program. 16081fdd08beSQuentin Monnet * 1609bdb7b79bSAndrii Nakryiko * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1610ad4a5223SQuentin Monnet * Description 1611ad4a5223SQuentin Monnet * Store *len* bytes from address *from* into the packet 1612ad4a5223SQuentin Monnet * associated to *skb*, at *offset*. *flags* are a combination of 1613ad4a5223SQuentin Monnet * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1614ad4a5223SQuentin Monnet * checksum for the packet after storing the bytes) and 1615ad4a5223SQuentin Monnet * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1616ad4a5223SQuentin Monnet * **->swhash** and *skb*\ **->l4hash** to 0). 1617ad4a5223SQuentin Monnet * 161832e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1619ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1620ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1621ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1622ad4a5223SQuentin Monnet * direct packet access. 1623ad4a5223SQuentin Monnet * Return 1624ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1625ad4a5223SQuentin Monnet * 1626bdb7b79bSAndrii Nakryiko * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1627ad4a5223SQuentin Monnet * Description 1628ad4a5223SQuentin Monnet * Recompute the layer 3 (e.g. IP) checksum for the packet 1629ad4a5223SQuentin Monnet * associated to *skb*. Computation is incremental, so the helper 1630ad4a5223SQuentin Monnet * must know the former value of the header field that was 1631ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 1632ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored in *size*. 1633ad4a5223SQuentin Monnet * Alternatively, it is possible to store the difference between 1634ad4a5223SQuentin Monnet * the previous and the new values of the header field in *to*, by 1635ad4a5223SQuentin Monnet * setting *from* and *size* to 0. For both methods, *offset* 1636ad4a5223SQuentin Monnet * indicates the location of the IP checksum within the packet. 1637ad4a5223SQuentin Monnet * 1638ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 1639ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 1640ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 1641ad4a5223SQuentin Monnet * checksum to update. 1642ad4a5223SQuentin Monnet * 164332e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1644ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1645ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1646ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1647ad4a5223SQuentin Monnet * direct packet access. 1648ad4a5223SQuentin Monnet * Return 1649ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1650ad4a5223SQuentin Monnet * 1651bdb7b79bSAndrii Nakryiko * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1652ad4a5223SQuentin Monnet * Description 1653ad4a5223SQuentin Monnet * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1654ad4a5223SQuentin Monnet * packet associated to *skb*. Computation is incremental, so the 1655ad4a5223SQuentin Monnet * helper must know the former value of the header field that was 1656ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 1657ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored on the lowest 1658ad4a5223SQuentin Monnet * four bits of *flags*. Alternatively, it is possible to store 1659ad4a5223SQuentin Monnet * the difference between the previous and the new values of the 1660ad4a5223SQuentin Monnet * header field in *to*, by setting *from* and the four lowest 1661ad4a5223SQuentin Monnet * bits of *flags* to 0. For both methods, *offset* indicates the 1662ad4a5223SQuentin Monnet * location of the IP checksum within the packet. In addition to 1663ad4a5223SQuentin Monnet * the size of the field, *flags* can be added (bitwise OR) actual 1664ad4a5223SQuentin Monnet * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1665ad4a5223SQuentin Monnet * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1666ad4a5223SQuentin Monnet * for updates resulting in a null checksum the value is set to 1667ad4a5223SQuentin Monnet * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1668ad4a5223SQuentin Monnet * the checksum is to be computed against a pseudo-header. 1669ad4a5223SQuentin Monnet * 1670ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 1671ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 1672ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 1673ad4a5223SQuentin Monnet * checksum to update. 1674ad4a5223SQuentin Monnet * 167532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1676ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1677ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1678ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1679ad4a5223SQuentin Monnet * direct packet access. 1680ad4a5223SQuentin Monnet * Return 1681ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1682ad4a5223SQuentin Monnet * 1683bdb7b79bSAndrii Nakryiko * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 1684ad4a5223SQuentin Monnet * Description 1685ad4a5223SQuentin Monnet * This special helper is used to trigger a "tail call", or in 1686ad4a5223SQuentin Monnet * other words, to jump into another eBPF program. The same stack 1687ad4a5223SQuentin Monnet * frame is used (but values on stack and in registers for the 1688ad4a5223SQuentin Monnet * caller are not accessible to the callee). This mechanism allows 1689ad4a5223SQuentin Monnet * for program chaining, either for raising the maximum number of 1690ad4a5223SQuentin Monnet * available eBPF instructions, or to execute given programs in 1691ad4a5223SQuentin Monnet * conditional blocks. For security reasons, there is an upper 1692ad4a5223SQuentin Monnet * limit to the number of successive tail calls that can be 1693ad4a5223SQuentin Monnet * performed. 1694ad4a5223SQuentin Monnet * 1695ad4a5223SQuentin Monnet * Upon call of this helper, the program attempts to jump into a 1696ad4a5223SQuentin Monnet * program referenced at index *index* in *prog_array_map*, a 1697ad4a5223SQuentin Monnet * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 1698ad4a5223SQuentin Monnet * *ctx*, a pointer to the context. 1699ad4a5223SQuentin Monnet * 1700ad4a5223SQuentin Monnet * If the call succeeds, the kernel immediately runs the first 1701ad4a5223SQuentin Monnet * instruction of the new program. This is not a function call, 1702ad4a5223SQuentin Monnet * and it never returns to the previous program. If the call 1703ad4a5223SQuentin Monnet * fails, then the helper has no effect, and the caller continues 1704ad4a5223SQuentin Monnet * to run its subsequent instructions. A call can fail if the 1705ad4a5223SQuentin Monnet * destination program for the jump does not exist (i.e. *index* 1706ad4a5223SQuentin Monnet * is superior to the number of entries in *prog_array_map*), or 1707ad4a5223SQuentin Monnet * if the maximum number of tail calls has been reached for this 1708ad4a5223SQuentin Monnet * chain of programs. This limit is defined in the kernel by the 1709ad4a5223SQuentin Monnet * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 1710ad4a5223SQuentin Monnet * which is currently set to 32. 1711ad4a5223SQuentin Monnet * Return 1712ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1713ad4a5223SQuentin Monnet * 1714bdb7b79bSAndrii Nakryiko * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 1715ad4a5223SQuentin Monnet * Description 1716ad4a5223SQuentin Monnet * Clone and redirect the packet associated to *skb* to another 1717ad4a5223SQuentin Monnet * net device of index *ifindex*. Both ingress and egress 1718ad4a5223SQuentin Monnet * interfaces can be used for redirection. The **BPF_F_INGRESS** 1719ad4a5223SQuentin Monnet * value in *flags* is used to make the distinction (ingress path 1720ad4a5223SQuentin Monnet * is selected if the flag is present, egress path otherwise). 1721ad4a5223SQuentin Monnet * This is the only flag supported for now. 1722ad4a5223SQuentin Monnet * 1723ad4a5223SQuentin Monnet * In comparison with **bpf_redirect**\ () helper, 1724ad4a5223SQuentin Monnet * **bpf_clone_redirect**\ () has the associated cost of 1725ad4a5223SQuentin Monnet * duplicating the packet buffer, but this can be executed out of 1726ad4a5223SQuentin Monnet * the eBPF program. Conversely, **bpf_redirect**\ () is more 1727ad4a5223SQuentin Monnet * efficient, but it is handled through an action code where the 1728ad4a5223SQuentin Monnet * redirection happens only after the eBPF program has returned. 1729ad4a5223SQuentin Monnet * 173032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1731ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1732ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1733ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1734ad4a5223SQuentin Monnet * direct packet access. 1735ad4a5223SQuentin Monnet * Return 1736ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1737c456dec4SQuentin Monnet * 1738c456dec4SQuentin Monnet * u64 bpf_get_current_pid_tgid(void) 1739c456dec4SQuentin Monnet * Return 1740c456dec4SQuentin Monnet * A 64-bit integer containing the current tgid and pid, and 1741c456dec4SQuentin Monnet * created as such: 1742c456dec4SQuentin Monnet * *current_task*\ **->tgid << 32 \|** 1743c456dec4SQuentin Monnet * *current_task*\ **->pid**. 1744c456dec4SQuentin Monnet * 1745c456dec4SQuentin Monnet * u64 bpf_get_current_uid_gid(void) 1746c456dec4SQuentin Monnet * Return 1747c456dec4SQuentin Monnet * A 64-bit integer containing the current GID and UID, and 1748c456dec4SQuentin Monnet * created as such: *current_gid* **<< 32 \|** *current_uid*. 1749c456dec4SQuentin Monnet * 1750bdb7b79bSAndrii Nakryiko * long bpf_get_current_comm(void *buf, u32 size_of_buf) 1751c456dec4SQuentin Monnet * Description 1752c456dec4SQuentin Monnet * Copy the **comm** attribute of the current task into *buf* of 1753c456dec4SQuentin Monnet * *size_of_buf*. The **comm** attribute contains the name of 1754c456dec4SQuentin Monnet * the executable (excluding the path) for the current task. The 1755c456dec4SQuentin Monnet * *size_of_buf* must be strictly positive. On success, the 1756c456dec4SQuentin Monnet * helper makes sure that the *buf* is NUL-terminated. On failure, 1757c456dec4SQuentin Monnet * it is filled with zeroes. 1758c456dec4SQuentin Monnet * Return 1759c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1760c456dec4SQuentin Monnet * 17611fdd08beSQuentin Monnet * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 17621fdd08beSQuentin Monnet * Description 17631fdd08beSQuentin Monnet * Retrieve the classid for the current task, i.e. for the net_cls 17641fdd08beSQuentin Monnet * cgroup to which *skb* belongs. 17651fdd08beSQuentin Monnet * 17661fdd08beSQuentin Monnet * This helper can be used on TC egress path, but not on ingress. 17671fdd08beSQuentin Monnet * 17681fdd08beSQuentin Monnet * The net_cls cgroup provides an interface to tag network packets 17691fdd08beSQuentin Monnet * based on a user-provided identifier for all traffic coming from 17701fdd08beSQuentin Monnet * the tasks belonging to the related cgroup. See also the related 17711fdd08beSQuentin Monnet * kernel documentation, available from the Linux sources in file 1772da82c92fSMauro Carvalho Chehab * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 17731fdd08beSQuentin Monnet * 17741fdd08beSQuentin Monnet * The Linux kernel has two versions for cgroups: there are 17751fdd08beSQuentin Monnet * cgroups v1 and cgroups v2. Both are available to users, who can 17761fdd08beSQuentin Monnet * use a mixture of them, but note that the net_cls cgroup is for 17771fdd08beSQuentin Monnet * cgroup v1 only. This makes it incompatible with BPF programs 17781fdd08beSQuentin Monnet * run on cgroups, which is a cgroup-v2-only feature (a socket can 17791fdd08beSQuentin Monnet * only hold data for one version of cgroups at a time). 17801fdd08beSQuentin Monnet * 17811fdd08beSQuentin Monnet * This helper is only available is the kernel was compiled with 17821fdd08beSQuentin Monnet * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 17831fdd08beSQuentin Monnet * "**y**" or to "**m**". 17841fdd08beSQuentin Monnet * Return 17851fdd08beSQuentin Monnet * The classid, or 0 for the default unconfigured classid. 17861fdd08beSQuentin Monnet * 1787bdb7b79bSAndrii Nakryiko * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1788c456dec4SQuentin Monnet * Description 1789c456dec4SQuentin Monnet * Push a *vlan_tci* (VLAN tag control information) of protocol 1790c456dec4SQuentin Monnet * *vlan_proto* to the packet associated to *skb*, then update 1791c456dec4SQuentin Monnet * the checksum. Note that if *vlan_proto* is different from 1792c456dec4SQuentin Monnet * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1793c456dec4SQuentin Monnet * be **ETH_P_8021Q**. 1794c456dec4SQuentin Monnet * 179532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1796c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1797c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 1798c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 1799c456dec4SQuentin Monnet * direct packet access. 1800c456dec4SQuentin Monnet * Return 1801c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1802c456dec4SQuentin Monnet * 1803bdb7b79bSAndrii Nakryiko * long bpf_skb_vlan_pop(struct sk_buff *skb) 1804c456dec4SQuentin Monnet * Description 1805c456dec4SQuentin Monnet * Pop a VLAN header from the packet associated to *skb*. 1806c456dec4SQuentin Monnet * 180732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1808c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1809c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 1810c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 1811c456dec4SQuentin Monnet * direct packet access. 1812c456dec4SQuentin Monnet * Return 1813c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1814c456dec4SQuentin Monnet * 1815bdb7b79bSAndrii Nakryiko * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1816c456dec4SQuentin Monnet * Description 1817c456dec4SQuentin Monnet * Get tunnel metadata. This helper takes a pointer *key* to an 1818c456dec4SQuentin Monnet * empty **struct bpf_tunnel_key** of **size**, that will be 1819c456dec4SQuentin Monnet * filled with tunnel metadata for the packet associated to *skb*. 1820c456dec4SQuentin Monnet * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1821c456dec4SQuentin Monnet * indicates that the tunnel is based on IPv6 protocol instead of 1822c456dec4SQuentin Monnet * IPv4. 1823c456dec4SQuentin Monnet * 1824c456dec4SQuentin Monnet * The **struct bpf_tunnel_key** is an object that generalizes the 1825c456dec4SQuentin Monnet * principal parameters used by various tunneling protocols into a 1826c456dec4SQuentin Monnet * single struct. This way, it can be used to easily make a 1827c456dec4SQuentin Monnet * decision based on the contents of the encapsulation header, 1828c456dec4SQuentin Monnet * "summarized" in this struct. In particular, it holds the IP 1829c456dec4SQuentin Monnet * address of the remote end (IPv4 or IPv6, depending on the case) 1830c456dec4SQuentin Monnet * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1831c456dec4SQuentin Monnet * this struct exposes the *key*\ **->tunnel_id**, which is 1832c456dec4SQuentin Monnet * generally mapped to a VNI (Virtual Network Identifier), making 1833c456dec4SQuentin Monnet * it programmable together with the **bpf_skb_set_tunnel_key**\ 1834c456dec4SQuentin Monnet * () helper. 1835c456dec4SQuentin Monnet * 1836c456dec4SQuentin Monnet * Let's imagine that the following code is part of a program 1837c456dec4SQuentin Monnet * attached to the TC ingress interface, on one end of a GRE 1838c456dec4SQuentin Monnet * tunnel, and is supposed to filter out all messages coming from 1839c456dec4SQuentin Monnet * remote ends with IPv4 address other than 10.0.0.1: 1840c456dec4SQuentin Monnet * 1841c456dec4SQuentin Monnet * :: 1842c456dec4SQuentin Monnet * 1843c456dec4SQuentin Monnet * int ret; 1844c456dec4SQuentin Monnet * struct bpf_tunnel_key key = {}; 1845c456dec4SQuentin Monnet * 1846c456dec4SQuentin Monnet * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1847c456dec4SQuentin Monnet * if (ret < 0) 1848c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 1849c456dec4SQuentin Monnet * 1850c456dec4SQuentin Monnet * if (key.remote_ipv4 != 0x0a000001) 1851c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 1852c456dec4SQuentin Monnet * 1853c456dec4SQuentin Monnet * return TC_ACT_OK; // accept packet 1854c456dec4SQuentin Monnet * 1855c456dec4SQuentin Monnet * This interface can also be used with all encapsulation devices 1856c456dec4SQuentin Monnet * that can operate in "collect metadata" mode: instead of having 1857c456dec4SQuentin Monnet * one network device per specific configuration, the "collect 1858c456dec4SQuentin Monnet * metadata" mode only requires a single device where the 1859c456dec4SQuentin Monnet * configuration can be extracted from this helper. 1860c456dec4SQuentin Monnet * 1861c456dec4SQuentin Monnet * This can be used together with various tunnels such as VXLan, 1862c456dec4SQuentin Monnet * Geneve, GRE or IP in IP (IPIP). 1863c456dec4SQuentin Monnet * Return 1864c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1865c456dec4SQuentin Monnet * 1866bdb7b79bSAndrii Nakryiko * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1867c456dec4SQuentin Monnet * Description 1868c456dec4SQuentin Monnet * Populate tunnel metadata for packet associated to *skb.* The 1869c456dec4SQuentin Monnet * tunnel metadata is set to the contents of *key*, of *size*. The 1870c456dec4SQuentin Monnet * *flags* can be set to a combination of the following values: 1871c456dec4SQuentin Monnet * 1872c456dec4SQuentin Monnet * **BPF_F_TUNINFO_IPV6** 1873c456dec4SQuentin Monnet * Indicate that the tunnel is based on IPv6 protocol 1874c456dec4SQuentin Monnet * instead of IPv4. 1875c456dec4SQuentin Monnet * **BPF_F_ZERO_CSUM_TX** 1876c456dec4SQuentin Monnet * For IPv4 packets, add a flag to tunnel metadata 1877c456dec4SQuentin Monnet * indicating that checksum computation should be skipped 1878c456dec4SQuentin Monnet * and checksum set to zeroes. 1879c456dec4SQuentin Monnet * **BPF_F_DONT_FRAGMENT** 1880c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that the 1881c456dec4SQuentin Monnet * packet should not be fragmented. 1882c456dec4SQuentin Monnet * **BPF_F_SEQ_NUMBER** 1883c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that a 1884c456dec4SQuentin Monnet * sequence number should be added to tunnel header before 1885c456dec4SQuentin Monnet * sending the packet. This flag was added for GRE 1886c456dec4SQuentin Monnet * encapsulation, but might be used with other protocols 1887c456dec4SQuentin Monnet * as well in the future. 1888c456dec4SQuentin Monnet * 1889c456dec4SQuentin Monnet * Here is a typical usage on the transmit path: 1890c456dec4SQuentin Monnet * 1891c456dec4SQuentin Monnet * :: 1892c456dec4SQuentin Monnet * 1893c456dec4SQuentin Monnet * struct bpf_tunnel_key key; 1894c456dec4SQuentin Monnet * populate key ... 1895c456dec4SQuentin Monnet * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1896c456dec4SQuentin Monnet * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1897c456dec4SQuentin Monnet * 1898c456dec4SQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_key**\ () 1899c456dec4SQuentin Monnet * helper for additional information. 1900c456dec4SQuentin Monnet * Return 1901c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1902c456dec4SQuentin Monnet * 1903c6b5fb86SQuentin Monnet * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1904c6b5fb86SQuentin Monnet * Description 1905c6b5fb86SQuentin Monnet * Read the value of a perf event counter. This helper relies on a 1906c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1907c6b5fb86SQuentin Monnet * the perf event counter is selected when *map* is updated with 1908c6b5fb86SQuentin Monnet * perf event file descriptors. The *map* is an array whose size 1909c6b5fb86SQuentin Monnet * is the number of available CPUs, and each cell contains a value 1910c6b5fb86SQuentin Monnet * relative to one CPU. The value to retrieve is indicated by 1911c6b5fb86SQuentin Monnet * *flags*, that contains the index of the CPU to look up, masked 1912c6b5fb86SQuentin Monnet * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1913c6b5fb86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 1914c6b5fb86SQuentin Monnet * current CPU should be retrieved. 1915c6b5fb86SQuentin Monnet * 1916c6b5fb86SQuentin Monnet * Note that before Linux 4.13, only hardware perf event can be 1917c6b5fb86SQuentin Monnet * retrieved. 1918c6b5fb86SQuentin Monnet * 1919c6b5fb86SQuentin Monnet * Also, be aware that the newer helper 1920c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () is recommended over 19213bd5a09bSQuentin Monnet * **bpf_perf_event_read**\ () in general. The latter has some ABI 1922c6b5fb86SQuentin Monnet * quirks where error and counter value are used as a return code 1923c6b5fb86SQuentin Monnet * (which is wrong to do since ranges may overlap). This issue is 19243bd5a09bSQuentin Monnet * fixed with **bpf_perf_event_read_value**\ (), which at the same 19253bd5a09bSQuentin Monnet * time provides more features over the **bpf_perf_event_read**\ 19263bd5a09bSQuentin Monnet * () interface. Please refer to the description of 1927c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () for details. 1928c6b5fb86SQuentin Monnet * Return 1929c6b5fb86SQuentin Monnet * The value of the perf event counter read from the map, or a 1930c6b5fb86SQuentin Monnet * negative error code in case of failure. 1931c6b5fb86SQuentin Monnet * 1932bdb7b79bSAndrii Nakryiko * long bpf_redirect(u32 ifindex, u64 flags) 1933c456dec4SQuentin Monnet * Description 1934c456dec4SQuentin Monnet * Redirect the packet to another net device of index *ifindex*. 1935c456dec4SQuentin Monnet * This helper is somewhat similar to **bpf_clone_redirect**\ 1936c456dec4SQuentin Monnet * (), except that the packet is not cloned, which provides 1937c456dec4SQuentin Monnet * increased performance. 1938c456dec4SQuentin Monnet * 1939c456dec4SQuentin Monnet * Except for XDP, both ingress and egress interfaces can be used 1940c456dec4SQuentin Monnet * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1941c456dec4SQuentin Monnet * to make the distinction (ingress path is selected if the flag 1942c456dec4SQuentin Monnet * is present, egress path otherwise). Currently, XDP only 1943c456dec4SQuentin Monnet * supports redirection to the egress interface, and accepts no 1944c456dec4SQuentin Monnet * flag at all. 1945c456dec4SQuentin Monnet * 1946f25975f4SToke Høiland-Jørgensen * The same effect can also be attained with the more generic 1947f25975f4SToke Høiland-Jørgensen * **bpf_redirect_map**\ (), which uses a BPF map to store the 1948f25975f4SToke Høiland-Jørgensen * redirect target instead of providing it directly to the helper. 1949c456dec4SQuentin Monnet * Return 1950c456dec4SQuentin Monnet * For XDP, the helper returns **XDP_REDIRECT** on success or 1951c456dec4SQuentin Monnet * **XDP_ABORTED** on error. For other program types, the values 1952c456dec4SQuentin Monnet * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1953c456dec4SQuentin Monnet * error. 1954c456dec4SQuentin Monnet * 19551fdd08beSQuentin Monnet * u32 bpf_get_route_realm(struct sk_buff *skb) 19561fdd08beSQuentin Monnet * Description 19571fdd08beSQuentin Monnet * Retrieve the realm or the route, that is to say the 19581fdd08beSQuentin Monnet * **tclassid** field of the destination for the *skb*. The 1959b16fc097STobias Klauser * identifier retrieved is a user-provided tag, similar to the 19601fdd08beSQuentin Monnet * one used with the net_cls cgroup (see description for 19611fdd08beSQuentin Monnet * **bpf_get_cgroup_classid**\ () helper), but here this tag is 19621fdd08beSQuentin Monnet * held by a route (a destination entry), not by a task. 19631fdd08beSQuentin Monnet * 19641fdd08beSQuentin Monnet * Retrieving this identifier works with the clsact TC egress hook 19651fdd08beSQuentin Monnet * (see also **tc-bpf(8)**), or alternatively on conventional 19661fdd08beSQuentin Monnet * classful egress qdiscs, but not on TC ingress path. In case of 19671fdd08beSQuentin Monnet * clsact TC egress hook, this has the advantage that, internally, 19681fdd08beSQuentin Monnet * the destination entry has not been dropped yet in the transmit 19691fdd08beSQuentin Monnet * path. Therefore, the destination entry does not need to be 19701fdd08beSQuentin Monnet * artificially held via **netif_keep_dst**\ () for a classful 19711fdd08beSQuentin Monnet * qdisc until the *skb* is freed. 19721fdd08beSQuentin Monnet * 19731fdd08beSQuentin Monnet * This helper is available only if the kernel was compiled with 19741fdd08beSQuentin Monnet * **CONFIG_IP_ROUTE_CLASSID** configuration option. 19751fdd08beSQuentin Monnet * Return 19761fdd08beSQuentin Monnet * The realm of the route for the packet associated to *skb*, or 0 19771fdd08beSQuentin Monnet * if none was found. 19781fdd08beSQuentin Monnet * 1979bdb7b79bSAndrii Nakryiko * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1980c456dec4SQuentin Monnet * Description 1981c456dec4SQuentin Monnet * Write raw *data* blob into a special BPF perf event held by 1982c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1983c456dec4SQuentin Monnet * event must have the following attributes: **PERF_SAMPLE_RAW** 1984c456dec4SQuentin Monnet * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1985c456dec4SQuentin Monnet * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1986c456dec4SQuentin Monnet * 1987c456dec4SQuentin Monnet * The *flags* are used to indicate the index in *map* for which 1988c456dec4SQuentin Monnet * the value must be put, masked with **BPF_F_INDEX_MASK**. 1989c456dec4SQuentin Monnet * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1990c456dec4SQuentin Monnet * to indicate that the index of the current CPU core should be 1991c456dec4SQuentin Monnet * used. 1992c456dec4SQuentin Monnet * 1993c456dec4SQuentin Monnet * The value to write, of *size*, is passed through eBPF stack and 1994c456dec4SQuentin Monnet * pointed by *data*. 1995c456dec4SQuentin Monnet * 1996c456dec4SQuentin Monnet * The context of the program *ctx* needs also be passed to the 1997c456dec4SQuentin Monnet * helper. 1998c456dec4SQuentin Monnet * 1999c456dec4SQuentin Monnet * On user space, a program willing to read the values needs to 2000c456dec4SQuentin Monnet * call **perf_event_open**\ () on the perf event (either for 2001c456dec4SQuentin Monnet * one or for all CPUs) and to store the file descriptor into the 2002c456dec4SQuentin Monnet * *map*. This must be done before the eBPF program can send data 2003c456dec4SQuentin Monnet * into it. An example is available in file 2004c456dec4SQuentin Monnet * *samples/bpf/trace_output_user.c* in the Linux kernel source 2005c456dec4SQuentin Monnet * tree (the eBPF program counterpart is in 2006c456dec4SQuentin Monnet * *samples/bpf/trace_output_kern.c*). 2007c456dec4SQuentin Monnet * 2008c456dec4SQuentin Monnet * **bpf_perf_event_output**\ () achieves better performance 2009c456dec4SQuentin Monnet * than **bpf_trace_printk**\ () for sharing data with user 2010c456dec4SQuentin Monnet * space, and is much better suitable for streaming data from eBPF 2011c456dec4SQuentin Monnet * programs. 2012c456dec4SQuentin Monnet * 2013c456dec4SQuentin Monnet * Note that this helper is not restricted to tracing use cases 2014c456dec4SQuentin Monnet * and can be used with programs attached to TC or XDP as well, 2015c456dec4SQuentin Monnet * where it allows for passing data to user space listeners. Data 2016c456dec4SQuentin Monnet * can be: 2017c456dec4SQuentin Monnet * 2018c456dec4SQuentin Monnet * * Only custom structs, 2019c456dec4SQuentin Monnet * * Only the packet payload, or 2020c456dec4SQuentin Monnet * * A combination of both. 2021c456dec4SQuentin Monnet * Return 2022c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 2023c456dec4SQuentin Monnet * 2024bdb7b79bSAndrii Nakryiko * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 20251fdd08beSQuentin Monnet * Description 20261fdd08beSQuentin Monnet * This helper was provided as an easy way to load data from a 20271fdd08beSQuentin Monnet * packet. It can be used to load *len* bytes from *offset* from 20281fdd08beSQuentin Monnet * the packet associated to *skb*, into the buffer pointed by 20291fdd08beSQuentin Monnet * *to*. 20301fdd08beSQuentin Monnet * 20311fdd08beSQuentin Monnet * Since Linux 4.7, usage of this helper has mostly been replaced 20321fdd08beSQuentin Monnet * by "direct packet access", enabling packet data to be 20331fdd08beSQuentin Monnet * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 20341fdd08beSQuentin Monnet * pointing respectively to the first byte of packet data and to 20351fdd08beSQuentin Monnet * the byte after the last byte of packet data. However, it 20361fdd08beSQuentin Monnet * remains useful if one wishes to read large quantities of data 20371fdd08beSQuentin Monnet * at once from a packet into the eBPF stack. 20381fdd08beSQuentin Monnet * Return 20391fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 20401fdd08beSQuentin Monnet * 2041bdb7b79bSAndrii Nakryiko * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2042c456dec4SQuentin Monnet * Description 2043c456dec4SQuentin Monnet * Walk a user or a kernel stack and return its id. To achieve 2044c456dec4SQuentin Monnet * this, the helper needs *ctx*, which is a pointer to the context 2045c456dec4SQuentin Monnet * on which the tracing program is executed, and a pointer to a 2046c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2047c456dec4SQuentin Monnet * 2048c456dec4SQuentin Monnet * The last argument, *flags*, holds the number of stack frames to 2049c456dec4SQuentin Monnet * skip (from 0 to 255), masked with 2050c456dec4SQuentin Monnet * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2051c456dec4SQuentin Monnet * a combination of the following flags: 2052c456dec4SQuentin Monnet * 2053c456dec4SQuentin Monnet * **BPF_F_USER_STACK** 2054c456dec4SQuentin Monnet * Collect a user space stack instead of a kernel stack. 2055c456dec4SQuentin Monnet * **BPF_F_FAST_STACK_CMP** 2056c456dec4SQuentin Monnet * Compare stacks by hash only. 2057c456dec4SQuentin Monnet * **BPF_F_REUSE_STACKID** 2058c456dec4SQuentin Monnet * If two different stacks hash into the same *stackid*, 2059c456dec4SQuentin Monnet * discard the old one. 2060c456dec4SQuentin Monnet * 2061c456dec4SQuentin Monnet * The stack id retrieved is a 32 bit long integer handle which 2062c456dec4SQuentin Monnet * can be further combined with other data (including other stack 2063c456dec4SQuentin Monnet * ids) and used as a key into maps. This can be useful for 2064c456dec4SQuentin Monnet * generating a variety of graphs (such as flame graphs or off-cpu 2065c456dec4SQuentin Monnet * graphs). 2066c456dec4SQuentin Monnet * 2067c456dec4SQuentin Monnet * For walking a stack, this helper is an improvement over 2068c456dec4SQuentin Monnet * **bpf_probe_read**\ (), which can be used with unrolled loops 2069c456dec4SQuentin Monnet * but is not efficient and consumes a lot of eBPF instructions. 2070c456dec4SQuentin Monnet * Instead, **bpf_get_stackid**\ () can collect up to 2071c456dec4SQuentin Monnet * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2072c456dec4SQuentin Monnet * this limit can be controlled with the **sysctl** program, and 2073c456dec4SQuentin Monnet * that it should be manually increased in order to profile long 2074c456dec4SQuentin Monnet * user stacks (such as stacks for Java programs). To do so, use: 2075c456dec4SQuentin Monnet * 2076c456dec4SQuentin Monnet * :: 2077c456dec4SQuentin Monnet * 2078c456dec4SQuentin Monnet * # sysctl kernel.perf_event_max_stack=<new value> 2079c456dec4SQuentin Monnet * Return 2080c456dec4SQuentin Monnet * The positive or null stack id on success, or a negative error 2081c456dec4SQuentin Monnet * in case of failure. 2082c456dec4SQuentin Monnet * 20831fdd08beSQuentin Monnet * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 20841fdd08beSQuentin Monnet * Description 20851fdd08beSQuentin Monnet * Compute a checksum difference, from the raw buffer pointed by 20861fdd08beSQuentin Monnet * *from*, of length *from_size* (that must be a multiple of 4), 20871fdd08beSQuentin Monnet * towards the raw buffer pointed by *to*, of size *to_size* 20881fdd08beSQuentin Monnet * (same remark). An optional *seed* can be added to the value 20891fdd08beSQuentin Monnet * (this can be cascaded, the seed may come from a previous call 20901fdd08beSQuentin Monnet * to the helper). 20911fdd08beSQuentin Monnet * 20921fdd08beSQuentin Monnet * This is flexible enough to be used in several ways: 20931fdd08beSQuentin Monnet * 20941fdd08beSQuentin Monnet * * With *from_size* == 0, *to_size* > 0 and *seed* set to 20951fdd08beSQuentin Monnet * checksum, it can be used when pushing new data. 20961fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* == 0 and *seed* set to 20971fdd08beSQuentin Monnet * checksum, it can be used when removing data from a packet. 20981fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 20991fdd08beSQuentin Monnet * can be used to compute a diff. Note that *from_size* and 21001fdd08beSQuentin Monnet * *to_size* do not need to be equal. 21011fdd08beSQuentin Monnet * 21021fdd08beSQuentin Monnet * This helper can be used in combination with 21031fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 21041fdd08beSQuentin Monnet * which one can feed in the difference computed with 21051fdd08beSQuentin Monnet * **bpf_csum_diff**\ (). 21061fdd08beSQuentin Monnet * Return 21071fdd08beSQuentin Monnet * The checksum result, or a negative error code in case of 21081fdd08beSQuentin Monnet * failure. 21091fdd08beSQuentin Monnet * 2110bdb7b79bSAndrii Nakryiko * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 21111fdd08beSQuentin Monnet * Description 21121fdd08beSQuentin Monnet * Retrieve tunnel options metadata for the packet associated to 21131fdd08beSQuentin Monnet * *skb*, and store the raw tunnel option data to the buffer *opt* 21141fdd08beSQuentin Monnet * of *size*. 21151fdd08beSQuentin Monnet * 21161fdd08beSQuentin Monnet * This helper can be used with encapsulation devices that can 21171fdd08beSQuentin Monnet * operate in "collect metadata" mode (please refer to the related 21181fdd08beSQuentin Monnet * note in the description of **bpf_skb_get_tunnel_key**\ () for 21191fdd08beSQuentin Monnet * more details). A particular example where this can be used is 21201fdd08beSQuentin Monnet * in combination with the Geneve encapsulation protocol, where it 21211fdd08beSQuentin Monnet * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 21221fdd08beSQuentin Monnet * and retrieving arbitrary TLVs (Type-Length-Value headers) from 21231fdd08beSQuentin Monnet * the eBPF program. This allows for full customization of these 21241fdd08beSQuentin Monnet * headers. 21251fdd08beSQuentin Monnet * Return 21261fdd08beSQuentin Monnet * The size of the option data retrieved. 21271fdd08beSQuentin Monnet * 2128bdb7b79bSAndrii Nakryiko * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 21291fdd08beSQuentin Monnet * Description 21301fdd08beSQuentin Monnet * Set tunnel options metadata for the packet associated to *skb* 21311fdd08beSQuentin Monnet * to the option data contained in the raw buffer *opt* of *size*. 21321fdd08beSQuentin Monnet * 21331fdd08beSQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_opt**\ () 21341fdd08beSQuentin Monnet * helper for additional information. 21351fdd08beSQuentin Monnet * Return 21361fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 21371fdd08beSQuentin Monnet * 2138bdb7b79bSAndrii Nakryiko * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 21391fdd08beSQuentin Monnet * Description 21401fdd08beSQuentin Monnet * Change the protocol of the *skb* to *proto*. Currently 21411fdd08beSQuentin Monnet * supported are transition from IPv4 to IPv6, and from IPv6 to 21421fdd08beSQuentin Monnet * IPv4. The helper takes care of the groundwork for the 21431fdd08beSQuentin Monnet * transition, including resizing the socket buffer. The eBPF 21441fdd08beSQuentin Monnet * program is expected to fill the new headers, if any, via 21451fdd08beSQuentin Monnet * **skb_store_bytes**\ () and to recompute the checksums with 21461fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 21471fdd08beSQuentin Monnet * (). The main case for this helper is to perform NAT64 21481fdd08beSQuentin Monnet * operations out of an eBPF program. 21491fdd08beSQuentin Monnet * 21501fdd08beSQuentin Monnet * Internally, the GSO type is marked as dodgy so that headers are 21511fdd08beSQuentin Monnet * checked and segments are recalculated by the GSO/GRO engine. 21521fdd08beSQuentin Monnet * The size for GSO target is adapted as well. 21531fdd08beSQuentin Monnet * 21541fdd08beSQuentin Monnet * All values for *flags* are reserved for future usage, and must 21551fdd08beSQuentin Monnet * be left at zero. 21561fdd08beSQuentin Monnet * 215732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 21581fdd08beSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 21591fdd08beSQuentin Monnet * previously done by the verifier are invalidated and must be 21601fdd08beSQuentin Monnet * performed again, if the helper is used in combination with 21611fdd08beSQuentin Monnet * direct packet access. 21621fdd08beSQuentin Monnet * Return 21631fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 21641fdd08beSQuentin Monnet * 2165bdb7b79bSAndrii Nakryiko * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 21661fdd08beSQuentin Monnet * Description 21671fdd08beSQuentin Monnet * Change the packet type for the packet associated to *skb*. This 21681fdd08beSQuentin Monnet * comes down to setting *skb*\ **->pkt_type** to *type*, except 21691fdd08beSQuentin Monnet * the eBPF program does not have a write access to *skb*\ 21701fdd08beSQuentin Monnet * **->pkt_type** beside this helper. Using a helper here allows 21711fdd08beSQuentin Monnet * for graceful handling of errors. 21721fdd08beSQuentin Monnet * 21731fdd08beSQuentin Monnet * The major use case is to change incoming *skb*s to 21741fdd08beSQuentin Monnet * **PACKET_HOST** in a programmatic way instead of having to 21751fdd08beSQuentin Monnet * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 21761fdd08beSQuentin Monnet * example. 21771fdd08beSQuentin Monnet * 21781fdd08beSQuentin Monnet * Note that *type* only allows certain values. At this time, they 21791fdd08beSQuentin Monnet * are: 21801fdd08beSQuentin Monnet * 21811fdd08beSQuentin Monnet * **PACKET_HOST** 21821fdd08beSQuentin Monnet * Packet is for us. 21831fdd08beSQuentin Monnet * **PACKET_BROADCAST** 21841fdd08beSQuentin Monnet * Send packet to all. 21851fdd08beSQuentin Monnet * **PACKET_MULTICAST** 21861fdd08beSQuentin Monnet * Send packet to group. 21871fdd08beSQuentin Monnet * **PACKET_OTHERHOST** 21881fdd08beSQuentin Monnet * Send packet to someone else. 21891fdd08beSQuentin Monnet * Return 21901fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 21911fdd08beSQuentin Monnet * 2192bdb7b79bSAndrii Nakryiko * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2193c6b5fb86SQuentin Monnet * Description 2194c6b5fb86SQuentin Monnet * Check whether *skb* is a descendant of the cgroup2 held by 2195c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2196c6b5fb86SQuentin Monnet * Return 2197c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 2198c6b5fb86SQuentin Monnet * 2199c6b5fb86SQuentin Monnet * * 0, if the *skb* failed the cgroup2 descendant test. 2200c6b5fb86SQuentin Monnet * * 1, if the *skb* succeeded the cgroup2 descendant test. 2201c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 2202c6b5fb86SQuentin Monnet * 2203fa15601aSQuentin Monnet * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2204fa15601aSQuentin Monnet * Description 2205fa15601aSQuentin Monnet * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2206fa15601aSQuentin Monnet * not set, in particular if the hash was cleared due to mangling, 2207fa15601aSQuentin Monnet * recompute this hash. Later accesses to the hash can be done 2208fa15601aSQuentin Monnet * directly with *skb*\ **->hash**. 2209fa15601aSQuentin Monnet * 2210fa15601aSQuentin Monnet * Calling **bpf_set_hash_invalid**\ (), changing a packet 2211fa15601aSQuentin Monnet * prototype with **bpf_skb_change_proto**\ (), or calling 2212fa15601aSQuentin Monnet * **bpf_skb_store_bytes**\ () with the 2213fa15601aSQuentin Monnet * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2214fa15601aSQuentin Monnet * the hash and to trigger a new computation for the next call to 2215fa15601aSQuentin Monnet * **bpf_get_hash_recalc**\ (). 2216fa15601aSQuentin Monnet * Return 2217fa15601aSQuentin Monnet * The 32-bit hash. 2218fa15601aSQuentin Monnet * 2219c456dec4SQuentin Monnet * u64 bpf_get_current_task(void) 2220c456dec4SQuentin Monnet * Return 2221c456dec4SQuentin Monnet * A pointer to the current task struct. 2222fa15601aSQuentin Monnet * 2223bdb7b79bSAndrii Nakryiko * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2224c6b5fb86SQuentin Monnet * Description 2225c6b5fb86SQuentin Monnet * Attempt in a safe way to write *len* bytes from the buffer 2226c6b5fb86SQuentin Monnet * *src* to *dst* in memory. It only works for threads that are in 2227c6b5fb86SQuentin Monnet * user context, and *dst* must be a valid user space address. 2228c6b5fb86SQuentin Monnet * 2229c6b5fb86SQuentin Monnet * This helper should not be used to implement any kind of 2230c6b5fb86SQuentin Monnet * security mechanism because of TOC-TOU attacks, but rather to 2231c6b5fb86SQuentin Monnet * debug, divert, and manipulate execution of semi-cooperative 2232c6b5fb86SQuentin Monnet * processes. 2233c6b5fb86SQuentin Monnet * 2234c6b5fb86SQuentin Monnet * Keep in mind that this feature is meant for experiments, and it 2235c6b5fb86SQuentin Monnet * has a risk of crashing the system and running programs. 2236c6b5fb86SQuentin Monnet * Therefore, when an eBPF program using this helper is attached, 2237c6b5fb86SQuentin Monnet * a warning including PID and process name is printed to kernel 2238c6b5fb86SQuentin Monnet * logs. 2239c6b5fb86SQuentin Monnet * Return 2240c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2241c6b5fb86SQuentin Monnet * 2242bdb7b79bSAndrii Nakryiko * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2243c6b5fb86SQuentin Monnet * Description 2244c6b5fb86SQuentin Monnet * Check whether the probe is being run is the context of a given 2245c6b5fb86SQuentin Monnet * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2246c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2247c6b5fb86SQuentin Monnet * Return 2248c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 2249c6b5fb86SQuentin Monnet * 22501aef5b43SSong Liu * * 0, if current task belongs to the cgroup2. 22511aef5b43SSong Liu * * 1, if current task does not belong to the cgroup2. 2252c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 2253c6b5fb86SQuentin Monnet * 2254bdb7b79bSAndrii Nakryiko * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2255fa15601aSQuentin Monnet * Description 2256fa15601aSQuentin Monnet * Resize (trim or grow) the packet associated to *skb* to the 2257fa15601aSQuentin Monnet * new *len*. The *flags* are reserved for future usage, and must 2258fa15601aSQuentin Monnet * be left at zero. 2259fa15601aSQuentin Monnet * 2260fa15601aSQuentin Monnet * The basic idea is that the helper performs the needed work to 2261fa15601aSQuentin Monnet * change the size of the packet, then the eBPF program rewrites 2262fa15601aSQuentin Monnet * the rest via helpers like **bpf_skb_store_bytes**\ (), 2263fa15601aSQuentin Monnet * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2264fa15601aSQuentin Monnet * and others. This helper is a slow path utility intended for 2265fa15601aSQuentin Monnet * replies with control messages. And because it is targeted for 2266fa15601aSQuentin Monnet * slow path, the helper itself can afford to be slow: it 2267fa15601aSQuentin Monnet * implicitly linearizes, unclones and drops offloads from the 2268fa15601aSQuentin Monnet * *skb*. 2269fa15601aSQuentin Monnet * 227032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2271fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2272fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2273fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2274fa15601aSQuentin Monnet * direct packet access. 2275fa15601aSQuentin Monnet * Return 2276fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2277fa15601aSQuentin Monnet * 2278bdb7b79bSAndrii Nakryiko * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2279fa15601aSQuentin Monnet * Description 2280fa15601aSQuentin Monnet * Pull in non-linear data in case the *skb* is non-linear and not 2281fa15601aSQuentin Monnet * all of *len* are part of the linear section. Make *len* bytes 2282fa15601aSQuentin Monnet * from *skb* readable and writable. If a zero value is passed for 2283fa15601aSQuentin Monnet * *len*, then the whole length of the *skb* is pulled. 2284fa15601aSQuentin Monnet * 2285fa15601aSQuentin Monnet * This helper is only needed for reading and writing with direct 2286fa15601aSQuentin Monnet * packet access. 2287fa15601aSQuentin Monnet * 2288fa15601aSQuentin Monnet * For direct packet access, testing that offsets to access 2289fa15601aSQuentin Monnet * are within packet boundaries (test on *skb*\ **->data_end**) is 2290fa15601aSQuentin Monnet * susceptible to fail if offsets are invalid, or if the requested 2291fa15601aSQuentin Monnet * data is in non-linear parts of the *skb*. On failure the 2292fa15601aSQuentin Monnet * program can just bail out, or in the case of a non-linear 2293fa15601aSQuentin Monnet * buffer, use a helper to make the data available. The 2294fa15601aSQuentin Monnet * **bpf_skb_load_bytes**\ () helper is a first solution to access 2295fa15601aSQuentin Monnet * the data. Another one consists in using **bpf_skb_pull_data** 2296fa15601aSQuentin Monnet * to pull in once the non-linear parts, then retesting and 2297fa15601aSQuentin Monnet * eventually access the data. 2298fa15601aSQuentin Monnet * 2299fa15601aSQuentin Monnet * At the same time, this also makes sure the *skb* is uncloned, 2300fa15601aSQuentin Monnet * which is a necessary condition for direct write. As this needs 2301fa15601aSQuentin Monnet * to be an invariant for the write part only, the verifier 2302fa15601aSQuentin Monnet * detects writes and adds a prologue that is calling 2303fa15601aSQuentin Monnet * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2304fa15601aSQuentin Monnet * the very beginning in case it is indeed cloned. 2305fa15601aSQuentin Monnet * 230632e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2307fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2308fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2309fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2310fa15601aSQuentin Monnet * direct packet access. 2311fa15601aSQuentin Monnet * Return 2312fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2313fa15601aSQuentin Monnet * 2314fa15601aSQuentin Monnet * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2315fa15601aSQuentin Monnet * Description 2316fa15601aSQuentin Monnet * Add the checksum *csum* into *skb*\ **->csum** in case the 2317fa15601aSQuentin Monnet * driver has supplied a checksum for the entire packet into that 2318fa15601aSQuentin Monnet * field. Return an error otherwise. This helper is intended to be 2319fa15601aSQuentin Monnet * used in combination with **bpf_csum_diff**\ (), in particular 2320fa15601aSQuentin Monnet * when the checksum needs to be updated after data has been 2321fa15601aSQuentin Monnet * written into the packet through direct packet access. 2322fa15601aSQuentin Monnet * Return 2323fa15601aSQuentin Monnet * The checksum on success, or a negative error code in case of 2324fa15601aSQuentin Monnet * failure. 2325fa15601aSQuentin Monnet * 2326fa15601aSQuentin Monnet * void bpf_set_hash_invalid(struct sk_buff *skb) 2327fa15601aSQuentin Monnet * Description 2328fa15601aSQuentin Monnet * Invalidate the current *skb*\ **->hash**. It can be used after 2329fa15601aSQuentin Monnet * mangling on headers through direct packet access, in order to 2330fa15601aSQuentin Monnet * indicate that the hash is outdated and to trigger a 2331fa15601aSQuentin Monnet * recalculation the next time the kernel tries to access this 2332fa15601aSQuentin Monnet * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2333fa15601aSQuentin Monnet * 2334bdb7b79bSAndrii Nakryiko * long bpf_get_numa_node_id(void) 2335fa15601aSQuentin Monnet * Description 2336fa15601aSQuentin Monnet * Return the id of the current NUMA node. The primary use case 2337fa15601aSQuentin Monnet * for this helper is the selection of sockets for the local NUMA 2338fa15601aSQuentin Monnet * node, when the program is attached to sockets using the 2339fa15601aSQuentin Monnet * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2340fa15601aSQuentin Monnet * but the helper is also available to other eBPF program types, 2341fa15601aSQuentin Monnet * similarly to **bpf_get_smp_processor_id**\ (). 2342fa15601aSQuentin Monnet * Return 2343fa15601aSQuentin Monnet * The id of current NUMA node. 2344fa15601aSQuentin Monnet * 2345bdb7b79bSAndrii Nakryiko * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2346c6b5fb86SQuentin Monnet * Description 2347c6b5fb86SQuentin Monnet * Grows headroom of packet associated to *skb* and adjusts the 2348c6b5fb86SQuentin Monnet * offset of the MAC header accordingly, adding *len* bytes of 2349c6b5fb86SQuentin Monnet * space. It automatically extends and reallocates memory as 2350c6b5fb86SQuentin Monnet * required. 2351c6b5fb86SQuentin Monnet * 2352c6b5fb86SQuentin Monnet * This helper can be used on a layer 3 *skb* to push a MAC header 2353c6b5fb86SQuentin Monnet * for redirection into a layer 2 device. 2354c6b5fb86SQuentin Monnet * 2355c6b5fb86SQuentin Monnet * All values for *flags* are reserved for future usage, and must 2356c6b5fb86SQuentin Monnet * be left at zero. 2357c6b5fb86SQuentin Monnet * 235832e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2359c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2360c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 2361c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 2362c6b5fb86SQuentin Monnet * direct packet access. 2363c6b5fb86SQuentin Monnet * Return 2364c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2365c6b5fb86SQuentin Monnet * 2366bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2367c6b5fb86SQuentin Monnet * Description 2368c6b5fb86SQuentin Monnet * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2369c6b5fb86SQuentin Monnet * it is possible to use a negative value for *delta*. This helper 2370c6b5fb86SQuentin Monnet * can be used to prepare the packet for pushing or popping 2371c6b5fb86SQuentin Monnet * headers. 2372c6b5fb86SQuentin Monnet * 237332e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2374c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2375c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 2376c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 2377c6b5fb86SQuentin Monnet * direct packet access. 2378c6b5fb86SQuentin Monnet * Return 2379c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2380c6b5fb86SQuentin Monnet * 2381bdb7b79bSAndrii Nakryiko * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2382c6b5fb86SQuentin Monnet * Description 23836ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe kernel address 2384ab8d7809SQuentin Monnet * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 23856ae08ae3SDaniel Borkmann * more details. 2386c6b5fb86SQuentin Monnet * 2387ab8d7809SQuentin Monnet * Generally, use **bpf_probe_read_user_str**\ () or 2388ab8d7809SQuentin Monnet * **bpf_probe_read_kernel_str**\ () instead. 2389c6b5fb86SQuentin Monnet * Return 2390c6b5fb86SQuentin Monnet * On success, the strictly positive length of the string, 2391c6b5fb86SQuentin Monnet * including the trailing NUL character. On error, a negative 2392c6b5fb86SQuentin Monnet * value. 2393c6b5fb86SQuentin Monnet * 2394c6b5fb86SQuentin Monnet * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2395c6b5fb86SQuentin Monnet * Description 2396c6b5fb86SQuentin Monnet * If the **struct sk_buff** pointed by *skb* has a known socket, 2397c6b5fb86SQuentin Monnet * retrieve the cookie (generated by the kernel) of this socket. 2398c6b5fb86SQuentin Monnet * If no cookie has been set yet, generate a new cookie. Once 2399c6b5fb86SQuentin Monnet * generated, the socket cookie remains stable for the life of the 2400c6b5fb86SQuentin Monnet * socket. This helper can be useful for monitoring per socket 2401cd48bddaSDaniel Borkmann * networking traffic statistics as it provides a global socket 2402cd48bddaSDaniel Borkmann * identifier that can be assumed unique. 2403c6b5fb86SQuentin Monnet * Return 240407881ccbSFlorent Revest * A 8-byte long unique number on success, or 0 if the socket 240507881ccbSFlorent Revest * field is missing inside *skb*. 2406c6b5fb86SQuentin Monnet * 2407d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2408d692f113SAndrey Ignatov * Description 2409d692f113SAndrey Ignatov * Equivalent to bpf_get_socket_cookie() helper that accepts 241062369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_addr** context. 2411d692f113SAndrey Ignatov * Return 241207881ccbSFlorent Revest * A 8-byte long unique number. 2413d692f113SAndrey Ignatov * 2414d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2415d692f113SAndrey Ignatov * Description 2416ab8d7809SQuentin Monnet * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 241762369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_ops** context. 2418d692f113SAndrey Ignatov * Return 241907881ccbSFlorent Revest * A 8-byte long unique number. 2420d692f113SAndrey Ignatov * 2421c5dbb89fSFlorent Revest * u64 bpf_get_socket_cookie(struct sock *sk) 2422c5dbb89fSFlorent Revest * Description 2423c5dbb89fSFlorent Revest * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2424c5dbb89fSFlorent Revest * *sk*, but gets socket from a BTF **struct sock**. This helper 2425c5dbb89fSFlorent Revest * also works for sleepable programs. 2426c5dbb89fSFlorent Revest * Return 2427c5dbb89fSFlorent Revest * A 8-byte long unique number or 0 if *sk* is NULL. 2428c5dbb89fSFlorent Revest * 2429c6b5fb86SQuentin Monnet * u32 bpf_get_socket_uid(struct sk_buff *skb) 2430c6b5fb86SQuentin Monnet * Return 2431c6b5fb86SQuentin Monnet * The owner UID of the socket associated to *skb*. If the socket 2432c6b5fb86SQuentin Monnet * is **NULL**, or if it is not a full socket (i.e. if it is a 2433c6b5fb86SQuentin Monnet * time-wait or a request socket instead), **overflowuid** value 2434c6b5fb86SQuentin Monnet * is returned (note that **overflowuid** might also be the actual 2435c6b5fb86SQuentin Monnet * UID value for the socket). 2436c6b5fb86SQuentin Monnet * 2437bdb7b79bSAndrii Nakryiko * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2438fa15601aSQuentin Monnet * Description 2439fa15601aSQuentin Monnet * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2440fa15601aSQuentin Monnet * to value *hash*. 2441fa15601aSQuentin Monnet * Return 2442fa15601aSQuentin Monnet * 0 2443fa15601aSQuentin Monnet * 2444bdb7b79bSAndrii Nakryiko * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 24457aa79a86SQuentin Monnet * Description 24467aa79a86SQuentin Monnet * Emulate a call to **setsockopt()** on the socket associated to 24477aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 24487aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 24497aa79a86SQuentin Monnet * must be specified, see **setsockopt(2)** for more information. 24507aa79a86SQuentin Monnet * The option value of length *optlen* is pointed by *optval*. 24517aa79a86SQuentin Monnet * 2452beecf11bSStanislav Fomichev * *bpf_socket* should be one of the following: 2453ab8d7809SQuentin Monnet * 2454beecf11bSStanislav Fomichev * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2455beecf11bSStanislav Fomichev * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2456beecf11bSStanislav Fomichev * and **BPF_CGROUP_INET6_CONNECT**. 2457beecf11bSStanislav Fomichev * 24587aa79a86SQuentin Monnet * This helper actually implements a subset of **setsockopt()**. 24597aa79a86SQuentin Monnet * It supports the following *level*\ s: 24607aa79a86SQuentin Monnet * 24617aa79a86SQuentin Monnet * * **SOL_SOCKET**, which supports the following *optname*\ s: 24627aa79a86SQuentin Monnet * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2463f9bcf968SDmitry Yakunin * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2464f9bcf968SDmitry Yakunin * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 24657aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports the following *optname*\ s: 24667aa79a86SQuentin Monnet * **TCP_CONGESTION**, **TCP_BPF_IW**, 2467f9bcf968SDmitry Yakunin * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2468f9bcf968SDmitry Yakunin * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2469eca43ee6SNikita V. Shirokov * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**. 24707aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 24717aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 24727aa79a86SQuentin Monnet * Return 24737aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 24747aa79a86SQuentin Monnet * 2475bdb7b79bSAndrii Nakryiko * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2476fa15601aSQuentin Monnet * Description 2477fa15601aSQuentin Monnet * Grow or shrink the room for data in the packet associated to 2478fa15601aSQuentin Monnet * *skb* by *len_diff*, and according to the selected *mode*. 2479fa15601aSQuentin Monnet * 2480836e66c2SDaniel Borkmann * By default, the helper will reset any offloaded checksum 2481836e66c2SDaniel Borkmann * indicator of the skb to CHECKSUM_NONE. This can be avoided 2482836e66c2SDaniel Borkmann * by the following flag: 2483836e66c2SDaniel Borkmann * 2484836e66c2SDaniel Borkmann * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2485836e66c2SDaniel Borkmann * checksum data of the skb to CHECKSUM_NONE. 2486836e66c2SDaniel Borkmann * 248714aa3192SWillem de Bruijn * There are two supported modes at this time: 248814aa3192SWillem de Bruijn * 248914aa3192SWillem de Bruijn * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 249014aa3192SWillem de Bruijn * (room space is added or removed below the layer 2 header). 2491fa15601aSQuentin Monnet * 2492fa15601aSQuentin Monnet * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2493fa15601aSQuentin Monnet * (room space is added or removed below the layer 3 header). 2494fa15601aSQuentin Monnet * 2495868d5235SWillem de Bruijn * The following flags are supported at this time: 24962278f6ccSWillem de Bruijn * 24972278f6ccSWillem de Bruijn * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 24982278f6ccSWillem de Bruijn * Adjusting mss in this way is not allowed for datagrams. 2499fa15601aSQuentin Monnet * 250080867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 250180867c5eSQuentin Monnet * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2502868d5235SWillem de Bruijn * Any new space is reserved to hold a tunnel header. 2503868d5235SWillem de Bruijn * Configure skb offsets and other fields accordingly. 2504868d5235SWillem de Bruijn * 250580867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 250680867c5eSQuentin Monnet * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2507868d5235SWillem de Bruijn * Use with ENCAP_L3 flags to further specify the tunnel type. 2508868d5235SWillem de Bruijn * 250980867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 251058dfc900SAlan Maguire * Use with ENCAP_L3/L4 flags to further specify the tunnel 251180867c5eSQuentin Monnet * type; *len* is the length of the inner MAC header. 251258dfc900SAlan Maguire * 2513d01b59c9SXuesen Huang * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2514d01b59c9SXuesen Huang * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2515d01b59c9SXuesen Huang * L2 type as Ethernet. 2516d01b59c9SXuesen Huang * 251732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2518fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2519fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2520fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2521fa15601aSQuentin Monnet * direct packet access. 2522fa15601aSQuentin Monnet * Return 2523fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2524fa15601aSQuentin Monnet * 2525bdb7b79bSAndrii Nakryiko * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 2526ab127040SQuentin Monnet * Description 2527ab127040SQuentin Monnet * Redirect the packet to the endpoint referenced by *map* at 2528ab127040SQuentin Monnet * index *key*. Depending on its type, this *map* can contain 2529ab127040SQuentin Monnet * references to net devices (for forwarding packets through other 2530ab127040SQuentin Monnet * ports), or to CPUs (for redirecting XDP frames to another CPU; 2531ab127040SQuentin Monnet * but this is only implemented for native XDP (with driver 2532ab127040SQuentin Monnet * support) as of this writing). 2533ab127040SQuentin Monnet * 253443e74c02SToke Høiland-Jørgensen * The lower two bits of *flags* are used as the return code if 253543e74c02SToke Høiland-Jørgensen * the map lookup fails. This is so that the return value can be 2536ab8d7809SQuentin Monnet * one of the XDP program return codes up to **XDP_TX**, as chosen 2537ab8d7809SQuentin Monnet * by the caller. Any higher bits in the *flags* argument must be 253843e74c02SToke Høiland-Jørgensen * unset. 2539ab127040SQuentin Monnet * 2540ab8d7809SQuentin Monnet * See also **bpf_redirect**\ (), which only supports redirecting 2541ab8d7809SQuentin Monnet * to an ifindex, but doesn't require a map to do so. 2542ab127040SQuentin Monnet * Return 2543f25975f4SToke Høiland-Jørgensen * **XDP_REDIRECT** on success, or the value of the two lower bits 2544a33d3147SJakub Wilk * of the *flags* argument on error. 2545ab127040SQuentin Monnet * 2546bdb7b79bSAndrii Nakryiko * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2547ab127040SQuentin Monnet * Description 2548ab127040SQuentin Monnet * Redirect the packet to the socket referenced by *map* (of type 2549ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2550ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 2551ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 2552ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 2553ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 2554ab127040SQuentin Monnet * Return 2555ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 2556ab127040SQuentin Monnet * 2557bdb7b79bSAndrii Nakryiko * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2558ab127040SQuentin Monnet * Description 2559ab127040SQuentin Monnet * Add an entry to, or update a *map* referencing sockets. The 2560ab127040SQuentin Monnet * *skops* is used as a new value for the entry associated to 2561ab127040SQuentin Monnet * *key*. *flags* is one of: 2562ab127040SQuentin Monnet * 2563ab127040SQuentin Monnet * **BPF_NOEXIST** 2564ab127040SQuentin Monnet * The entry for *key* must not exist in the map. 2565ab127040SQuentin Monnet * **BPF_EXIST** 2566ab127040SQuentin Monnet * The entry for *key* must already exist in the map. 2567ab127040SQuentin Monnet * **BPF_ANY** 2568ab127040SQuentin Monnet * No condition on the existence of the entry for *key*. 2569ab127040SQuentin Monnet * 2570ab127040SQuentin Monnet * If the *map* has eBPF programs (parser and verdict), those will 2571ab127040SQuentin Monnet * be inherited by the socket being added. If the socket is 2572ab127040SQuentin Monnet * already attached to eBPF programs, this results in an error. 2573ab127040SQuentin Monnet * Return 2574ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 2575ab127040SQuentin Monnet * 2576bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2577fa15601aSQuentin Monnet * Description 2578fa15601aSQuentin Monnet * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2579fa15601aSQuentin Monnet * *delta* (which can be positive or negative). Note that this 2580fa15601aSQuentin Monnet * operation modifies the address stored in *xdp_md*\ **->data**, 2581fa15601aSQuentin Monnet * so the latter must be loaded only after the helper has been 2582fa15601aSQuentin Monnet * called. 2583fa15601aSQuentin Monnet * 2584fa15601aSQuentin Monnet * The use of *xdp_md*\ **->data_meta** is optional and programs 2585fa15601aSQuentin Monnet * are not required to use it. The rationale is that when the 2586fa15601aSQuentin Monnet * packet is processed with XDP (e.g. as DoS filter), it is 2587fa15601aSQuentin Monnet * possible to push further meta data along with it before passing 2588fa15601aSQuentin Monnet * to the stack, and to give the guarantee that an ingress eBPF 2589fa15601aSQuentin Monnet * program attached as a TC classifier on the same device can pick 2590fa15601aSQuentin Monnet * this up for further post-processing. Since TC works with socket 2591fa15601aSQuentin Monnet * buffers, it remains possible to set from XDP the **mark** or 2592fa15601aSQuentin Monnet * **priority** pointers, or other pointers for the socket buffer. 2593fa15601aSQuentin Monnet * Having this scratch space generic and programmable allows for 2594fa15601aSQuentin Monnet * more flexibility as the user is free to store whatever meta 2595fa15601aSQuentin Monnet * data they need. 2596fa15601aSQuentin Monnet * 259732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2598fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2599fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2600fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2601fa15601aSQuentin Monnet * direct packet access. 2602fa15601aSQuentin Monnet * Return 2603fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 26047aa79a86SQuentin Monnet * 2605bdb7b79bSAndrii Nakryiko * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 26067aa79a86SQuentin Monnet * Description 26077aa79a86SQuentin Monnet * Read the value of a perf event counter, and store it into *buf* 26087aa79a86SQuentin Monnet * of size *buf_size*. This helper relies on a *map* of type 26097aa79a86SQuentin Monnet * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 26107aa79a86SQuentin Monnet * counter is selected when *map* is updated with perf event file 26117aa79a86SQuentin Monnet * descriptors. The *map* is an array whose size is the number of 26127aa79a86SQuentin Monnet * available CPUs, and each cell contains a value relative to one 26137aa79a86SQuentin Monnet * CPU. The value to retrieve is indicated by *flags*, that 26147aa79a86SQuentin Monnet * contains the index of the CPU to look up, masked with 26157aa79a86SQuentin Monnet * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 26167aa79a86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 26177aa79a86SQuentin Monnet * current CPU should be retrieved. 26187aa79a86SQuentin Monnet * 26197aa79a86SQuentin Monnet * This helper behaves in a way close to 26207aa79a86SQuentin Monnet * **bpf_perf_event_read**\ () helper, save that instead of 26217aa79a86SQuentin Monnet * just returning the value observed, it fills the *buf* 26227aa79a86SQuentin Monnet * structure. This allows for additional data to be retrieved: in 26237aa79a86SQuentin Monnet * particular, the enabled and running times (in *buf*\ 26247aa79a86SQuentin Monnet * **->enabled** and *buf*\ **->running**, respectively) are 26257aa79a86SQuentin Monnet * copied. In general, **bpf_perf_event_read_value**\ () is 26267aa79a86SQuentin Monnet * recommended over **bpf_perf_event_read**\ (), which has some 26277aa79a86SQuentin Monnet * ABI issues and provides fewer functionalities. 26287aa79a86SQuentin Monnet * 26297aa79a86SQuentin Monnet * These values are interesting, because hardware PMU (Performance 26307aa79a86SQuentin Monnet * Monitoring Unit) counters are limited resources. When there are 26317aa79a86SQuentin Monnet * more PMU based perf events opened than available counters, 26327aa79a86SQuentin Monnet * kernel will multiplex these events so each event gets certain 26337aa79a86SQuentin Monnet * percentage (but not all) of the PMU time. In case that 26347aa79a86SQuentin Monnet * multiplexing happens, the number of samples or counter value 26357aa79a86SQuentin Monnet * will not reflect the case compared to when no multiplexing 26367aa79a86SQuentin Monnet * occurs. This makes comparison between different runs difficult. 26377aa79a86SQuentin Monnet * Typically, the counter value should be normalized before 26387aa79a86SQuentin Monnet * comparing to other experiments. The usual normalization is done 26397aa79a86SQuentin Monnet * as follows. 26407aa79a86SQuentin Monnet * 26417aa79a86SQuentin Monnet * :: 26427aa79a86SQuentin Monnet * 26437aa79a86SQuentin Monnet * normalized_counter = counter * t_enabled / t_running 26447aa79a86SQuentin Monnet * 26457aa79a86SQuentin Monnet * Where t_enabled is the time enabled for event and t_running is 26467aa79a86SQuentin Monnet * the time running for event since last normalization. The 26477aa79a86SQuentin Monnet * enabled and running times are accumulated since the perf event 26487aa79a86SQuentin Monnet * open. To achieve scaling factor between two invocations of an 2649ab8d7809SQuentin Monnet * eBPF program, users can use CPU id as the key (which is 26507aa79a86SQuentin Monnet * typical for perf array usage model) to remember the previous 26517aa79a86SQuentin Monnet * value and do the calculation inside the eBPF program. 26527aa79a86SQuentin Monnet * Return 26537aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 26547aa79a86SQuentin Monnet * 2655bdb7b79bSAndrii Nakryiko * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 26567aa79a86SQuentin Monnet * Description 26577aa79a86SQuentin Monnet * For en eBPF program attached to a perf event, retrieve the 26587aa79a86SQuentin Monnet * value of the event counter associated to *ctx* and store it in 26597aa79a86SQuentin Monnet * the structure pointed by *buf* and of size *buf_size*. Enabled 26607aa79a86SQuentin Monnet * and running times are also stored in the structure (see 26617aa79a86SQuentin Monnet * description of helper **bpf_perf_event_read_value**\ () for 26627aa79a86SQuentin Monnet * more details). 26637aa79a86SQuentin Monnet * Return 26647aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 26657aa79a86SQuentin Monnet * 2666bdb7b79bSAndrii Nakryiko * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 26677aa79a86SQuentin Monnet * Description 26687aa79a86SQuentin Monnet * Emulate a call to **getsockopt()** on the socket associated to 26697aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 26707aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 26717aa79a86SQuentin Monnet * must be specified, see **getsockopt(2)** for more information. 26727aa79a86SQuentin Monnet * The retrieved value is stored in the structure pointed by 26737aa79a86SQuentin Monnet * *opval* and of length *optlen*. 26747aa79a86SQuentin Monnet * 2675beecf11bSStanislav Fomichev * *bpf_socket* should be one of the following: 2676ab8d7809SQuentin Monnet * 2677beecf11bSStanislav Fomichev * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2678beecf11bSStanislav Fomichev * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2679beecf11bSStanislav Fomichev * and **BPF_CGROUP_INET6_CONNECT**. 2680beecf11bSStanislav Fomichev * 26817aa79a86SQuentin Monnet * This helper actually implements a subset of **getsockopt()**. 26827aa79a86SQuentin Monnet * It supports the following *level*\ s: 26837aa79a86SQuentin Monnet * 26847aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports *optname* 26857aa79a86SQuentin Monnet * **TCP_CONGESTION**. 26867aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 26877aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 26887aa79a86SQuentin Monnet * Return 26897aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 26907aa79a86SQuentin Monnet * 2691bdb7b79bSAndrii Nakryiko * long bpf_override_return(struct pt_regs *regs, u64 rc) 26927aa79a86SQuentin Monnet * Description 26937aa79a86SQuentin Monnet * Used for error injection, this helper uses kprobes to override 26947aa79a86SQuentin Monnet * the return value of the probed function, and to set it to *rc*. 26957aa79a86SQuentin Monnet * The first argument is the context *regs* on which the kprobe 26967aa79a86SQuentin Monnet * works. 26977aa79a86SQuentin Monnet * 2698ab8d7809SQuentin Monnet * This helper works by setting the PC (program counter) 26997aa79a86SQuentin Monnet * to an override function which is run in place of the original 27007aa79a86SQuentin Monnet * probed function. This means the probed function is not run at 27017aa79a86SQuentin Monnet * all. The replacement function just returns with the required 27027aa79a86SQuentin Monnet * value. 27037aa79a86SQuentin Monnet * 27047aa79a86SQuentin Monnet * This helper has security implications, and thus is subject to 27057aa79a86SQuentin Monnet * restrictions. It is only available if the kernel was compiled 27067aa79a86SQuentin Monnet * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 27077aa79a86SQuentin Monnet * option, and in this case it only works on functions tagged with 27087aa79a86SQuentin Monnet * **ALLOW_ERROR_INJECTION** in the kernel code. 27097aa79a86SQuentin Monnet * 27107aa79a86SQuentin Monnet * Also, the helper is only available for the architectures having 27117aa79a86SQuentin Monnet * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 27127aa79a86SQuentin Monnet * x86 architecture is the only one to support this feature. 27137aa79a86SQuentin Monnet * Return 27147aa79a86SQuentin Monnet * 0 27157aa79a86SQuentin Monnet * 2716bdb7b79bSAndrii Nakryiko * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 27177aa79a86SQuentin Monnet * Description 27187aa79a86SQuentin Monnet * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 27197aa79a86SQuentin Monnet * for the full TCP socket associated to *bpf_sock_ops* to 27207aa79a86SQuentin Monnet * *argval*. 27217aa79a86SQuentin Monnet * 27227aa79a86SQuentin Monnet * The primary use of this field is to determine if there should 27237aa79a86SQuentin Monnet * be calls to eBPF programs of type 27247aa79a86SQuentin Monnet * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 27257aa79a86SQuentin Monnet * code. A program of the same type can change its value, per 27267aa79a86SQuentin Monnet * connection and as necessary, when the connection is 27277aa79a86SQuentin Monnet * established. This field is directly accessible for reading, but 27287aa79a86SQuentin Monnet * this helper must be used for updates in order to return an 27297aa79a86SQuentin Monnet * error if an eBPF program tries to set a callback that is not 27307aa79a86SQuentin Monnet * supported in the current kernel. 27317aa79a86SQuentin Monnet * 2732725721a6SViet Hoang Tran * *argval* is a flag array which can combine these flags: 27337aa79a86SQuentin Monnet * 27347aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 27357aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 27367aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 273723729ff2SStanislav Fomichev * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 27387aa79a86SQuentin Monnet * 2739725721a6SViet Hoang Tran * Therefore, this function can be used to clear a callback flag by 2740725721a6SViet Hoang Tran * setting the appropriate bit to zero. e.g. to disable the RTO 2741725721a6SViet Hoang Tran * callback: 2742725721a6SViet Hoang Tran * 2743725721a6SViet Hoang Tran * **bpf_sock_ops_cb_flags_set(bpf_sock,** 2744725721a6SViet Hoang Tran * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 2745725721a6SViet Hoang Tran * 27467aa79a86SQuentin Monnet * Here are some examples of where one could call such eBPF 27477aa79a86SQuentin Monnet * program: 27487aa79a86SQuentin Monnet * 27497aa79a86SQuentin Monnet * * When RTO fires. 27507aa79a86SQuentin Monnet * * When a packet is retransmitted. 27517aa79a86SQuentin Monnet * * When the connection terminates. 27527aa79a86SQuentin Monnet * * When a packet is sent. 27537aa79a86SQuentin Monnet * * When a packet is received. 27547aa79a86SQuentin Monnet * Return 27557aa79a86SQuentin Monnet * Code **-EINVAL** if the socket is not a full TCP socket; 27567aa79a86SQuentin Monnet * otherwise, a positive number containing the bits that could not 27577aa79a86SQuentin Monnet * be set is returned (which comes down to 0 if all bits were set 27587aa79a86SQuentin Monnet * as required). 27597aa79a86SQuentin Monnet * 2760bdb7b79bSAndrii Nakryiko * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 2761ab127040SQuentin Monnet * Description 2762ab127040SQuentin Monnet * This helper is used in programs implementing policies at the 2763ab127040SQuentin Monnet * socket level. If the message *msg* is allowed to pass (i.e. if 2764ab127040SQuentin Monnet * the verdict eBPF program returns **SK_PASS**), redirect it to 2765ab127040SQuentin Monnet * the socket referenced by *map* (of type 2766ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2767ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 2768ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 2769ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 2770ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 2771ab127040SQuentin Monnet * Return 2772ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 2773ab127040SQuentin Monnet * 2774bdb7b79bSAndrii Nakryiko * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 2775ab127040SQuentin Monnet * Description 2776ab127040SQuentin Monnet * For socket policies, apply the verdict of the eBPF program to 2777ab127040SQuentin Monnet * the next *bytes* (number of bytes) of message *msg*. 2778ab127040SQuentin Monnet * 2779ab127040SQuentin Monnet * For example, this helper can be used in the following cases: 2780ab127040SQuentin Monnet * 2781ab127040SQuentin Monnet * * A single **sendmsg**\ () or **sendfile**\ () system call 2782ab127040SQuentin Monnet * contains multiple logical messages that the eBPF program is 2783ab127040SQuentin Monnet * supposed to read and for which it should apply a verdict. 2784ab127040SQuentin Monnet * * An eBPF program only cares to read the first *bytes* of a 2785ab127040SQuentin Monnet * *msg*. If the message has a large payload, then setting up 2786ab127040SQuentin Monnet * and calling the eBPF program repeatedly for all bytes, even 2787ab127040SQuentin Monnet * though the verdict is already known, would create unnecessary 2788ab127040SQuentin Monnet * overhead. 2789ab127040SQuentin Monnet * 2790ab127040SQuentin Monnet * When called from within an eBPF program, the helper sets a 2791ab127040SQuentin Monnet * counter internal to the BPF infrastructure, that is used to 2792ab127040SQuentin Monnet * apply the last verdict to the next *bytes*. If *bytes* is 2793ab127040SQuentin Monnet * smaller than the current data being processed from a 2794ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () system call, the first 2795ab127040SQuentin Monnet * *bytes* will be sent and the eBPF program will be re-run with 2796ab127040SQuentin Monnet * the pointer for start of data pointing to byte number *bytes* 2797ab127040SQuentin Monnet * **+ 1**. If *bytes* is larger than the current data being 2798ab127040SQuentin Monnet * processed, then the eBPF verdict will be applied to multiple 2799ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2800ab127040SQuentin Monnet * consumed. 2801ab127040SQuentin Monnet * 2802ab127040SQuentin Monnet * Note that if a socket closes with the internal counter holding 2803ab127040SQuentin Monnet * a non-zero value, this is not a problem because data is not 2804ab127040SQuentin Monnet * being buffered for *bytes* and is sent as it is received. 2805ab127040SQuentin Monnet * Return 2806ab127040SQuentin Monnet * 0 2807ab127040SQuentin Monnet * 2808bdb7b79bSAndrii Nakryiko * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 2809ab127040SQuentin Monnet * Description 2810ab127040SQuentin Monnet * For socket policies, prevent the execution of the verdict eBPF 2811ab127040SQuentin Monnet * program for message *msg* until *bytes* (byte number) have been 2812ab127040SQuentin Monnet * accumulated. 2813ab127040SQuentin Monnet * 2814ab127040SQuentin Monnet * This can be used when one needs a specific number of bytes 2815ab127040SQuentin Monnet * before a verdict can be assigned, even if the data spans 2816ab127040SQuentin Monnet * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 2817ab127040SQuentin Monnet * case would be a user calling **sendmsg**\ () repeatedly with 2818ab127040SQuentin Monnet * 1-byte long message segments. Obviously, this is bad for 2819ab127040SQuentin Monnet * performance, but it is still valid. If the eBPF program needs 2820ab127040SQuentin Monnet * *bytes* bytes to validate a header, this helper can be used to 2821ab127040SQuentin Monnet * prevent the eBPF program to be called again until *bytes* have 2822ab127040SQuentin Monnet * been accumulated. 2823ab127040SQuentin Monnet * Return 2824ab127040SQuentin Monnet * 0 2825ab127040SQuentin Monnet * 2826bdb7b79bSAndrii Nakryiko * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2827ab127040SQuentin Monnet * Description 2828ab127040SQuentin Monnet * For socket policies, pull in non-linear data from user space 2829ab127040SQuentin Monnet * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2830ab127040SQuentin Monnet * **->data_end** to *start* and *end* bytes offsets into *msg*, 2831ab127040SQuentin Monnet * respectively. 2832ab127040SQuentin Monnet * 2833ab127040SQuentin Monnet * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2834ab127040SQuentin Monnet * *msg* it can only parse data that the (**data**, **data_end**) 2835ab127040SQuentin Monnet * pointers have already consumed. For **sendmsg**\ () hooks this 2836ab127040SQuentin Monnet * is likely the first scatterlist element. But for calls relying 2837ab127040SQuentin Monnet * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2838ab127040SQuentin Monnet * be the range (**0**, **0**) because the data is shared with 2839ab127040SQuentin Monnet * user space and by default the objective is to avoid allowing 2840ab127040SQuentin Monnet * user space to modify data while (or after) eBPF verdict is 2841ab127040SQuentin Monnet * being decided. This helper can be used to pull in data and to 2842ab127040SQuentin Monnet * set the start and end pointer to given values. Data will be 2843ab127040SQuentin Monnet * copied if necessary (i.e. if data was not linear and if start 2844ab127040SQuentin Monnet * and end pointers do not point to the same chunk). 2845ab127040SQuentin Monnet * 284632e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2847ab127040SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2848ab127040SQuentin Monnet * previously done by the verifier are invalidated and must be 2849ab127040SQuentin Monnet * performed again, if the helper is used in combination with 2850ab127040SQuentin Monnet * direct packet access. 2851ab127040SQuentin Monnet * 2852ab127040SQuentin Monnet * All values for *flags* are reserved for future usage, and must 2853ab127040SQuentin Monnet * be left at zero. 2854ab127040SQuentin Monnet * Return 2855ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 2856ab127040SQuentin Monnet * 2857bdb7b79bSAndrii Nakryiko * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 28587aa79a86SQuentin Monnet * Description 28597aa79a86SQuentin Monnet * Bind the socket associated to *ctx* to the address pointed by 28607aa79a86SQuentin Monnet * *addr*, of length *addr_len*. This allows for making outgoing 28617aa79a86SQuentin Monnet * connection from the desired IP address, which can be useful for 28627aa79a86SQuentin Monnet * example when all processes inside a cgroup should use one 28637aa79a86SQuentin Monnet * single IP address on a host that has multiple IP configured. 28647aa79a86SQuentin Monnet * 28657aa79a86SQuentin Monnet * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 28667aa79a86SQuentin Monnet * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 28678086fbafSStanislav Fomichev * **AF_INET6**). It's advised to pass zero port (**sin_port** 28688086fbafSStanislav Fomichev * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 28698086fbafSStanislav Fomichev * behavior and lets the kernel efficiently pick up an unused 28708086fbafSStanislav Fomichev * port as long as 4-tuple is unique. Passing non-zero port might 28718086fbafSStanislav Fomichev * lead to degraded performance. 28727aa79a86SQuentin Monnet * Return 28737aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 28742d020dd7SQuentin Monnet * 2875bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 28762d020dd7SQuentin Monnet * Description 28772d020dd7SQuentin Monnet * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2878c8741e2bSJesper Dangaard Brouer * possible to both shrink and grow the packet tail. 2879c8741e2bSJesper Dangaard Brouer * Shrink done via *delta* being a negative integer. 28802d020dd7SQuentin Monnet * 288132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 28822d020dd7SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 28832d020dd7SQuentin Monnet * previously done by the verifier are invalidated and must be 28842d020dd7SQuentin Monnet * performed again, if the helper is used in combination with 28852d020dd7SQuentin Monnet * direct packet access. 28862d020dd7SQuentin Monnet * Return 28872d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 28882d020dd7SQuentin Monnet * 2889bdb7b79bSAndrii Nakryiko * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 28902d020dd7SQuentin Monnet * Description 28912d020dd7SQuentin Monnet * Retrieve the XFRM state (IP transform framework, see also 28922d020dd7SQuentin Monnet * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 28932d020dd7SQuentin Monnet * 28942d020dd7SQuentin Monnet * The retrieved value is stored in the **struct bpf_xfrm_state** 28952d020dd7SQuentin Monnet * pointed by *xfrm_state* and of length *size*. 28962d020dd7SQuentin Monnet * 28972d020dd7SQuentin Monnet * All values for *flags* are reserved for future usage, and must 28982d020dd7SQuentin Monnet * be left at zero. 28992d020dd7SQuentin Monnet * 29002d020dd7SQuentin Monnet * This helper is available only if the kernel was compiled with 29012d020dd7SQuentin Monnet * **CONFIG_XFRM** configuration option. 29022d020dd7SQuentin Monnet * Return 29032d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 2904c195651eSYonghong Song * 2905bdb7b79bSAndrii Nakryiko * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2906c195651eSYonghong Song * Description 2907c195651eSYonghong Song * Return a user or a kernel stack in bpf program provided buffer. 2908c195651eSYonghong Song * To achieve this, the helper needs *ctx*, which is a pointer 2909c195651eSYonghong Song * to the context on which the tracing program is executed. 2910c195651eSYonghong Song * To store the stacktrace, the bpf program provides *buf* with 2911c195651eSYonghong Song * a nonnegative *size*. 2912c195651eSYonghong Song * 2913c195651eSYonghong Song * The last argument, *flags*, holds the number of stack frames to 2914c195651eSYonghong Song * skip (from 0 to 255), masked with 2915c195651eSYonghong Song * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2916c195651eSYonghong Song * the following flags: 2917c195651eSYonghong Song * 2918c195651eSYonghong Song * **BPF_F_USER_STACK** 2919c195651eSYonghong Song * Collect a user space stack instead of a kernel stack. 2920c195651eSYonghong Song * **BPF_F_USER_BUILD_ID** 2921c195651eSYonghong Song * Collect buildid+offset instead of ips for user stack, 2922c195651eSYonghong Song * only valid if **BPF_F_USER_STACK** is also specified. 2923c195651eSYonghong Song * 2924c195651eSYonghong Song * **bpf_get_stack**\ () can collect up to 2925c195651eSYonghong Song * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 2926c195651eSYonghong Song * to sufficient large buffer size. Note that 2927c195651eSYonghong Song * this limit can be controlled with the **sysctl** program, and 2928c195651eSYonghong Song * that it should be manually increased in order to profile long 2929c195651eSYonghong Song * user stacks (such as stacks for Java programs). To do so, use: 2930c195651eSYonghong Song * 2931c195651eSYonghong Song * :: 2932c195651eSYonghong Song * 2933c195651eSYonghong Song * # sysctl kernel.perf_event_max_stack=<new value> 2934c195651eSYonghong Song * Return 29357a279e93SQuentin Monnet * A non-negative value equal to or less than *size* on success, 29367a279e93SQuentin Monnet * or a negative error in case of failure. 29374e1ec56cSDaniel Borkmann * 2938bdb7b79bSAndrii Nakryiko * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 29394e1ec56cSDaniel Borkmann * Description 29404e1ec56cSDaniel Borkmann * This helper is similar to **bpf_skb_load_bytes**\ () in that 29414e1ec56cSDaniel Borkmann * it provides an easy way to load *len* bytes from *offset* 29424e1ec56cSDaniel Borkmann * from the packet associated to *skb*, into the buffer pointed 29434e1ec56cSDaniel Borkmann * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 29444e1ec56cSDaniel Borkmann * a fifth argument *start_header* exists in order to select a 29454e1ec56cSDaniel Borkmann * base offset to start from. *start_header* can be one of: 29464e1ec56cSDaniel Borkmann * 29474e1ec56cSDaniel Borkmann * **BPF_HDR_START_MAC** 29484e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s mac header. 29494e1ec56cSDaniel Borkmann * **BPF_HDR_START_NET** 29504e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s network header. 29514e1ec56cSDaniel Borkmann * 29524e1ec56cSDaniel Borkmann * In general, "direct packet access" is the preferred method to 29534e1ec56cSDaniel Borkmann * access packet data, however, this helper is in particular useful 29544e1ec56cSDaniel Borkmann * in socket filters where *skb*\ **->data** does not always point 29554e1ec56cSDaniel Borkmann * to the start of the mac header and where "direct packet access" 29564e1ec56cSDaniel Borkmann * is not available. 29574e1ec56cSDaniel Borkmann * Return 29584e1ec56cSDaniel Borkmann * 0 on success, or a negative error in case of failure. 29594e1ec56cSDaniel Borkmann * 2960bdb7b79bSAndrii Nakryiko * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 296187f5fc7eSDavid Ahern * Description 296287f5fc7eSDavid Ahern * Do FIB lookup in kernel tables using parameters in *params*. 296387f5fc7eSDavid Ahern * If lookup is successful and result shows packet is to be 296487f5fc7eSDavid Ahern * forwarded, the neighbor tables are searched for the nexthop. 296587f5fc7eSDavid Ahern * If successful (ie., FIB lookup shows forwarding and nexthop 2966fa898d76SDavid Ahern * is resolved), the nexthop address is returned in ipv4_dst 2967fa898d76SDavid Ahern * or ipv6_dst based on family, smac is set to mac address of 2968fa898d76SDavid Ahern * egress device, dmac is set to nexthop mac address, rt_metric 29694c79579bSDavid Ahern * is set to metric from route (IPv4/IPv6 only), and ifindex 29704c79579bSDavid Ahern * is set to the device index of the nexthop from the FIB lookup. 297187f5fc7eSDavid Ahern * 297287f5fc7eSDavid Ahern * *plen* argument is the size of the passed in struct. 29737a279e93SQuentin Monnet * *flags* argument can be a combination of one or more of the 29747a279e93SQuentin Monnet * following values: 297587f5fc7eSDavid Ahern * 29767a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_DIRECT** 29777a279e93SQuentin Monnet * Do a direct table lookup vs full lookup using FIB 29787a279e93SQuentin Monnet * rules. 29797a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_OUTPUT** 29807a279e93SQuentin Monnet * Perform lookup from an egress perspective (default is 29817a279e93SQuentin Monnet * ingress). 298287f5fc7eSDavid Ahern * 298387f5fc7eSDavid Ahern * *ctx* is either **struct xdp_md** for XDP programs or 298487f5fc7eSDavid Ahern * **struct sk_buff** tc cls_act programs. 298587f5fc7eSDavid Ahern * Return 29864c79579bSDavid Ahern * * < 0 if any input argument is invalid 29874c79579bSDavid Ahern * * 0 on success (packet is forwarded, nexthop neighbor exists) 29884c79579bSDavid Ahern * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 29892bae79d2SQuentin Monnet * packet is not forwarded or needs assist from full stack 299081110384SJohn Fastabend * 2991e1850ea9SJesper Dangaard Brouer * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 2992e1850ea9SJesper Dangaard Brouer * was exceeded and output params->mtu_result contains the MTU. 2993e1850ea9SJesper Dangaard Brouer * 2994bdb7b79bSAndrii Nakryiko * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 299581110384SJohn Fastabend * Description 299681110384SJohn Fastabend * Add an entry to, or update a sockhash *map* referencing sockets. 299781110384SJohn Fastabend * The *skops* is used as a new value for the entry associated to 299881110384SJohn Fastabend * *key*. *flags* is one of: 299981110384SJohn Fastabend * 300081110384SJohn Fastabend * **BPF_NOEXIST** 300181110384SJohn Fastabend * The entry for *key* must not exist in the map. 300281110384SJohn Fastabend * **BPF_EXIST** 300381110384SJohn Fastabend * The entry for *key* must already exist in the map. 300481110384SJohn Fastabend * **BPF_ANY** 300581110384SJohn Fastabend * No condition on the existence of the entry for *key*. 300681110384SJohn Fastabend * 300781110384SJohn Fastabend * If the *map* has eBPF programs (parser and verdict), those will 300881110384SJohn Fastabend * be inherited by the socket being added. If the socket is 300981110384SJohn Fastabend * already attached to eBPF programs, this results in an error. 301081110384SJohn Fastabend * Return 301181110384SJohn Fastabend * 0 on success, or a negative error in case of failure. 301281110384SJohn Fastabend * 3013bdb7b79bSAndrii Nakryiko * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 301481110384SJohn Fastabend * Description 301581110384SJohn Fastabend * This helper is used in programs implementing policies at the 301681110384SJohn Fastabend * socket level. If the message *msg* is allowed to pass (i.e. if 301781110384SJohn Fastabend * the verdict eBPF program returns **SK_PASS**), redirect it to 301881110384SJohn Fastabend * the socket referenced by *map* (of type 301981110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 302081110384SJohn Fastabend * egress interfaces can be used for redirection. The 302181110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 302281110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 302381110384SJohn Fastabend * egress path otherwise). This is the only flag supported for now. 302481110384SJohn Fastabend * Return 302581110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 302681110384SJohn Fastabend * 3027bdb7b79bSAndrii Nakryiko * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 302881110384SJohn Fastabend * Description 302981110384SJohn Fastabend * This helper is used in programs implementing policies at the 303081110384SJohn Fastabend * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 303149f3d12bSJakub Wilk * if the verdict eBPF program returns **SK_PASS**), redirect it 303281110384SJohn Fastabend * to the socket referenced by *map* (of type 303381110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 303481110384SJohn Fastabend * egress interfaces can be used for redirection. The 303581110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 303681110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 303781110384SJohn Fastabend * egress otherwise). This is the only flag supported for now. 303881110384SJohn Fastabend * Return 303981110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 3040fe94cc29SMathieu Xhonneux * 3041bdb7b79bSAndrii Nakryiko * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3042fe94cc29SMathieu Xhonneux * Description 3043fe94cc29SMathieu Xhonneux * Encapsulate the packet associated to *skb* within a Layer 3 3044fe94cc29SMathieu Xhonneux * protocol header. This header is provided in the buffer at 3045fe94cc29SMathieu Xhonneux * address *hdr*, with *len* its size in bytes. *type* indicates 3046fe94cc29SMathieu Xhonneux * the protocol of the header and can be one of: 3047fe94cc29SMathieu Xhonneux * 3048fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6** 3049fe94cc29SMathieu Xhonneux * IPv6 encapsulation with Segment Routing Header 3050fe94cc29SMathieu Xhonneux * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3051fe94cc29SMathieu Xhonneux * the IPv6 header is computed by the kernel. 3052fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6_INLINE** 3053fe94cc29SMathieu Xhonneux * Only works if *skb* contains an IPv6 packet. Insert a 3054fe94cc29SMathieu Xhonneux * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3055fe94cc29SMathieu Xhonneux * the IPv6 header. 30563e0bd37cSPeter Oskolkov * **BPF_LWT_ENCAP_IP** 30573e0bd37cSPeter Oskolkov * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 30583e0bd37cSPeter Oskolkov * must be IPv4 or IPv6, followed by zero or more 305980867c5eSQuentin Monnet * additional headers, up to **LWT_BPF_MAX_HEADROOM** 306080867c5eSQuentin Monnet * total bytes in all prepended headers. Please note that 306180867c5eSQuentin Monnet * if **skb_is_gso**\ (*skb*) is true, no more than two 306280867c5eSQuentin Monnet * headers can be prepended, and the inner header, if 306380867c5eSQuentin Monnet * present, should be either GRE or UDP/GUE. 30643e0bd37cSPeter Oskolkov * 306580867c5eSQuentin Monnet * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 306680867c5eSQuentin Monnet * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 306780867c5eSQuentin Monnet * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 306880867c5eSQuentin Monnet * **BPF_PROG_TYPE_LWT_XMIT**. 3069fe94cc29SMathieu Xhonneux * 307032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3071fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3072fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3073fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3074fe94cc29SMathieu Xhonneux * direct packet access. 3075fe94cc29SMathieu Xhonneux * Return 3076fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3077fe94cc29SMathieu Xhonneux * 3078bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3079fe94cc29SMathieu Xhonneux * Description 3080fe94cc29SMathieu Xhonneux * Store *len* bytes from address *from* into the packet 3081fe94cc29SMathieu Xhonneux * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3082fe94cc29SMathieu Xhonneux * inside the outermost IPv6 Segment Routing Header can be 3083fe94cc29SMathieu Xhonneux * modified through this helper. 3084fe94cc29SMathieu Xhonneux * 308532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3086fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3087fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3088fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3089fe94cc29SMathieu Xhonneux * direct packet access. 3090fe94cc29SMathieu Xhonneux * Return 3091fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3092fe94cc29SMathieu Xhonneux * 3093bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3094fe94cc29SMathieu Xhonneux * Description 3095fe94cc29SMathieu Xhonneux * Adjust the size allocated to TLVs in the outermost IPv6 3096fe94cc29SMathieu Xhonneux * Segment Routing Header contained in the packet associated to 3097fe94cc29SMathieu Xhonneux * *skb*, at position *offset* by *delta* bytes. Only offsets 3098fe94cc29SMathieu Xhonneux * after the segments are accepted. *delta* can be as well 3099fe94cc29SMathieu Xhonneux * positive (growing) as negative (shrinking). 3100fe94cc29SMathieu Xhonneux * 310132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3102fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3103fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3104fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3105fe94cc29SMathieu Xhonneux * direct packet access. 3106fe94cc29SMathieu Xhonneux * Return 3107fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3108fe94cc29SMathieu Xhonneux * 3109bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3110fe94cc29SMathieu Xhonneux * Description 3111fe94cc29SMathieu Xhonneux * Apply an IPv6 Segment Routing action of type *action* to the 3112fe94cc29SMathieu Xhonneux * packet associated to *skb*. Each action takes a parameter 3113fe94cc29SMathieu Xhonneux * contained at address *param*, and of length *param_len* bytes. 3114fe94cc29SMathieu Xhonneux * *action* can be one of: 3115fe94cc29SMathieu Xhonneux * 3116fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_X** 3117fe94cc29SMathieu Xhonneux * End.X action: Endpoint with Layer-3 cross-connect. 3118fe94cc29SMathieu Xhonneux * Type of *param*: **struct in6_addr**. 3119fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_T** 3120fe94cc29SMathieu Xhonneux * End.T action: Endpoint with specific IPv6 table lookup. 3121fe94cc29SMathieu Xhonneux * Type of *param*: **int**. 3122fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6** 3123fe94cc29SMathieu Xhonneux * End.B6 action: Endpoint bound to an SRv6 policy. 312480867c5eSQuentin Monnet * Type of *param*: **struct ipv6_sr_hdr**. 3125fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3126fe94cc29SMathieu Xhonneux * End.B6.Encap action: Endpoint bound to an SRv6 3127fe94cc29SMathieu Xhonneux * encapsulation policy. 312880867c5eSQuentin Monnet * Type of *param*: **struct ipv6_sr_hdr**. 3129fe94cc29SMathieu Xhonneux * 313032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3131fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3132fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3133fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3134fe94cc29SMathieu Xhonneux * direct packet access. 3135fe94cc29SMathieu Xhonneux * Return 3136fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3137f4364dcfSSean Young * 3138bdb7b79bSAndrii Nakryiko * long bpf_rc_repeat(void *ctx) 313962369db2SQuentin Monnet * Description 314062369db2SQuentin Monnet * This helper is used in programs implementing IR decoding, to 314162369db2SQuentin Monnet * report a successfully decoded repeat key message. This delays 314262369db2SQuentin Monnet * the generation of a key up event for previously generated 314362369db2SQuentin Monnet * key down event. 314462369db2SQuentin Monnet * 314562369db2SQuentin Monnet * Some IR protocols like NEC have a special IR message for 314662369db2SQuentin Monnet * repeating last button, for when a button is held down. 314762369db2SQuentin Monnet * 314862369db2SQuentin Monnet * The *ctx* should point to the lirc sample as passed into 314962369db2SQuentin Monnet * the program. 315062369db2SQuentin Monnet * 315162369db2SQuentin Monnet * This helper is only available is the kernel was compiled with 315262369db2SQuentin Monnet * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 315362369db2SQuentin Monnet * "**y**". 315462369db2SQuentin Monnet * Return 315562369db2SQuentin Monnet * 0 315662369db2SQuentin Monnet * 3157bdb7b79bSAndrii Nakryiko * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3158f4364dcfSSean Young * Description 3159f4364dcfSSean Young * This helper is used in programs implementing IR decoding, to 3160f4364dcfSSean Young * report a successfully decoded key press with *scancode*, 3161f4364dcfSSean Young * *toggle* value in the given *protocol*. The scancode will be 3162f4364dcfSSean Young * translated to a keycode using the rc keymap, and reported as 3163f4364dcfSSean Young * an input key down event. After a period a key up event is 3164f4364dcfSSean Young * generated. This period can be extended by calling either 316590b1023fSQuentin Monnet * **bpf_rc_keydown**\ () again with the same values, or calling 316690b1023fSQuentin Monnet * **bpf_rc_repeat**\ (). 3167f4364dcfSSean Young * 3168f4364dcfSSean Young * Some protocols include a toggle bit, in case the button was 3169f4364dcfSSean Young * released and pressed again between consecutive scancodes. 3170f4364dcfSSean Young * 3171f4364dcfSSean Young * The *ctx* should point to the lirc sample as passed into 3172f4364dcfSSean Young * the program. 3173f4364dcfSSean Young * 3174f4364dcfSSean Young * The *protocol* is the decoded protocol number (see 3175f4364dcfSSean Young * **enum rc_proto** for some predefined values). 3176f4364dcfSSean Young * 3177f4364dcfSSean Young * This helper is only available is the kernel was compiled with 3178f4364dcfSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3179f4364dcfSSean Young * "**y**". 3180f4364dcfSSean Young * Return 3181f4364dcfSSean Young * 0 3182f4364dcfSSean Young * 318362369db2SQuentin Monnet * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3184cb20b08eSDaniel Borkmann * Description 3185cb20b08eSDaniel Borkmann * Return the cgroup v2 id of the socket associated with the *skb*. 3186cb20b08eSDaniel Borkmann * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3187cb20b08eSDaniel Borkmann * helper for cgroup v1 by providing a tag resp. identifier that 3188cb20b08eSDaniel Borkmann * can be matched on or used for map lookups e.g. to implement 3189cb20b08eSDaniel Borkmann * policy. The cgroup v2 id of a given path in the hierarchy is 3190cb20b08eSDaniel Borkmann * exposed in user space through the f_handle API in order to get 3191cb20b08eSDaniel Borkmann * to the same 64-bit id. 3192cb20b08eSDaniel Borkmann * 3193cb20b08eSDaniel Borkmann * This helper can be used on TC egress path, but not on ingress, 3194cb20b08eSDaniel Borkmann * and is available only if the kernel was compiled with the 3195cb20b08eSDaniel Borkmann * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3196cb20b08eSDaniel Borkmann * Return 3197cb20b08eSDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 3198bf6fa2c8SYonghong Song * 3199bf6fa2c8SYonghong Song * u64 bpf_get_current_cgroup_id(void) 3200bf6fa2c8SYonghong Song * Return 3201bf6fa2c8SYonghong Song * A 64-bit integer containing the current cgroup id based 3202bf6fa2c8SYonghong Song * on the cgroup within which the current task is running. 3203cd339431SRoman Gushchin * 320462369db2SQuentin Monnet * void *bpf_get_local_storage(void *map, u64 flags) 3205cd339431SRoman Gushchin * Description 3206cd339431SRoman Gushchin * Get the pointer to the local storage area. 3207cd339431SRoman Gushchin * The type and the size of the local storage is defined 3208cd339431SRoman Gushchin * by the *map* argument. 3209cd339431SRoman Gushchin * The *flags* meaning is specific for each map type, 3210cd339431SRoman Gushchin * and has to be 0 for cgroup local storage. 3211cd339431SRoman Gushchin * 321290b1023fSQuentin Monnet * Depending on the BPF program type, a local storage area 321390b1023fSQuentin Monnet * can be shared between multiple instances of the BPF program, 3214cd339431SRoman Gushchin * running simultaneously. 3215cd339431SRoman Gushchin * 3216cd339431SRoman Gushchin * A user should care about the synchronization by himself. 321791c960b0SBrendan Jackman * For example, by using the **BPF_ATOMIC** instructions to alter 3218cd339431SRoman Gushchin * the shared data. 3219cd339431SRoman Gushchin * Return 322090b1023fSQuentin Monnet * A pointer to the local storage area. 32212dbb9b9eSMartin KaFai Lau * 3222bdb7b79bSAndrii Nakryiko * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 32232dbb9b9eSMartin KaFai Lau * Description 322490b1023fSQuentin Monnet * Select a **SO_REUSEPORT** socket from a 322590b1023fSQuentin Monnet * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 322690b1023fSQuentin Monnet * It checks the selected socket is matching the incoming 322790b1023fSQuentin Monnet * request in the socket buffer. 32282dbb9b9eSMartin KaFai Lau * Return 32292dbb9b9eSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 32306acc9b43SJoe Stringer * 323162369db2SQuentin Monnet * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 323262369db2SQuentin Monnet * Description 323362369db2SQuentin Monnet * Return id of cgroup v2 that is ancestor of cgroup associated 323462369db2SQuentin Monnet * with the *skb* at the *ancestor_level*. The root cgroup is at 323562369db2SQuentin Monnet * *ancestor_level* zero and each step down the hierarchy 323662369db2SQuentin Monnet * increments the level. If *ancestor_level* == level of cgroup 323762369db2SQuentin Monnet * associated with *skb*, then return value will be same as that 323862369db2SQuentin Monnet * of **bpf_skb_cgroup_id**\ (). 323962369db2SQuentin Monnet * 324062369db2SQuentin Monnet * The helper is useful to implement policies based on cgroups 324162369db2SQuentin Monnet * that are upper in hierarchy than immediate cgroup associated 324262369db2SQuentin Monnet * with *skb*. 324362369db2SQuentin Monnet * 324462369db2SQuentin Monnet * The format of returned id and helper limitations are same as in 324562369db2SQuentin Monnet * **bpf_skb_cgroup_id**\ (). 324662369db2SQuentin Monnet * Return 324762369db2SQuentin Monnet * The id is returned or 0 in case the id could not be retrieved. 324862369db2SQuentin Monnet * 3249f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 32506acc9b43SJoe Stringer * Description 32516acc9b43SJoe Stringer * Look for TCP socket matching *tuple*, optionally in a child 32526acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 325390b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 32546acc9b43SJoe Stringer * 32556acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 32566acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 32576acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 32586acc9b43SJoe Stringer * 32596acc9b43SJoe Stringer * *tuple_size* must be one of: 32606acc9b43SJoe Stringer * 32616acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 32626acc9b43SJoe Stringer * Look for an IPv4 socket. 32636acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 32646acc9b43SJoe Stringer * Look for an IPv6 socket. 32656acc9b43SJoe Stringer * 3266f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 3267bfdfa517SRandy Dunlap * socket lookup table in the netns associated with the *ctx* 3268f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 3269f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 3270f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 3271f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 3272f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 3273f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 32746acc9b43SJoe Stringer * 32756acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 32766acc9b43SJoe Stringer * be left at zero. 32776acc9b43SJoe Stringer * 32786acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 32796acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 32806acc9b43SJoe Stringer * Return 32810bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 32820bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 328380867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 328480867c5eSQuentin Monnet * tuple. 32856acc9b43SJoe Stringer * 3286f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 32876acc9b43SJoe Stringer * Description 32886acc9b43SJoe Stringer * Look for UDP socket matching *tuple*, optionally in a child 32896acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 329090b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 32916acc9b43SJoe Stringer * 32926acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 32936acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 32946acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 32956acc9b43SJoe Stringer * 32966acc9b43SJoe Stringer * *tuple_size* must be one of: 32976acc9b43SJoe Stringer * 32986acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 32996acc9b43SJoe Stringer * Look for an IPv4 socket. 33006acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 33016acc9b43SJoe Stringer * Look for an IPv6 socket. 33026acc9b43SJoe Stringer * 3303f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 3304bfdfa517SRandy Dunlap * socket lookup table in the netns associated with the *ctx* 3305f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 3306f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 3307f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 3308f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 3309f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 3310f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 33116acc9b43SJoe Stringer * 33126acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 33136acc9b43SJoe Stringer * be left at zero. 33146acc9b43SJoe Stringer * 33156acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 33166acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 33176acc9b43SJoe Stringer * Return 33180bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 33190bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 332080867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 332180867c5eSQuentin Monnet * tuple. 33226acc9b43SJoe Stringer * 3323a5fa25adSMartin KaFai Lau * long bpf_sk_release(void *sock) 33246acc9b43SJoe Stringer * Description 332590b1023fSQuentin Monnet * Release the reference held by *sock*. *sock* must be a 332690b1023fSQuentin Monnet * non-**NULL** pointer that was returned from 332790b1023fSQuentin Monnet * **bpf_sk_lookup_xxx**\ (). 332890b1023fSQuentin Monnet * Return 332990b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 333090b1023fSQuentin Monnet * 3331bdb7b79bSAndrii Nakryiko * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 333262369db2SQuentin Monnet * Description 333362369db2SQuentin Monnet * Push an element *value* in *map*. *flags* is one of: 333462369db2SQuentin Monnet * 333562369db2SQuentin Monnet * **BPF_EXIST** 333662369db2SQuentin Monnet * If the queue/stack is full, the oldest element is 333762369db2SQuentin Monnet * removed to make room for this. 333862369db2SQuentin Monnet * Return 333962369db2SQuentin Monnet * 0 on success, or a negative error in case of failure. 334062369db2SQuentin Monnet * 3341bdb7b79bSAndrii Nakryiko * long bpf_map_pop_elem(struct bpf_map *map, void *value) 334290b1023fSQuentin Monnet * Description 334390b1023fSQuentin Monnet * Pop an element from *map*. 334490b1023fSQuentin Monnet * Return 334590b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 334690b1023fSQuentin Monnet * 3347bdb7b79bSAndrii Nakryiko * long bpf_map_peek_elem(struct bpf_map *map, void *value) 334890b1023fSQuentin Monnet * Description 334990b1023fSQuentin Monnet * Get an element from *map* without removing it. 33506acc9b43SJoe Stringer * Return 33516acc9b43SJoe Stringer * 0 on success, or a negative error in case of failure. 33526fff607eSJohn Fastabend * 3353bdb7b79bSAndrii Nakryiko * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 33546fff607eSJohn Fastabend * Description 335590b1023fSQuentin Monnet * For socket policies, insert *len* bytes into *msg* at offset 33566fff607eSJohn Fastabend * *start*. 33576fff607eSJohn Fastabend * 33586fff607eSJohn Fastabend * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 335990b1023fSQuentin Monnet * *msg* it may want to insert metadata or options into the *msg*. 33606fff607eSJohn Fastabend * This can later be read and used by any of the lower layer BPF 33616fff607eSJohn Fastabend * hooks. 33626fff607eSJohn Fastabend * 33636fff607eSJohn Fastabend * This helper may fail if under memory pressure (a malloc 33646fff607eSJohn Fastabend * fails) in these cases BPF programs will get an appropriate 33656fff607eSJohn Fastabend * error and BPF programs will need to handle them. 33666fff607eSJohn Fastabend * Return 33676fff607eSJohn Fastabend * 0 on success, or a negative error in case of failure. 33687246d8edSJohn Fastabend * 3369bdb7b79bSAndrii Nakryiko * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 33707246d8edSJohn Fastabend * Description 33715f0e5412SAndrii Nakryiko * Will remove *len* bytes from a *msg* starting at byte *start*. 33727246d8edSJohn Fastabend * This may result in **ENOMEM** errors under certain situations if 33737246d8edSJohn Fastabend * an allocation and copy are required due to a full ring buffer. 33747246d8edSJohn Fastabend * However, the helper will try to avoid doing the allocation 33757246d8edSJohn Fastabend * if possible. Other errors can occur if input parameters are 337690b1023fSQuentin Monnet * invalid either due to *start* byte not being valid part of *msg* 33777246d8edSJohn Fastabend * payload and/or *pop* value being to large. 33787246d8edSJohn Fastabend * Return 337990b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 338001d3240aSSean Young * 3381bdb7b79bSAndrii Nakryiko * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 338201d3240aSSean Young * Description 338301d3240aSSean Young * This helper is used in programs implementing IR decoding, to 338401d3240aSSean Young * report a successfully decoded pointer movement. 338501d3240aSSean Young * 338601d3240aSSean Young * The *ctx* should point to the lirc sample as passed into 338701d3240aSSean Young * the program. 338801d3240aSSean Young * 338901d3240aSSean Young * This helper is only available is the kernel was compiled with 339001d3240aSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 339101d3240aSSean Young * "**y**". 339201d3240aSSean Young * Return 339301d3240aSSean Young * 0 339446f8bc92SMartin KaFai Lau * 3395bdb7b79bSAndrii Nakryiko * long bpf_spin_lock(struct bpf_spin_lock *lock) 33960eb09785SQuentin Monnet * Description 33970eb09785SQuentin Monnet * Acquire a spinlock represented by the pointer *lock*, which is 33980eb09785SQuentin Monnet * stored as part of a value of a map. Taking the lock allows to 33990eb09785SQuentin Monnet * safely update the rest of the fields in that value. The 34000eb09785SQuentin Monnet * spinlock can (and must) later be released with a call to 34010eb09785SQuentin Monnet * **bpf_spin_unlock**\ (\ *lock*\ ). 34020eb09785SQuentin Monnet * 34030eb09785SQuentin Monnet * Spinlocks in BPF programs come with a number of restrictions 34040eb09785SQuentin Monnet * and constraints: 34050eb09785SQuentin Monnet * 34060eb09785SQuentin Monnet * * **bpf_spin_lock** objects are only allowed inside maps of 34070eb09785SQuentin Monnet * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 34080eb09785SQuentin Monnet * list could be extended in the future). 34090eb09785SQuentin Monnet * * BTF description of the map is mandatory. 34100eb09785SQuentin Monnet * * The BPF program can take ONE lock at a time, since taking two 34110eb09785SQuentin Monnet * or more could cause dead locks. 34120eb09785SQuentin Monnet * * Only one **struct bpf_spin_lock** is allowed per map element. 34130eb09785SQuentin Monnet * * When the lock is taken, calls (either BPF to BPF or helpers) 34140eb09785SQuentin Monnet * are not allowed. 34150eb09785SQuentin Monnet * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 34160eb09785SQuentin Monnet * allowed inside a spinlock-ed region. 34170eb09785SQuentin Monnet * * The BPF program MUST call **bpf_spin_unlock**\ () to release 34180eb09785SQuentin Monnet * the lock, on all execution paths, before it returns. 34190eb09785SQuentin Monnet * * The BPF program can access **struct bpf_spin_lock** only via 34200eb09785SQuentin Monnet * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 34210eb09785SQuentin Monnet * helpers. Loading or storing data into the **struct 34220eb09785SQuentin Monnet * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 34230eb09785SQuentin Monnet * * To use the **bpf_spin_lock**\ () helper, the BTF description 34240eb09785SQuentin Monnet * of the map value must be a struct and have **struct 34250eb09785SQuentin Monnet * bpf_spin_lock** *anyname*\ **;** field at the top level. 34260eb09785SQuentin Monnet * Nested lock inside another struct is not allowed. 34270eb09785SQuentin Monnet * * The **struct bpf_spin_lock** *lock* field in a map value must 34280eb09785SQuentin Monnet * be aligned on a multiple of 4 bytes in that value. 34290eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 34300eb09785SQuentin Monnet * the **bpf_spin_lock** field to user space. 34310eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 34320eb09785SQuentin Monnet * a BPF program, do not update the **bpf_spin_lock** field. 34330eb09785SQuentin Monnet * * **bpf_spin_lock** cannot be on the stack or inside a 34340eb09785SQuentin Monnet * networking packet (it can only be inside of a map values). 34350eb09785SQuentin Monnet * * **bpf_spin_lock** is available to root only. 34360eb09785SQuentin Monnet * * Tracing programs and socket filter programs cannot use 34370eb09785SQuentin Monnet * **bpf_spin_lock**\ () due to insufficient preemption checks 34380eb09785SQuentin Monnet * (but this may change in the future). 34390eb09785SQuentin Monnet * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 34400eb09785SQuentin Monnet * Return 34410eb09785SQuentin Monnet * 0 34420eb09785SQuentin Monnet * 3443bdb7b79bSAndrii Nakryiko * long bpf_spin_unlock(struct bpf_spin_lock *lock) 34440eb09785SQuentin Monnet * Description 34450eb09785SQuentin Monnet * Release the *lock* previously locked by a call to 34460eb09785SQuentin Monnet * **bpf_spin_lock**\ (\ *lock*\ ). 34470eb09785SQuentin Monnet * Return 34480eb09785SQuentin Monnet * 0 34490eb09785SQuentin Monnet * 345046f8bc92SMartin KaFai Lau * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 345146f8bc92SMartin KaFai Lau * Description 345246f8bc92SMartin KaFai Lau * This helper gets a **struct bpf_sock** pointer such 345362369db2SQuentin Monnet * that all the fields in this **bpf_sock** can be accessed. 345446f8bc92SMartin KaFai Lau * Return 345562369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 345646f8bc92SMartin KaFai Lau * case of failure. 3457655a51e5SMartin KaFai Lau * 3458655a51e5SMartin KaFai Lau * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3459655a51e5SMartin KaFai Lau * Description 3460655a51e5SMartin KaFai Lau * This helper gets a **struct bpf_tcp_sock** pointer from a 3461655a51e5SMartin KaFai Lau * **struct bpf_sock** pointer. 3462655a51e5SMartin KaFai Lau * Return 346362369db2SQuentin Monnet * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3464655a51e5SMartin KaFai Lau * case of failure. 3465f7c917baSbrakmo * 3466bdb7b79bSAndrii Nakryiko * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3467f7c917baSbrakmo * Description 346862369db2SQuentin Monnet * Set ECN (Explicit Congestion Notification) field of IP header 346962369db2SQuentin Monnet * to **CE** (Congestion Encountered) if current value is **ECT** 347062369db2SQuentin Monnet * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 347162369db2SQuentin Monnet * and IPv4. 3472f7c917baSbrakmo * Return 347362369db2SQuentin Monnet * 1 if the **CE** flag is set (either by the current helper call 347462369db2SQuentin Monnet * or because it was already present), 0 if it is not set. 3475dbafd7ddSMartin KaFai Lau * 3476dbafd7ddSMartin KaFai Lau * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3477dbafd7ddSMartin KaFai Lau * Description 347862369db2SQuentin Monnet * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 347962369db2SQuentin Monnet * **bpf_sk_release**\ () is unnecessary and not allowed. 3480dbafd7ddSMartin KaFai Lau * Return 348162369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 3482dbafd7ddSMartin KaFai Lau * case of failure. 3483edbf8c01SLorenz Bauer * 3484edbf8c01SLorenz Bauer * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3485edbf8c01SLorenz Bauer * Description 3486edbf8c01SLorenz Bauer * Look for TCP socket matching *tuple*, optionally in a child 3487edbf8c01SLorenz Bauer * network namespace *netns*. The return value must be checked, 3488edbf8c01SLorenz Bauer * and if non-**NULL**, released via **bpf_sk_release**\ (). 3489edbf8c01SLorenz Bauer * 349080867c5eSQuentin Monnet * This function is identical to **bpf_sk_lookup_tcp**\ (), except 349180867c5eSQuentin Monnet * that it also returns timewait or request sockets. Use 349280867c5eSQuentin Monnet * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 349380867c5eSQuentin Monnet * full structure. 3494edbf8c01SLorenz Bauer * 3495edbf8c01SLorenz Bauer * This helper is available only if the kernel was compiled with 3496edbf8c01SLorenz Bauer * **CONFIG_NET** configuration option. 3497edbf8c01SLorenz Bauer * Return 3498edbf8c01SLorenz Bauer * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3499edbf8c01SLorenz Bauer * For sockets with reuseport option, the **struct bpf_sock** 350080867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 350180867c5eSQuentin Monnet * tuple. 350239904084SLorenz Bauer * 3503c0df236eSMartin KaFai Lau * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 350439904084SLorenz Bauer * Description 350580867c5eSQuentin Monnet * Check whether *iph* and *th* contain a valid SYN cookie ACK for 350680867c5eSQuentin Monnet * the listening socket in *sk*. 350739904084SLorenz Bauer * 350880867c5eSQuentin Monnet * *iph* points to the start of the IPv4 or IPv6 header, while 350980867c5eSQuentin Monnet * *iph_len* contains **sizeof**\ (**struct iphdr**) or 351080867c5eSQuentin Monnet * **sizeof**\ (**struct ip6hdr**). 351139904084SLorenz Bauer * 351280867c5eSQuentin Monnet * *th* points to the start of the TCP header, while *th_len* 351380867c5eSQuentin Monnet * contains **sizeof**\ (**struct tcphdr**). 351439904084SLorenz Bauer * Return 351580867c5eSQuentin Monnet * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 351680867c5eSQuentin Monnet * error otherwise. 3517808649fbSAndrey Ignatov * 3518bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3519808649fbSAndrey Ignatov * Description 3520808649fbSAndrey Ignatov * Get name of sysctl in /proc/sys/ and copy it into provided by 3521808649fbSAndrey Ignatov * program buffer *buf* of size *buf_len*. 3522808649fbSAndrey Ignatov * 3523808649fbSAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 3524808649fbSAndrey Ignatov * 3525808649fbSAndrey Ignatov * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3526808649fbSAndrey Ignatov * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3527808649fbSAndrey Ignatov * only (e.g. "tcp_mem"). 3528808649fbSAndrey Ignatov * Return 3529808649fbSAndrey Ignatov * Number of character copied (not including the trailing NUL). 3530808649fbSAndrey Ignatov * 3531808649fbSAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3532808649fbSAndrey Ignatov * truncated name in this case). 35331d11b301SAndrey Ignatov * 3534bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 35351d11b301SAndrey Ignatov * Description 35361d11b301SAndrey Ignatov * Get current value of sysctl as it is presented in /proc/sys 35371d11b301SAndrey Ignatov * (incl. newline, etc), and copy it as a string into provided 35381d11b301SAndrey Ignatov * by program buffer *buf* of size *buf_len*. 35391d11b301SAndrey Ignatov * 35401d11b301SAndrey Ignatov * The whole value is copied, no matter what file position user 35411d11b301SAndrey Ignatov * space issued e.g. sys_read at. 35421d11b301SAndrey Ignatov * 35431d11b301SAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 35441d11b301SAndrey Ignatov * Return 35451d11b301SAndrey Ignatov * Number of character copied (not including the trailing NUL). 35461d11b301SAndrey Ignatov * 35471d11b301SAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 35481d11b301SAndrey Ignatov * truncated name in this case). 35491d11b301SAndrey Ignatov * 35501d11b301SAndrey Ignatov * **-EINVAL** if current value was unavailable, e.g. because 35511d11b301SAndrey Ignatov * sysctl is uninitialized and read returns -EIO for it. 35524e63acdfSAndrey Ignatov * 3553bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 35544e63acdfSAndrey Ignatov * Description 35554e63acdfSAndrey Ignatov * Get new value being written by user space to sysctl (before 35564e63acdfSAndrey Ignatov * the actual write happens) and copy it as a string into 35574e63acdfSAndrey Ignatov * provided by program buffer *buf* of size *buf_len*. 35584e63acdfSAndrey Ignatov * 35594e63acdfSAndrey Ignatov * User space may write new value at file position > 0. 35604e63acdfSAndrey Ignatov * 35614e63acdfSAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 35624e63acdfSAndrey Ignatov * Return 35634e63acdfSAndrey Ignatov * Number of character copied (not including the trailing NUL). 35644e63acdfSAndrey Ignatov * 35654e63acdfSAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 35664e63acdfSAndrey Ignatov * truncated name in this case). 35674e63acdfSAndrey Ignatov * 35684e63acdfSAndrey Ignatov * **-EINVAL** if sysctl is being read. 35694e63acdfSAndrey Ignatov * 3570bdb7b79bSAndrii Nakryiko * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 35714e63acdfSAndrey Ignatov * Description 35724e63acdfSAndrey Ignatov * Override new value being written by user space to sysctl with 35734e63acdfSAndrey Ignatov * value provided by program in buffer *buf* of size *buf_len*. 35744e63acdfSAndrey Ignatov * 35754e63acdfSAndrey Ignatov * *buf* should contain a string in same form as provided by user 35764e63acdfSAndrey Ignatov * space on sysctl write. 35774e63acdfSAndrey Ignatov * 35784e63acdfSAndrey Ignatov * User space may write new value at file position > 0. To override 35794e63acdfSAndrey Ignatov * the whole sysctl value file position should be set to zero. 35804e63acdfSAndrey Ignatov * Return 35814e63acdfSAndrey Ignatov * 0 on success. 35824e63acdfSAndrey Ignatov * 35834e63acdfSAndrey Ignatov * **-E2BIG** if the *buf_len* is too big. 35844e63acdfSAndrey Ignatov * 35854e63acdfSAndrey Ignatov * **-EINVAL** if sysctl is being read. 3586d7a4cb9bSAndrey Ignatov * 3587bdb7b79bSAndrii Nakryiko * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3588d7a4cb9bSAndrey Ignatov * Description 3589d7a4cb9bSAndrey Ignatov * Convert the initial part of the string from buffer *buf* of 3590d7a4cb9bSAndrey Ignatov * size *buf_len* to a long integer according to the given base 3591d7a4cb9bSAndrey Ignatov * and save the result in *res*. 3592d7a4cb9bSAndrey Ignatov * 3593d7a4cb9bSAndrey Ignatov * The string may begin with an arbitrary amount of white space 359480867c5eSQuentin Monnet * (as determined by **isspace**\ (3)) followed by a single 359580867c5eSQuentin Monnet * optional '**-**' sign. 3596d7a4cb9bSAndrey Ignatov * 3597d7a4cb9bSAndrey Ignatov * Five least significant bits of *flags* encode base, other bits 3598d7a4cb9bSAndrey Ignatov * are currently unused. 3599d7a4cb9bSAndrey Ignatov * 3600d7a4cb9bSAndrey Ignatov * Base must be either 8, 10, 16 or 0 to detect it automatically 360180867c5eSQuentin Monnet * similar to user space **strtol**\ (3). 3602d7a4cb9bSAndrey Ignatov * Return 3603d7a4cb9bSAndrey Ignatov * Number of characters consumed on success. Must be positive but 360480867c5eSQuentin Monnet * no more than *buf_len*. 3605d7a4cb9bSAndrey Ignatov * 3606d7a4cb9bSAndrey Ignatov * **-EINVAL** if no valid digits were found or unsupported base 3607d7a4cb9bSAndrey Ignatov * was provided. 3608d7a4cb9bSAndrey Ignatov * 3609d7a4cb9bSAndrey Ignatov * **-ERANGE** if resulting value was out of range. 3610d7a4cb9bSAndrey Ignatov * 3611bdb7b79bSAndrii Nakryiko * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3612d7a4cb9bSAndrey Ignatov * Description 3613d7a4cb9bSAndrey Ignatov * Convert the initial part of the string from buffer *buf* of 3614d7a4cb9bSAndrey Ignatov * size *buf_len* to an unsigned long integer according to the 3615d7a4cb9bSAndrey Ignatov * given base 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)). 3619d7a4cb9bSAndrey Ignatov * 3620d7a4cb9bSAndrey Ignatov * Five least significant bits of *flags* encode base, other bits 3621d7a4cb9bSAndrey Ignatov * are currently unused. 3622d7a4cb9bSAndrey Ignatov * 3623d7a4cb9bSAndrey Ignatov * Base must be either 8, 10, 16 or 0 to detect it automatically 362480867c5eSQuentin Monnet * similar to user space **strtoul**\ (3). 3625d7a4cb9bSAndrey Ignatov * Return 3626d7a4cb9bSAndrey Ignatov * Number of characters consumed on success. Must be positive but 362780867c5eSQuentin Monnet * no more than *buf_len*. 3628d7a4cb9bSAndrey Ignatov * 3629d7a4cb9bSAndrey Ignatov * **-EINVAL** if no valid digits were found or unsupported base 3630d7a4cb9bSAndrey Ignatov * was provided. 3631d7a4cb9bSAndrey Ignatov * 3632d7a4cb9bSAndrey Ignatov * **-ERANGE** if resulting value was out of range. 36336ac99e8fSMartin KaFai Lau * 363430897832SKP Singh * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 36356ac99e8fSMartin KaFai Lau * Description 363680867c5eSQuentin Monnet * Get a bpf-local-storage from a *sk*. 36376ac99e8fSMartin KaFai Lau * 36386ac99e8fSMartin KaFai Lau * Logically, it could be thought of getting the value from 36396ac99e8fSMartin KaFai Lau * a *map* with *sk* as the **key**. From this 36406ac99e8fSMartin KaFai Lau * perspective, the usage is not much different from 364180867c5eSQuentin Monnet * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 364280867c5eSQuentin Monnet * helper enforces the key must be a full socket and the map must 364380867c5eSQuentin Monnet * be a **BPF_MAP_TYPE_SK_STORAGE** also. 36446ac99e8fSMartin KaFai Lau * 36456ac99e8fSMartin KaFai Lau * Underneath, the value is stored locally at *sk* instead of 364680867c5eSQuentin Monnet * the *map*. The *map* is used as the bpf-local-storage 364780867c5eSQuentin Monnet * "type". The bpf-local-storage "type" (i.e. the *map*) is 364880867c5eSQuentin Monnet * searched against all bpf-local-storages residing at *sk*. 36496ac99e8fSMartin KaFai Lau * 365030897832SKP Singh * *sk* is a kernel **struct sock** pointer for LSM program. 365130897832SKP Singh * *sk* is a **struct bpf_sock** pointer for other program types. 365230897832SKP Singh * 365380867c5eSQuentin Monnet * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 36546ac99e8fSMartin KaFai Lau * used such that a new bpf-local-storage will be 36556ac99e8fSMartin KaFai Lau * created if one does not exist. *value* can be used 365680867c5eSQuentin Monnet * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 36576ac99e8fSMartin KaFai Lau * the initial value of a bpf-local-storage. If *value* is 365880867c5eSQuentin Monnet * **NULL**, the new bpf-local-storage will be zero initialized. 36596ac99e8fSMartin KaFai Lau * Return 36606ac99e8fSMartin KaFai Lau * A bpf-local-storage pointer is returned on success. 36616ac99e8fSMartin KaFai Lau * 36626ac99e8fSMartin KaFai Lau * **NULL** if not found or there was an error in adding 36636ac99e8fSMartin KaFai Lau * a new bpf-local-storage. 36646ac99e8fSMartin KaFai Lau * 366530897832SKP Singh * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 36666ac99e8fSMartin KaFai Lau * Description 366780867c5eSQuentin Monnet * Delete a bpf-local-storage from a *sk*. 36686ac99e8fSMartin KaFai Lau * Return 36696ac99e8fSMartin KaFai Lau * 0 on success. 36706ac99e8fSMartin KaFai Lau * 36716ac99e8fSMartin KaFai Lau * **-ENOENT** if the bpf-local-storage cannot be found. 3672592a3498SMartin KaFai Lau * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 36738b401f9eSYonghong Song * 3674bdb7b79bSAndrii Nakryiko * long bpf_send_signal(u32 sig) 36758b401f9eSYonghong Song * Description 36768482941fSYonghong Song * Send signal *sig* to the process of the current task. 36778482941fSYonghong Song * The signal may be delivered to any of this process's threads. 36788b401f9eSYonghong Song * Return 36798b401f9eSYonghong Song * 0 on success or successfully queued. 36808b401f9eSYonghong Song * 36818b401f9eSYonghong Song * **-EBUSY** if work queue under nmi is full. 36828b401f9eSYonghong Song * 36838b401f9eSYonghong Song * **-EINVAL** if *sig* is invalid. 36848b401f9eSYonghong Song * 36858b401f9eSYonghong Song * **-EPERM** if no permission to send the *sig*. 36868b401f9eSYonghong Song * 36878b401f9eSYonghong Song * **-EAGAIN** if bpf program can try again. 368870d66244SPetar Penkov * 3689c0df236eSMartin KaFai Lau * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 369070d66244SPetar Penkov * Description 369170d66244SPetar Penkov * Try to issue a SYN cookie for the packet with corresponding 369270d66244SPetar Penkov * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 369370d66244SPetar Penkov * 369470d66244SPetar Penkov * *iph* points to the start of the IPv4 or IPv6 header, while 369570d66244SPetar Penkov * *iph_len* contains **sizeof**\ (**struct iphdr**) or 369670d66244SPetar Penkov * **sizeof**\ (**struct ip6hdr**). 369770d66244SPetar Penkov * 369870d66244SPetar Penkov * *th* points to the start of the TCP header, while *th_len* 369970d66244SPetar Penkov * contains the length of the TCP header. 370070d66244SPetar Penkov * Return 370170d66244SPetar Penkov * On success, lower 32 bits hold the generated SYN cookie in 370270d66244SPetar Penkov * followed by 16 bits which hold the MSS value for that cookie, 370370d66244SPetar Penkov * and the top 16 bits are unused. 370470d66244SPetar Penkov * 370570d66244SPetar Penkov * On failure, the returned value is one of the following: 370670d66244SPetar Penkov * 370770d66244SPetar Penkov * **-EINVAL** SYN cookie cannot be issued due to error 370870d66244SPetar Penkov * 370970d66244SPetar Penkov * **-ENOENT** SYN cookie should not be issued (no SYN flood) 371070d66244SPetar Penkov * 371170d66244SPetar Penkov * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 371270d66244SPetar Penkov * 371370d66244SPetar Penkov * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 3714a7658e1aSAlexei Starovoitov * 3715bdb7b79bSAndrii Nakryiko * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3716a7658e1aSAlexei Starovoitov * Description 3717a7658e1aSAlexei Starovoitov * Write raw *data* blob into a special BPF perf event held by 3718a7658e1aSAlexei Starovoitov * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3719a7658e1aSAlexei Starovoitov * event must have the following attributes: **PERF_SAMPLE_RAW** 3720a7658e1aSAlexei Starovoitov * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3721a7658e1aSAlexei Starovoitov * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3722a7658e1aSAlexei Starovoitov * 3723a7658e1aSAlexei Starovoitov * The *flags* are used to indicate the index in *map* for which 3724a7658e1aSAlexei Starovoitov * the value must be put, masked with **BPF_F_INDEX_MASK**. 3725a7658e1aSAlexei Starovoitov * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3726a7658e1aSAlexei Starovoitov * to indicate that the index of the current CPU core should be 3727a7658e1aSAlexei Starovoitov * used. 3728a7658e1aSAlexei Starovoitov * 3729a7658e1aSAlexei Starovoitov * The value to write, of *size*, is passed through eBPF stack and 3730a7658e1aSAlexei Starovoitov * pointed by *data*. 3731a7658e1aSAlexei Starovoitov * 3732a7658e1aSAlexei Starovoitov * *ctx* is a pointer to in-kernel struct sk_buff. 3733a7658e1aSAlexei Starovoitov * 3734a7658e1aSAlexei Starovoitov * This helper is similar to **bpf_perf_event_output**\ () but 3735a7658e1aSAlexei Starovoitov * restricted to raw_tracepoint bpf programs. 3736a7658e1aSAlexei Starovoitov * Return 3737a7658e1aSAlexei Starovoitov * 0 on success, or a negative error in case of failure. 37386ae08ae3SDaniel Borkmann * 3739bdb7b79bSAndrii Nakryiko * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 37406ae08ae3SDaniel Borkmann * Description 37416ae08ae3SDaniel Borkmann * Safely attempt to read *size* bytes from user space address 37426ae08ae3SDaniel Borkmann * *unsafe_ptr* and store the data in *dst*. 37436ae08ae3SDaniel Borkmann * Return 37446ae08ae3SDaniel Borkmann * 0 on success, or a negative error in case of failure. 37456ae08ae3SDaniel Borkmann * 3746bdb7b79bSAndrii Nakryiko * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 37476ae08ae3SDaniel Borkmann * Description 37486ae08ae3SDaniel Borkmann * Safely attempt to read *size* bytes from kernel space address 37496ae08ae3SDaniel Borkmann * *unsafe_ptr* and store the data in *dst*. 37506ae08ae3SDaniel Borkmann * Return 37516ae08ae3SDaniel Borkmann * 0 on success, or a negative error in case of failure. 37526ae08ae3SDaniel Borkmann * 3753bdb7b79bSAndrii Nakryiko * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 37546ae08ae3SDaniel Borkmann * Description 37556ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe user address 37566ae08ae3SDaniel Borkmann * *unsafe_ptr* to *dst*. The *size* should include the 37576ae08ae3SDaniel Borkmann * terminating NUL byte. In case the string length is smaller than 37586ae08ae3SDaniel Borkmann * *size*, the target is not padded with further NUL bytes. If the 37596ae08ae3SDaniel Borkmann * string length is larger than *size*, just *size*-1 bytes are 37606ae08ae3SDaniel Borkmann * copied and the last byte is set to NUL. 37616ae08ae3SDaniel Borkmann * 3762c6458e72SBrendan Jackman * On success, returns the number of bytes that were written, 3763c6458e72SBrendan Jackman * including the terminal NUL. This makes this helper useful in 3764c6458e72SBrendan Jackman * tracing programs for reading strings, and more importantly to 3765c6458e72SBrendan Jackman * get its length at runtime. See the following snippet: 37666ae08ae3SDaniel Borkmann * 37676ae08ae3SDaniel Borkmann * :: 37686ae08ae3SDaniel Borkmann * 37696ae08ae3SDaniel Borkmann * SEC("kprobe/sys_open") 37706ae08ae3SDaniel Borkmann * void bpf_sys_open(struct pt_regs *ctx) 37716ae08ae3SDaniel Borkmann * { 37726ae08ae3SDaniel Borkmann * char buf[PATHLEN]; // PATHLEN is defined to 256 37736ae08ae3SDaniel Borkmann * int res = bpf_probe_read_user_str(buf, sizeof(buf), 37746ae08ae3SDaniel Borkmann * ctx->di); 37756ae08ae3SDaniel Borkmann * 37766ae08ae3SDaniel Borkmann * // Consume buf, for example push it to 37776ae08ae3SDaniel Borkmann * // userspace via bpf_perf_event_output(); we 37786ae08ae3SDaniel Borkmann * // can use res (the string length) as event 37796ae08ae3SDaniel Borkmann * // size, after checking its boundaries. 37806ae08ae3SDaniel Borkmann * } 37816ae08ae3SDaniel Borkmann * 3782ab8d7809SQuentin Monnet * In comparison, using **bpf_probe_read_user**\ () helper here 37836ae08ae3SDaniel Borkmann * instead to read the string would require to estimate the length 37846ae08ae3SDaniel Borkmann * at compile time, and would often result in copying more memory 37856ae08ae3SDaniel Borkmann * than necessary. 37866ae08ae3SDaniel Borkmann * 37876ae08ae3SDaniel Borkmann * Another useful use case is when parsing individual process 37886ae08ae3SDaniel Borkmann * arguments or individual environment variables navigating 37896ae08ae3SDaniel Borkmann * *current*\ **->mm->arg_start** and *current*\ 37906ae08ae3SDaniel Borkmann * **->mm->env_start**: using this helper and the return value, 37916ae08ae3SDaniel Borkmann * one can quickly iterate at the right offset of the memory area. 37926ae08ae3SDaniel Borkmann * Return 3793c6458e72SBrendan Jackman * On success, the strictly positive length of the output string, 37946ae08ae3SDaniel Borkmann * including the trailing NUL character. On error, a negative 37956ae08ae3SDaniel Borkmann * value. 37966ae08ae3SDaniel Borkmann * 3797bdb7b79bSAndrii Nakryiko * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 37986ae08ae3SDaniel Borkmann * Description 37996ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 3800ab8d7809SQuentin Monnet * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 38016ae08ae3SDaniel Borkmann * Return 38026ae08ae3SDaniel Borkmann * On success, the strictly positive length of the string, including 38036ae08ae3SDaniel Borkmann * the trailing NUL character. On error, a negative value. 3804206057feSMartin KaFai Lau * 3805bdb7b79bSAndrii Nakryiko * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 3806206057feSMartin KaFai Lau * Description 3807ab8d7809SQuentin Monnet * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 3808206057feSMartin KaFai Lau * *rcv_nxt* is the ack_seq to be sent out. 3809206057feSMartin KaFai Lau * Return 3810206057feSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 3811206057feSMartin KaFai Lau * 3812bdb7b79bSAndrii Nakryiko * long bpf_send_signal_thread(u32 sig) 38138482941fSYonghong Song * Description 38148482941fSYonghong Song * Send signal *sig* to the thread corresponding to the current task. 38158482941fSYonghong Song * Return 38168482941fSYonghong Song * 0 on success or successfully queued. 38178482941fSYonghong Song * 38188482941fSYonghong Song * **-EBUSY** if work queue under nmi is full. 38198482941fSYonghong Song * 38208482941fSYonghong Song * **-EINVAL** if *sig* is invalid. 38218482941fSYonghong Song * 38228482941fSYonghong Song * **-EPERM** if no permission to send the *sig*. 38238482941fSYonghong Song * 38248482941fSYonghong Song * **-EAGAIN** if bpf program can try again. 38255576b991SMartin KaFai Lau * 38265576b991SMartin KaFai Lau * u64 bpf_jiffies64(void) 38275576b991SMartin KaFai Lau * Description 38285576b991SMartin KaFai Lau * Obtain the 64bit jiffies 38295576b991SMartin KaFai Lau * Return 38305576b991SMartin KaFai Lau * The 64 bit jiffies 3831fff7b643SDaniel Xu * 3832bdb7b79bSAndrii Nakryiko * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3833fff7b643SDaniel Xu * Description 3834fff7b643SDaniel Xu * For an eBPF program attached to a perf event, retrieve the 3835ab8d7809SQuentin Monnet * branch records (**struct perf_branch_entry**) associated to *ctx* 3836fff7b643SDaniel Xu * and store it in the buffer pointed by *buf* up to size 3837fff7b643SDaniel Xu * *size* bytes. 3838fff7b643SDaniel Xu * Return 3839fff7b643SDaniel Xu * On success, number of bytes written to *buf*. On error, a 3840fff7b643SDaniel Xu * negative value. 3841fff7b643SDaniel Xu * 3842fff7b643SDaniel Xu * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3843fff7b643SDaniel Xu * instead return the number of bytes required to store all the 3844fff7b643SDaniel Xu * branch entries. If this flag is set, *buf* may be NULL. 3845fff7b643SDaniel Xu * 3846fff7b643SDaniel Xu * **-EINVAL** if arguments invalid or **size** not a multiple 3847ab8d7809SQuentin Monnet * of **sizeof**\ (**struct perf_branch_entry**\ ). 3848fff7b643SDaniel Xu * 3849fff7b643SDaniel Xu * **-ENOENT** if architecture does not support branch records. 3850b4490c5cSCarlos Neira * 3851bdb7b79bSAndrii Nakryiko * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3852b4490c5cSCarlos Neira * Description 3853b4490c5cSCarlos Neira * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3854b4490c5cSCarlos Neira * *namespace* will be returned in *nsdata*. 3855ab8d7809SQuentin Monnet * Return 3856ab8d7809SQuentin Monnet * 0 on success, or one of the following in case of failure: 3857b4490c5cSCarlos Neira * 3858b4490c5cSCarlos Neira * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3859b4490c5cSCarlos Neira * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3860b4490c5cSCarlos Neira * 3861b4490c5cSCarlos Neira * **-ENOENT** if pidns does not exists for the current task. 3862b4490c5cSCarlos Neira * 3863bdb7b79bSAndrii Nakryiko * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3864d831ee84SEelco Chaudron * Description 3865d831ee84SEelco Chaudron * Write raw *data* blob into a special BPF perf event held by 3866d831ee84SEelco Chaudron * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3867d831ee84SEelco Chaudron * event must have the following attributes: **PERF_SAMPLE_RAW** 3868d831ee84SEelco Chaudron * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3869d831ee84SEelco Chaudron * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3870d831ee84SEelco Chaudron * 3871d831ee84SEelco Chaudron * The *flags* are used to indicate the index in *map* for which 3872d831ee84SEelco Chaudron * the value must be put, masked with **BPF_F_INDEX_MASK**. 3873d831ee84SEelco Chaudron * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3874d831ee84SEelco Chaudron * to indicate that the index of the current CPU core should be 3875d831ee84SEelco Chaudron * used. 3876d831ee84SEelco Chaudron * 3877d831ee84SEelco Chaudron * The value to write, of *size*, is passed through eBPF stack and 3878d831ee84SEelco Chaudron * pointed by *data*. 3879d831ee84SEelco Chaudron * 3880d831ee84SEelco Chaudron * *ctx* is a pointer to in-kernel struct xdp_buff. 3881d831ee84SEelco Chaudron * 3882d831ee84SEelco Chaudron * This helper is similar to **bpf_perf_eventoutput**\ () but 3883d831ee84SEelco Chaudron * restricted to raw_tracepoint bpf programs. 3884d831ee84SEelco Chaudron * Return 3885d831ee84SEelco Chaudron * 0 on success, or a negative error in case of failure. 3886f318903cSDaniel Borkmann * 3887f318903cSDaniel Borkmann * u64 bpf_get_netns_cookie(void *ctx) 3888f318903cSDaniel Borkmann * Description 3889f318903cSDaniel Borkmann * Retrieve the cookie (generated by the kernel) of the network 3890f318903cSDaniel Borkmann * namespace the input *ctx* is associated with. The network 3891f318903cSDaniel Borkmann * namespace cookie remains stable for its lifetime and provides 3892f318903cSDaniel Borkmann * a global identifier that can be assumed unique. If *ctx* is 3893f318903cSDaniel Borkmann * NULL, then the helper returns the cookie for the initial 3894f318903cSDaniel Borkmann * network namespace. The cookie itself is very similar to that 3895ab8d7809SQuentin Monnet * of **bpf_get_socket_cookie**\ () helper, but for network 3896ab8d7809SQuentin Monnet * namespaces instead of sockets. 3897f318903cSDaniel Borkmann * Return 3898f318903cSDaniel Borkmann * A 8-byte long opaque number. 38990f09abd1SDaniel Borkmann * 39000f09abd1SDaniel Borkmann * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 39010f09abd1SDaniel Borkmann * Description 39020f09abd1SDaniel Borkmann * Return id of cgroup v2 that is ancestor of the cgroup associated 39030f09abd1SDaniel Borkmann * with the current task at the *ancestor_level*. The root cgroup 39040f09abd1SDaniel Borkmann * is at *ancestor_level* zero and each step down the hierarchy 39050f09abd1SDaniel Borkmann * increments the level. If *ancestor_level* == level of cgroup 39060f09abd1SDaniel Borkmann * associated with the current task, then return value will be the 39070f09abd1SDaniel Borkmann * same as that of **bpf_get_current_cgroup_id**\ (). 39080f09abd1SDaniel Borkmann * 39090f09abd1SDaniel Borkmann * The helper is useful to implement policies based on cgroups 39100f09abd1SDaniel Borkmann * that are upper in hierarchy than immediate cgroup associated 39110f09abd1SDaniel Borkmann * with the current task. 39120f09abd1SDaniel Borkmann * 39130f09abd1SDaniel Borkmann * The format of returned id and helper limitations are same as in 39140f09abd1SDaniel Borkmann * **bpf_get_current_cgroup_id**\ (). 39150f09abd1SDaniel Borkmann * Return 39160f09abd1SDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 3917cf7fbe66SJoe Stringer * 391827e5203bSMartin KaFai Lau * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 3919cf7fbe66SJoe Stringer * Description 3920e9ddbb77SJakub Sitnicki * Helper is overloaded depending on BPF program type. This 3921e9ddbb77SJakub Sitnicki * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 3922e9ddbb77SJakub Sitnicki * **BPF_PROG_TYPE_SCHED_ACT** programs. 3923e9ddbb77SJakub Sitnicki * 3924cf7fbe66SJoe Stringer * Assign the *sk* to the *skb*. When combined with appropriate 3925cf7fbe66SJoe Stringer * routing configuration to receive the packet towards the socket, 3926cf7fbe66SJoe Stringer * will cause *skb* to be delivered to the specified socket. 3927cf7fbe66SJoe Stringer * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3928cf7fbe66SJoe Stringer * **bpf_clone_redirect**\ () or other methods outside of BPF may 3929cf7fbe66SJoe Stringer * interfere with successful delivery to the socket. 3930cf7fbe66SJoe Stringer * 3931cf7fbe66SJoe Stringer * This operation is only valid from TC ingress path. 3932cf7fbe66SJoe Stringer * 3933cf7fbe66SJoe Stringer * The *flags* argument must be zero. 3934cf7fbe66SJoe Stringer * Return 3935ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 3936cf7fbe66SJoe Stringer * 3937ab8d7809SQuentin Monnet * **-EINVAL** if specified *flags* are not supported. 3938ab8d7809SQuentin Monnet * 3939ab8d7809SQuentin Monnet * **-ENOENT** if the socket is unavailable for assignment. 3940ab8d7809SQuentin Monnet * 3941ab8d7809SQuentin Monnet * **-ENETUNREACH** if the socket is unreachable (wrong netns). 3942ab8d7809SQuentin Monnet * 3943ab8d7809SQuentin Monnet * **-EOPNOTSUPP** if the operation is not supported, for example 3944ab8d7809SQuentin Monnet * a call from outside of TC ingress. 3945ab8d7809SQuentin Monnet * 3946ab8d7809SQuentin Monnet * **-ESOCKTNOSUPPORT** if the socket type is not supported 3947ab8d7809SQuentin Monnet * (reuseport). 394871d19214SMaciej Żenczykowski * 3949e9ddbb77SJakub Sitnicki * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 3950e9ddbb77SJakub Sitnicki * Description 3951e9ddbb77SJakub Sitnicki * Helper is overloaded depending on BPF program type. This 3952e9ddbb77SJakub Sitnicki * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 3953e9ddbb77SJakub Sitnicki * 3954e9ddbb77SJakub Sitnicki * Select the *sk* as a result of a socket lookup. 3955e9ddbb77SJakub Sitnicki * 3956e9ddbb77SJakub Sitnicki * For the operation to succeed passed socket must be compatible 3957e9ddbb77SJakub Sitnicki * with the packet description provided by the *ctx* object. 3958e9ddbb77SJakub Sitnicki * 3959e9ddbb77SJakub Sitnicki * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 3960e9ddbb77SJakub Sitnicki * be an exact match. While IP family (**AF_INET** or 3961e9ddbb77SJakub Sitnicki * **AF_INET6**) must be compatible, that is IPv6 sockets 3962e9ddbb77SJakub Sitnicki * that are not v6-only can be selected for IPv4 packets. 3963e9ddbb77SJakub Sitnicki * 3964e9ddbb77SJakub Sitnicki * Only TCP listeners and UDP unconnected sockets can be 3965e9ddbb77SJakub Sitnicki * selected. *sk* can also be NULL to reset any previous 3966e9ddbb77SJakub Sitnicki * selection. 3967e9ddbb77SJakub Sitnicki * 3968e9ddbb77SJakub Sitnicki * *flags* argument can combination of following values: 3969e9ddbb77SJakub Sitnicki * 3970e9ddbb77SJakub Sitnicki * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 3971e9ddbb77SJakub Sitnicki * socket selection, potentially done by a BPF program 3972e9ddbb77SJakub Sitnicki * that ran before us. 3973e9ddbb77SJakub Sitnicki * 3974e9ddbb77SJakub Sitnicki * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 3975e9ddbb77SJakub Sitnicki * load-balancing within reuseport group for the socket 3976e9ddbb77SJakub Sitnicki * being selected. 3977e9ddbb77SJakub Sitnicki * 3978e9ddbb77SJakub Sitnicki * On success *ctx->sk* will point to the selected socket. 3979e9ddbb77SJakub Sitnicki * 3980e9ddbb77SJakub Sitnicki * Return 3981e9ddbb77SJakub Sitnicki * 0 on success, or a negative errno in case of failure. 3982e9ddbb77SJakub Sitnicki * 3983e9ddbb77SJakub Sitnicki * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 3984e9ddbb77SJakub Sitnicki * not compatible with packet family (*ctx->family*). 3985e9ddbb77SJakub Sitnicki * 3986e9ddbb77SJakub Sitnicki * * **-EEXIST** if socket has been already selected, 3987e9ddbb77SJakub Sitnicki * potentially by another program, and 3988e9ddbb77SJakub Sitnicki * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 3989e9ddbb77SJakub Sitnicki * 3990e9ddbb77SJakub Sitnicki * * **-EINVAL** if unsupported flags were specified. 3991e9ddbb77SJakub Sitnicki * 3992e9ddbb77SJakub Sitnicki * * **-EPROTOTYPE** if socket L4 protocol 3993e9ddbb77SJakub Sitnicki * (*sk->protocol*) doesn't match packet protocol 3994e9ddbb77SJakub Sitnicki * (*ctx->protocol*). 3995e9ddbb77SJakub Sitnicki * 3996e9ddbb77SJakub Sitnicki * * **-ESOCKTNOSUPPORT** if socket is not in allowed 3997e9ddbb77SJakub Sitnicki * state (TCP listening or UDP unconnected). 3998e9ddbb77SJakub Sitnicki * 399971d19214SMaciej Żenczykowski * u64 bpf_ktime_get_boot_ns(void) 400071d19214SMaciej Żenczykowski * Description 400171d19214SMaciej Żenczykowski * Return the time elapsed since system boot, in nanoseconds. 400271d19214SMaciej Żenczykowski * Does include the time the system was suspended. 4003ab8d7809SQuentin Monnet * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 400471d19214SMaciej Żenczykowski * Return 400571d19214SMaciej Żenczykowski * Current *ktime*. 4006492e639fSYonghong Song * 4007bdb7b79bSAndrii Nakryiko * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4008492e639fSYonghong Song * Description 4009ab8d7809SQuentin Monnet * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4010ab8d7809SQuentin Monnet * out the format string. 4011492e639fSYonghong Song * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4012492e639fSYonghong Song * the format string itself. The *data* and *data_len* are format string 4013ab8d7809SQuentin Monnet * arguments. The *data* are a **u64** array and corresponding format string 4014492e639fSYonghong Song * values are stored in the array. For strings and pointers where pointees 4015492e639fSYonghong Song * are accessed, only the pointer values are stored in the *data* array. 4016ab8d7809SQuentin Monnet * The *data_len* is the size of *data* in bytes. 4017492e639fSYonghong Song * 4018492e639fSYonghong Song * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4019492e639fSYonghong Song * Reading kernel memory may fail due to either invalid address or 4020492e639fSYonghong Song * valid address but requiring a major memory fault. If reading kernel memory 4021492e639fSYonghong Song * fails, the string for **%s** will be an empty string, and the ip 4022492e639fSYonghong Song * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4023ab8d7809SQuentin Monnet * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4024492e639fSYonghong Song * Return 4025ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 4026492e639fSYonghong Song * 4027ab8d7809SQuentin Monnet * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4028492e639fSYonghong Song * by returning 1 from bpf program. 4029ab8d7809SQuentin Monnet * 4030ab8d7809SQuentin Monnet * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4031ab8d7809SQuentin Monnet * 4032ab8d7809SQuentin Monnet * **-E2BIG** if *fmt* contains too many format specifiers. 4033ab8d7809SQuentin Monnet * 4034ab8d7809SQuentin Monnet * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4035492e639fSYonghong Song * 4036bdb7b79bSAndrii Nakryiko * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4037492e639fSYonghong Song * Description 4038ab8d7809SQuentin Monnet * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4039492e639fSYonghong Song * The *m* represents the seq_file. The *data* and *len* represent the 4040492e639fSYonghong Song * data to write in bytes. 4041492e639fSYonghong Song * Return 4042ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 4043492e639fSYonghong Song * 4044ab8d7809SQuentin Monnet * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4045f307fa2cSAndrey Ignatov * 4046a5fa25adSMartin KaFai Lau * u64 bpf_sk_cgroup_id(void *sk) 4047f307fa2cSAndrey Ignatov * Description 4048f307fa2cSAndrey Ignatov * Return the cgroup v2 id of the socket *sk*. 4049f307fa2cSAndrey Ignatov * 4050a5fa25adSMartin KaFai Lau * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4051f307fa2cSAndrey Ignatov * returned from **bpf_sk_lookup_xxx**\ (), 4052f307fa2cSAndrey Ignatov * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4053f307fa2cSAndrey Ignatov * same as in **bpf_skb_cgroup_id**\ (). 4054f307fa2cSAndrey Ignatov * 4055f307fa2cSAndrey Ignatov * This helper is available only if the kernel was compiled with 4056f307fa2cSAndrey Ignatov * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4057f307fa2cSAndrey Ignatov * Return 4058f307fa2cSAndrey Ignatov * The id is returned or 0 in case the id could not be retrieved. 4059f307fa2cSAndrey Ignatov * 4060a5fa25adSMartin KaFai Lau * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4061f307fa2cSAndrey Ignatov * Description 4062f307fa2cSAndrey Ignatov * Return id of cgroup v2 that is ancestor of cgroup associated 4063f307fa2cSAndrey Ignatov * with the *sk* at the *ancestor_level*. The root cgroup is at 4064f307fa2cSAndrey Ignatov * *ancestor_level* zero and each step down the hierarchy 4065f307fa2cSAndrey Ignatov * increments the level. If *ancestor_level* == level of cgroup 4066f307fa2cSAndrey Ignatov * associated with *sk*, then return value will be same as that 4067f307fa2cSAndrey Ignatov * of **bpf_sk_cgroup_id**\ (). 4068f307fa2cSAndrey Ignatov * 4069f307fa2cSAndrey Ignatov * The helper is useful to implement policies based on cgroups 4070f307fa2cSAndrey Ignatov * that are upper in hierarchy than immediate cgroup associated 4071f307fa2cSAndrey Ignatov * with *sk*. 4072f307fa2cSAndrey Ignatov * 4073f307fa2cSAndrey Ignatov * The format of returned id and helper limitations are same as in 4074f307fa2cSAndrey Ignatov * **bpf_sk_cgroup_id**\ (). 4075f307fa2cSAndrey Ignatov * Return 4076f307fa2cSAndrey Ignatov * The id is returned or 0 in case the id could not be retrieved. 4077457f4436SAndrii Nakryiko * 4078e1613b57SAndrii Nakryiko * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4079457f4436SAndrii Nakryiko * Description 4080457f4436SAndrii Nakryiko * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4081bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4082bcc7f554SQuentin Monnet * of new data availability is sent. 4083bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4084bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 40855c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 40865c507329SPedro Tammela * of new data availability is sent. 40875c507329SPedro Tammela * 40885c507329SPedro Tammela * An adaptive notification is a notification sent whenever the user-space 40895c507329SPedro Tammela * process has caught up and consumed all available payloads. In case the user-space 40905c507329SPedro Tammela * process is still processing a previous payload, then no notification is needed 40915c507329SPedro Tammela * as it will process the newly added payload automatically. 4092457f4436SAndrii Nakryiko * Return 4093bcc7f554SQuentin Monnet * 0 on success, or a negative error in case of failure. 4094457f4436SAndrii Nakryiko * 4095457f4436SAndrii Nakryiko * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4096457f4436SAndrii Nakryiko * Description 4097457f4436SAndrii Nakryiko * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 40985c507329SPedro Tammela * *flags* must be 0. 4099457f4436SAndrii Nakryiko * Return 4100457f4436SAndrii Nakryiko * Valid pointer with *size* bytes of memory available; NULL, 4101457f4436SAndrii Nakryiko * otherwise. 4102457f4436SAndrii Nakryiko * 4103457f4436SAndrii Nakryiko * void bpf_ringbuf_submit(void *data, u64 flags) 4104457f4436SAndrii Nakryiko * Description 4105457f4436SAndrii Nakryiko * Submit reserved ring buffer sample, pointed to by *data*. 4106bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4107bcc7f554SQuentin Monnet * of new data availability is sent. 4108bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4109bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 41105c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 41115c507329SPedro Tammela * of new data availability is sent. 41125c507329SPedro Tammela * 41135c507329SPedro Tammela * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4114457f4436SAndrii Nakryiko * Return 4115457f4436SAndrii Nakryiko * Nothing. Always succeeds. 4116457f4436SAndrii Nakryiko * 4117457f4436SAndrii Nakryiko * void bpf_ringbuf_discard(void *data, u64 flags) 4118457f4436SAndrii Nakryiko * Description 4119457f4436SAndrii Nakryiko * Discard reserved ring buffer sample, pointed to by *data*. 4120bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4121bcc7f554SQuentin Monnet * of new data availability is sent. 4122bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4123bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 41245c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 41255c507329SPedro Tammela * of new data availability is sent. 41265c507329SPedro Tammela * 41275c507329SPedro Tammela * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4128457f4436SAndrii Nakryiko * Return 4129457f4436SAndrii Nakryiko * Nothing. Always succeeds. 4130457f4436SAndrii Nakryiko * 4131457f4436SAndrii Nakryiko * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4132457f4436SAndrii Nakryiko * Description 4133457f4436SAndrii Nakryiko * Query various characteristics of provided ring buffer. What 4134457f4436SAndrii Nakryiko * exactly is queries is determined by *flags*: 4135bcc7f554SQuentin Monnet * 4136bcc7f554SQuentin Monnet * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4137bcc7f554SQuentin Monnet * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4138bcc7f554SQuentin Monnet * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4139bcc7f554SQuentin Monnet * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4140bcc7f554SQuentin Monnet * 4141bcc7f554SQuentin Monnet * Data returned is just a momentary snapshot of actual values 4142457f4436SAndrii Nakryiko * and could be inaccurate, so this facility should be used to 4143457f4436SAndrii Nakryiko * power heuristics and for reporting, not to make 100% correct 4144457f4436SAndrii Nakryiko * calculation. 4145457f4436SAndrii Nakryiko * Return 4146bcc7f554SQuentin Monnet * Requested value, or 0, if *flags* are not recognized. 41477cdec54fSDaniel Borkmann * 4148bdb7b79bSAndrii Nakryiko * long bpf_csum_level(struct sk_buff *skb, u64 level) 41497cdec54fSDaniel Borkmann * Description 41507cdec54fSDaniel Borkmann * Change the skbs checksum level by one layer up or down, or 41517cdec54fSDaniel Borkmann * reset it entirely to none in order to have the stack perform 41527cdec54fSDaniel Borkmann * checksum validation. The level is applicable to the following 41537cdec54fSDaniel Borkmann * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 41547cdec54fSDaniel Borkmann * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 41557cdec54fSDaniel Borkmann * through **bpf_skb_adjust_room**\ () helper with passing in 41567cdec54fSDaniel Borkmann * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 41577cdec54fSDaniel Borkmann * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 41587cdec54fSDaniel Borkmann * the UDP header is removed. Similarly, an encap of the latter 41597cdec54fSDaniel Borkmann * into the former could be accompanied by a helper call to 41607cdec54fSDaniel Borkmann * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 41617cdec54fSDaniel Borkmann * skb is still intended to be processed in higher layers of the 41627cdec54fSDaniel Borkmann * stack instead of just egressing at tc. 41637cdec54fSDaniel Borkmann * 41647cdec54fSDaniel Borkmann * There are three supported level settings at this time: 41657cdec54fSDaniel Borkmann * 41667cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 41677cdec54fSDaniel Borkmann * with CHECKSUM_UNNECESSARY. 41687cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 41697cdec54fSDaniel Borkmann * with CHECKSUM_UNNECESSARY. 41707cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 41717cdec54fSDaniel Borkmann * sets CHECKSUM_NONE to force checksum validation by the stack. 41727cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 41737cdec54fSDaniel Borkmann * skb->csum_level. 41747cdec54fSDaniel Borkmann * Return 41757cdec54fSDaniel Borkmann * 0 on success, or a negative error in case of failure. In the 41767cdec54fSDaniel Borkmann * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 41777cdec54fSDaniel Borkmann * is returned or the error code -EACCES in case the skb is not 41787cdec54fSDaniel Borkmann * subject to CHECKSUM_UNNECESSARY. 4179af7ec138SYonghong Song * 4180af7ec138SYonghong Song * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4181af7ec138SYonghong Song * Description 4182af7ec138SYonghong Song * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4183af7ec138SYonghong Song * Return 4184938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4185478cfbdfSYonghong Song * 4186478cfbdfSYonghong Song * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4187478cfbdfSYonghong Song * Description 4188478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4189478cfbdfSYonghong Song * Return 4190938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4191478cfbdfSYonghong Song * 4192478cfbdfSYonghong Song * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4193478cfbdfSYonghong Song * Description 4194478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4195478cfbdfSYonghong Song * Return 4196938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4197478cfbdfSYonghong Song * 4198478cfbdfSYonghong Song * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4199478cfbdfSYonghong Song * Description 4200478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4201478cfbdfSYonghong Song * Return 4202938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 42030d4fad3eSYonghong Song * 42040d4fad3eSYonghong Song * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 42050d4fad3eSYonghong Song * Description 42060d4fad3eSYonghong Song * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 42070d4fad3eSYonghong Song * Return 4208938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4209fa28dcb8SSong Liu * 4210fa28dcb8SSong Liu * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4211fa28dcb8SSong Liu * Description 4212fa28dcb8SSong Liu * Return a user or a kernel stack in bpf program provided buffer. 4213fa28dcb8SSong Liu * To achieve this, the helper needs *task*, which is a valid 4214938c3efdSQuentin Monnet * pointer to **struct task_struct**. To store the stacktrace, the 4215fa28dcb8SSong Liu * bpf program provides *buf* with a nonnegative *size*. 4216fa28dcb8SSong Liu * 4217fa28dcb8SSong Liu * The last argument, *flags*, holds the number of stack frames to 4218fa28dcb8SSong Liu * skip (from 0 to 255), masked with 4219fa28dcb8SSong Liu * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4220fa28dcb8SSong Liu * the following flags: 4221fa28dcb8SSong Liu * 4222fa28dcb8SSong Liu * **BPF_F_USER_STACK** 4223fa28dcb8SSong Liu * Collect a user space stack instead of a kernel stack. 4224fa28dcb8SSong Liu * **BPF_F_USER_BUILD_ID** 4225fa28dcb8SSong Liu * Collect buildid+offset instead of ips for user stack, 4226fa28dcb8SSong Liu * only valid if **BPF_F_USER_STACK** is also specified. 4227fa28dcb8SSong Liu * 4228fa28dcb8SSong Liu * **bpf_get_task_stack**\ () can collect up to 4229fa28dcb8SSong Liu * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4230fa28dcb8SSong Liu * to sufficient large buffer size. Note that 4231fa28dcb8SSong Liu * this limit can be controlled with the **sysctl** program, and 4232fa28dcb8SSong Liu * that it should be manually increased in order to profile long 4233fa28dcb8SSong Liu * user stacks (such as stacks for Java programs). To do so, use: 4234fa28dcb8SSong Liu * 4235fa28dcb8SSong Liu * :: 4236fa28dcb8SSong Liu * 4237fa28dcb8SSong Liu * # sysctl kernel.perf_event_max_stack=<new value> 4238fa28dcb8SSong Liu * Return 4239fa28dcb8SSong Liu * A non-negative value equal to or less than *size* on success, 4240fa28dcb8SSong Liu * or a negative error in case of failure. 4241fa28dcb8SSong Liu * 42420813a841SMartin KaFai Lau * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 42430813a841SMartin KaFai Lau * Description 42440813a841SMartin KaFai Lau * Load header option. Support reading a particular TCP header 4245938c3efdSQuentin Monnet * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 42460813a841SMartin KaFai Lau * 42470813a841SMartin KaFai Lau * If *flags* is 0, it will search the option from the 4248938c3efdSQuentin Monnet * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 42490813a841SMartin KaFai Lau * has details on what skb_data contains under different 4250938c3efdSQuentin Monnet * *skops*\ **->op**. 42510813a841SMartin KaFai Lau * 42520813a841SMartin KaFai Lau * The first byte of the *searchby_res* specifies the 42530813a841SMartin KaFai Lau * kind that it wants to search. 42540813a841SMartin KaFai Lau * 42550813a841SMartin KaFai Lau * If the searching kind is an experimental kind 42560813a841SMartin KaFai Lau * (i.e. 253 or 254 according to RFC6994). It also 42570813a841SMartin KaFai Lau * needs to specify the "magic" which is either 42580813a841SMartin KaFai Lau * 2 bytes or 4 bytes. It then also needs to 42590813a841SMartin KaFai Lau * specify the size of the magic by using 42600813a841SMartin KaFai Lau * the 2nd byte which is "kind-length" of a TCP 42610813a841SMartin KaFai Lau * header option and the "kind-length" also 42620813a841SMartin KaFai Lau * includes the first 2 bytes "kind" and "kind-length" 42630813a841SMartin KaFai Lau * itself as a normal TCP header option also does. 42640813a841SMartin KaFai Lau * 42650813a841SMartin KaFai Lau * For example, to search experimental kind 254 with 42660813a841SMartin KaFai Lau * 2 byte magic 0xeB9F, the searchby_res should be 42670813a841SMartin KaFai Lau * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 42680813a841SMartin KaFai Lau * 42690813a841SMartin KaFai Lau * To search for the standard window scale option (3), 4270938c3efdSQuentin Monnet * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 42710813a841SMartin KaFai Lau * Note, kind-length must be 0 for regular option. 42720813a841SMartin KaFai Lau * 42730813a841SMartin KaFai Lau * Searching for No-Op (0) and End-of-Option-List (1) are 42740813a841SMartin KaFai Lau * not supported. 42750813a841SMartin KaFai Lau * 42760813a841SMartin KaFai Lau * *len* must be at least 2 bytes which is the minimal size 42770813a841SMartin KaFai Lau * of a header option. 42780813a841SMartin KaFai Lau * 42790813a841SMartin KaFai Lau * Supported flags: 4280938c3efdSQuentin Monnet * 42810813a841SMartin KaFai Lau * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 42820813a841SMartin KaFai Lau * saved_syn packet or the just-received syn packet. 42830813a841SMartin KaFai Lau * 42840813a841SMartin KaFai Lau * Return 42850813a841SMartin KaFai Lau * > 0 when found, the header option is copied to *searchby_res*. 4286938c3efdSQuentin Monnet * The return value is the total length copied. On failure, a 4287938c3efdSQuentin Monnet * negative error code is returned: 42880813a841SMartin KaFai Lau * 4289938c3efdSQuentin Monnet * **-EINVAL** if a parameter is invalid. 42900813a841SMartin KaFai Lau * 4291938c3efdSQuentin Monnet * **-ENOMSG** if the option is not found. 42920813a841SMartin KaFai Lau * 4293938c3efdSQuentin Monnet * **-ENOENT** if no syn packet is available when 4294938c3efdSQuentin Monnet * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 42950813a841SMartin KaFai Lau * 4296938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space. Only *len* number of 42970813a841SMartin KaFai Lau * bytes are copied. 42980813a841SMartin KaFai Lau * 4299938c3efdSQuentin Monnet * **-EFAULT** on failure to parse the header options in the 4300938c3efdSQuentin Monnet * packet. 43010813a841SMartin KaFai Lau * 4302938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4303938c3efdSQuentin Monnet * *skops*\ **->op**. 43040813a841SMartin KaFai Lau * 43050813a841SMartin KaFai Lau * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 43060813a841SMartin KaFai Lau * Description 43070813a841SMartin KaFai Lau * Store header option. The data will be copied 43080813a841SMartin KaFai Lau * from buffer *from* with length *len* to the TCP header. 43090813a841SMartin KaFai Lau * 43100813a841SMartin KaFai Lau * The buffer *from* should have the whole option that 43110813a841SMartin KaFai Lau * includes the kind, kind-length, and the actual 43120813a841SMartin KaFai Lau * option data. The *len* must be at least kind-length 43130813a841SMartin KaFai Lau * long. The kind-length does not have to be 4 byte 43140813a841SMartin KaFai Lau * aligned. The kernel will take care of the padding 43150813a841SMartin KaFai Lau * and setting the 4 bytes aligned value to th->doff. 43160813a841SMartin KaFai Lau * 43170813a841SMartin KaFai Lau * This helper will check for duplicated option 43180813a841SMartin KaFai Lau * by searching the same option in the outgoing skb. 43190813a841SMartin KaFai Lau * 43200813a841SMartin KaFai Lau * This helper can only be called during 4321938c3efdSQuentin Monnet * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 43220813a841SMartin KaFai Lau * 43230813a841SMartin KaFai Lau * Return 43240813a841SMartin KaFai Lau * 0 on success, or negative error in case of failure: 43250813a841SMartin KaFai Lau * 4326938c3efdSQuentin Monnet * **-EINVAL** If param is invalid. 43270813a841SMartin KaFai Lau * 4328938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space in the header. 43290813a841SMartin KaFai Lau * Nothing has been written 43300813a841SMartin KaFai Lau * 4331938c3efdSQuentin Monnet * **-EEXIST** if the option already exists. 43320813a841SMartin KaFai Lau * 4333938c3efdSQuentin Monnet * **-EFAULT** on failrue to parse the existing header options. 43340813a841SMartin KaFai Lau * 4335938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4336938c3efdSQuentin Monnet * *skops*\ **->op**. 43370813a841SMartin KaFai Lau * 43380813a841SMartin KaFai Lau * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 43390813a841SMartin KaFai Lau * Description 43400813a841SMartin KaFai Lau * Reserve *len* bytes for the bpf header option. The 4341938c3efdSQuentin Monnet * space will be used by **bpf_store_hdr_opt**\ () later in 4342938c3efdSQuentin Monnet * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 43430813a841SMartin KaFai Lau * 4344938c3efdSQuentin Monnet * If **bpf_reserve_hdr_opt**\ () is called multiple times, 43450813a841SMartin KaFai Lau * the total number of bytes will be reserved. 43460813a841SMartin KaFai Lau * 43470813a841SMartin KaFai Lau * This helper can only be called during 4348938c3efdSQuentin Monnet * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 43490813a841SMartin KaFai Lau * 43500813a841SMartin KaFai Lau * Return 43510813a841SMartin KaFai Lau * 0 on success, or negative error in case of failure: 43520813a841SMartin KaFai Lau * 4353938c3efdSQuentin Monnet * **-EINVAL** if a parameter is invalid. 43540813a841SMartin KaFai Lau * 4355938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space in the header. 43560813a841SMartin KaFai Lau * 4357938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4358938c3efdSQuentin Monnet * *skops*\ **->op**. 43596e22ab9dSJiri Olsa * 43608ea63684SKP Singh * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 43618ea63684SKP Singh * Description 43628ea63684SKP Singh * Get a bpf_local_storage from an *inode*. 43638ea63684SKP Singh * 43648ea63684SKP Singh * Logically, it could be thought of as getting the value from 43658ea63684SKP Singh * a *map* with *inode* as the **key**. From this 43668ea63684SKP Singh * perspective, the usage is not much different from 43678ea63684SKP Singh * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 43688ea63684SKP Singh * helper enforces the key must be an inode and the map must also 43698ea63684SKP Singh * be a **BPF_MAP_TYPE_INODE_STORAGE**. 43708ea63684SKP Singh * 43718ea63684SKP Singh * Underneath, the value is stored locally at *inode* instead of 43728ea63684SKP Singh * the *map*. The *map* is used as the bpf-local-storage 43738ea63684SKP Singh * "type". The bpf-local-storage "type" (i.e. the *map*) is 43748ea63684SKP Singh * searched against all bpf_local_storage residing at *inode*. 43758ea63684SKP Singh * 43768ea63684SKP Singh * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 43778ea63684SKP Singh * used such that a new bpf_local_storage will be 43788ea63684SKP Singh * created if one does not exist. *value* can be used 43798ea63684SKP Singh * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 43808ea63684SKP Singh * the initial value of a bpf_local_storage. If *value* is 43818ea63684SKP Singh * **NULL**, the new bpf_local_storage will be zero initialized. 43828ea63684SKP Singh * Return 43838ea63684SKP Singh * A bpf_local_storage pointer is returned on success. 43848ea63684SKP Singh * 43858ea63684SKP Singh * **NULL** if not found or there was an error in adding 43868ea63684SKP Singh * a new bpf_local_storage. 43878ea63684SKP Singh * 43888ea63684SKP Singh * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 43898ea63684SKP Singh * Description 43908ea63684SKP Singh * Delete a bpf_local_storage from an *inode*. 43918ea63684SKP Singh * Return 43928ea63684SKP Singh * 0 on success. 43938ea63684SKP Singh * 43948ea63684SKP Singh * **-ENOENT** if the bpf_local_storage cannot be found. 43956e22ab9dSJiri Olsa * 43966e22ab9dSJiri Olsa * long bpf_d_path(struct path *path, char *buf, u32 sz) 43976e22ab9dSJiri Olsa * Description 4398938c3efdSQuentin Monnet * Return full path for given **struct path** object, which 4399938c3efdSQuentin Monnet * needs to be the kernel BTF *path* object. The path is 4400938c3efdSQuentin Monnet * returned in the provided buffer *buf* of size *sz* and 44016e22ab9dSJiri Olsa * is zero terminated. 44026e22ab9dSJiri Olsa * 44036e22ab9dSJiri Olsa * Return 44046e22ab9dSJiri Olsa * On success, the strictly positive length of the string, 44056e22ab9dSJiri Olsa * including the trailing NUL character. On error, a negative 44066e22ab9dSJiri Olsa * value. 440707be4c4aSAlexei Starovoitov * 440807be4c4aSAlexei Starovoitov * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 440907be4c4aSAlexei Starovoitov * Description 441007be4c4aSAlexei Starovoitov * Read *size* bytes from user space address *user_ptr* and store 4411938c3efdSQuentin Monnet * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 441207be4c4aSAlexei Starovoitov * Return 441307be4c4aSAlexei Starovoitov * 0 on success, or a negative error in case of failure. 4414c4d0bfb4SAlan Maguire * 4415c4d0bfb4SAlan Maguire * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4416c4d0bfb4SAlan Maguire * Description 4417c4d0bfb4SAlan Maguire * Use BTF to store a string representation of *ptr*->ptr in *str*, 4418c4d0bfb4SAlan Maguire * using *ptr*->type_id. This value should specify the type 4419c4d0bfb4SAlan Maguire * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4420c4d0bfb4SAlan Maguire * can be used to look up vmlinux BTF type ids. Traversing the 4421c4d0bfb4SAlan Maguire * data structure using BTF, the type information and values are 4422c4d0bfb4SAlan Maguire * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4423c4d0bfb4SAlan Maguire * the pointer data is carried out to avoid kernel crashes during 4424c4d0bfb4SAlan Maguire * operation. Smaller types can use string space on the stack; 4425c4d0bfb4SAlan Maguire * larger programs can use map data to store the string 4426c4d0bfb4SAlan Maguire * representation. 4427c4d0bfb4SAlan Maguire * 4428c4d0bfb4SAlan Maguire * The string can be subsequently shared with userspace via 4429c4d0bfb4SAlan Maguire * bpf_perf_event_output() or ring buffer interfaces. 4430c4d0bfb4SAlan Maguire * bpf_trace_printk() is to be avoided as it places too small 4431c4d0bfb4SAlan Maguire * a limit on string size to be useful. 4432c4d0bfb4SAlan Maguire * 4433c4d0bfb4SAlan Maguire * *flags* is a combination of 4434c4d0bfb4SAlan Maguire * 4435c4d0bfb4SAlan Maguire * **BTF_F_COMPACT** 4436c4d0bfb4SAlan Maguire * no formatting around type information 4437c4d0bfb4SAlan Maguire * **BTF_F_NONAME** 4438c4d0bfb4SAlan Maguire * no struct/union member names/types 4439c4d0bfb4SAlan Maguire * **BTF_F_PTR_RAW** 4440c4d0bfb4SAlan Maguire * show raw (unobfuscated) pointer values; 4441c4d0bfb4SAlan Maguire * equivalent to printk specifier %px. 4442c4d0bfb4SAlan Maguire * **BTF_F_ZERO** 4443c4d0bfb4SAlan Maguire * show zero-valued struct/union members; they 4444c4d0bfb4SAlan Maguire * are not displayed by default 4445c4d0bfb4SAlan Maguire * 4446c4d0bfb4SAlan Maguire * Return 4447c4d0bfb4SAlan Maguire * The number of bytes that were written (or would have been 4448c4d0bfb4SAlan Maguire * written if output had to be truncated due to string size), 4449c4d0bfb4SAlan Maguire * or a negative error in cases of failure. 4450eb411377SAlan Maguire * 4451eb411377SAlan Maguire * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4452eb411377SAlan Maguire * Description 4453eb411377SAlan Maguire * Use BTF to write to seq_write a string representation of 4454eb411377SAlan Maguire * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4455eb411377SAlan Maguire * *flags* are identical to those used for bpf_snprintf_btf. 4456eb411377SAlan Maguire * Return 4457eb411377SAlan Maguire * 0 on success or a negative error in case of failure. 4458b426ce83SDaniel Borkmann * 4459b426ce83SDaniel Borkmann * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4460b426ce83SDaniel Borkmann * Description 4461b426ce83SDaniel Borkmann * See **bpf_get_cgroup_classid**\ () for the main description. 4462b426ce83SDaniel Borkmann * This helper differs from **bpf_get_cgroup_classid**\ () in that 4463b426ce83SDaniel Borkmann * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4464b426ce83SDaniel Borkmann * associated socket instead of the current process. 4465b426ce83SDaniel Borkmann * Return 4466b426ce83SDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 4467b4ab3141SDaniel Borkmann * 4468ba452c9eSToke Høiland-Jørgensen * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4469b4ab3141SDaniel Borkmann * Description 4470b4ab3141SDaniel Borkmann * Redirect the packet to another net device of index *ifindex* 4471b4ab3141SDaniel Borkmann * and fill in L2 addresses from neighboring subsystem. This helper 4472b4ab3141SDaniel Borkmann * is somewhat similar to **bpf_redirect**\ (), except that it 4473dd2ce6a5SDaniel Borkmann * populates L2 addresses as well, meaning, internally, the helper 4474ba452c9eSToke Høiland-Jørgensen * relies on the neighbor lookup for the L2 address of the nexthop. 4475ba452c9eSToke Høiland-Jørgensen * 4476ba452c9eSToke Høiland-Jørgensen * The helper will perform a FIB lookup based on the skb's 4477ba452c9eSToke Høiland-Jørgensen * networking header to get the address of the next hop, unless 4478ba452c9eSToke Høiland-Jørgensen * this is supplied by the caller in the *params* argument. The 4479ba452c9eSToke Høiland-Jørgensen * *plen* argument indicates the len of *params* and should be set 4480ba452c9eSToke Høiland-Jørgensen * to 0 if *params* is NULL. 4481dd2ce6a5SDaniel Borkmann * 4482b4ab3141SDaniel Borkmann * The *flags* argument is reserved and must be 0. The helper is 4483dd2ce6a5SDaniel Borkmann * currently only supported for tc BPF program types, and enabled 4484dd2ce6a5SDaniel Borkmann * for IPv4 and IPv6 protocols. 4485b4ab3141SDaniel Borkmann * Return 4486b4ab3141SDaniel Borkmann * The helper returns **TC_ACT_REDIRECT** on success or 4487b4ab3141SDaniel Borkmann * **TC_ACT_SHOT** on error. 4488eaa6bcb7SHao Luo * 4489eaa6bcb7SHao Luo * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4490eaa6bcb7SHao Luo * Description 4491eaa6bcb7SHao Luo * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4492eaa6bcb7SHao Luo * pointer to the percpu kernel variable on *cpu*. A ksym is an 4493eaa6bcb7SHao Luo * extern variable decorated with '__ksym'. For ksym, there is a 4494eaa6bcb7SHao Luo * global var (either static or global) defined of the same name 4495eaa6bcb7SHao Luo * in the kernel. The ksym is percpu if the global var is percpu. 4496eaa6bcb7SHao Luo * The returned pointer points to the global percpu var on *cpu*. 4497eaa6bcb7SHao Luo * 4498eaa6bcb7SHao Luo * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4499eaa6bcb7SHao Luo * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4500eaa6bcb7SHao Luo * happens if *cpu* is larger than nr_cpu_ids. The caller of 4501eaa6bcb7SHao Luo * bpf_per_cpu_ptr() must check the returned value. 4502eaa6bcb7SHao Luo * Return 4503eaa6bcb7SHao Luo * A pointer pointing to the kernel percpu variable on *cpu*, or 4504eaa6bcb7SHao Luo * NULL, if *cpu* is invalid. 450563d9b80dSHao Luo * 450663d9b80dSHao Luo * void *bpf_this_cpu_ptr(const void *percpu_ptr) 450763d9b80dSHao Luo * Description 450863d9b80dSHao Luo * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 450963d9b80dSHao Luo * pointer to the percpu kernel variable on this cpu. See the 451063d9b80dSHao Luo * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 451163d9b80dSHao Luo * 451263d9b80dSHao Luo * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 451363d9b80dSHao Luo * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 451463d9b80dSHao Luo * never return NULL. 451563d9b80dSHao Luo * Return 451663d9b80dSHao Luo * A pointer pointing to the kernel percpu variable on this cpu. 45179aa1206eSDaniel Borkmann * 45189aa1206eSDaniel Borkmann * long bpf_redirect_peer(u32 ifindex, u64 flags) 45199aa1206eSDaniel Borkmann * Description 45209aa1206eSDaniel Borkmann * Redirect the packet to another net device of index *ifindex*. 45219aa1206eSDaniel Borkmann * This helper is somewhat similar to **bpf_redirect**\ (), except 45229aa1206eSDaniel Borkmann * that the redirection happens to the *ifindex*' peer device and 45239aa1206eSDaniel Borkmann * the netns switch takes place from ingress to ingress without 45249aa1206eSDaniel Borkmann * going through the CPU's backlog queue. 45259aa1206eSDaniel Borkmann * 45269aa1206eSDaniel Borkmann * The *flags* argument is reserved and must be 0. The helper is 45279aa1206eSDaniel Borkmann * currently only supported for tc BPF program types at the ingress 45289aa1206eSDaniel Borkmann * hook and for veth device types. The peer device must reside in a 45299aa1206eSDaniel Borkmann * different network namespace. 45309aa1206eSDaniel Borkmann * Return 45319aa1206eSDaniel Borkmann * The helper returns **TC_ACT_REDIRECT** on success or 45329aa1206eSDaniel Borkmann * **TC_ACT_SHOT** on error. 45334cf1bc1fSKP Singh * 45344cf1bc1fSKP Singh * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 45354cf1bc1fSKP Singh * Description 45364cf1bc1fSKP Singh * Get a bpf_local_storage from the *task*. 45374cf1bc1fSKP Singh * 45384cf1bc1fSKP Singh * Logically, it could be thought of as getting the value from 45394cf1bc1fSKP Singh * a *map* with *task* as the **key**. From this 45404cf1bc1fSKP Singh * perspective, the usage is not much different from 45414cf1bc1fSKP Singh * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 45424cf1bc1fSKP Singh * helper enforces the key must be an task_struct and the map must also 45434cf1bc1fSKP Singh * be a **BPF_MAP_TYPE_TASK_STORAGE**. 45444cf1bc1fSKP Singh * 45454cf1bc1fSKP Singh * Underneath, the value is stored locally at *task* instead of 45464cf1bc1fSKP Singh * the *map*. The *map* is used as the bpf-local-storage 45474cf1bc1fSKP Singh * "type". The bpf-local-storage "type" (i.e. the *map*) is 45484cf1bc1fSKP Singh * searched against all bpf_local_storage residing at *task*. 45494cf1bc1fSKP Singh * 45504cf1bc1fSKP Singh * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 45514cf1bc1fSKP Singh * used such that a new bpf_local_storage will be 45524cf1bc1fSKP Singh * created if one does not exist. *value* can be used 45534cf1bc1fSKP Singh * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 45544cf1bc1fSKP Singh * the initial value of a bpf_local_storage. If *value* is 45554cf1bc1fSKP Singh * **NULL**, the new bpf_local_storage will be zero initialized. 45564cf1bc1fSKP Singh * Return 45574cf1bc1fSKP Singh * A bpf_local_storage pointer is returned on success. 45584cf1bc1fSKP Singh * 45594cf1bc1fSKP Singh * **NULL** if not found or there was an error in adding 45604cf1bc1fSKP Singh * a new bpf_local_storage. 45614cf1bc1fSKP Singh * 45624cf1bc1fSKP Singh * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 45634cf1bc1fSKP Singh * Description 45644cf1bc1fSKP Singh * Delete a bpf_local_storage from a *task*. 45654cf1bc1fSKP Singh * Return 45664cf1bc1fSKP Singh * 0 on success. 45674cf1bc1fSKP Singh * 45684cf1bc1fSKP Singh * **-ENOENT** if the bpf_local_storage cannot be found. 45693ca1032aSKP Singh * 45703ca1032aSKP Singh * struct task_struct *bpf_get_current_task_btf(void) 45713ca1032aSKP Singh * Description 45723ca1032aSKP Singh * Return a BTF pointer to the "current" task. 45733ca1032aSKP Singh * This pointer can also be used in helpers that accept an 45743ca1032aSKP Singh * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 45753ca1032aSKP Singh * Return 45763ca1032aSKP Singh * Pointer to the current task. 45773f6719c7SKP Singh * 45783f6719c7SKP Singh * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 45793f6719c7SKP Singh * Description 45803f6719c7SKP Singh * Set or clear certain options on *bprm*: 45813f6719c7SKP Singh * 45823f6719c7SKP Singh * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 45833f6719c7SKP Singh * which sets the **AT_SECURE** auxv for glibc. The bit 45843f6719c7SKP Singh * is cleared if the flag is not specified. 45853f6719c7SKP Singh * Return 45863f6719c7SKP Singh * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4587d0551261SDmitrii Banshchikov * 4588d0551261SDmitrii Banshchikov * u64 bpf_ktime_get_coarse_ns(void) 4589d0551261SDmitrii Banshchikov * Description 4590d0551261SDmitrii Banshchikov * Return a coarse-grained version of the time elapsed since 4591d0551261SDmitrii Banshchikov * system boot, in nanoseconds. Does not include time the system 4592d0551261SDmitrii Banshchikov * was suspended. 4593d0551261SDmitrii Banshchikov * 4594d0551261SDmitrii Banshchikov * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4595d0551261SDmitrii Banshchikov * Return 4596d0551261SDmitrii Banshchikov * Current *ktime*. 459727672f0dSKP Singh * 459827672f0dSKP Singh * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 459927672f0dSKP Singh * Description 460027672f0dSKP Singh * Returns the stored IMA hash of the *inode* (if it's avaialable). 460127672f0dSKP Singh * If the hash is larger than *size*, then only *size* 460227672f0dSKP Singh * bytes will be copied to *dst* 460327672f0dSKP Singh * Return 460427672f0dSKP Singh * The **hash_algo** is returned on success, 460527672f0dSKP Singh * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 460627672f0dSKP Singh * invalid arguments are passed. 46074f19cab7SFlorent Revest * 46084f19cab7SFlorent Revest * struct socket *bpf_sock_from_file(struct file *file) 46094f19cab7SFlorent Revest * Description 46104f19cab7SFlorent Revest * If the given file represents a socket, returns the associated 46114f19cab7SFlorent Revest * socket. 46124f19cab7SFlorent Revest * Return 46134f19cab7SFlorent Revest * A pointer to a struct socket on success or NULL if the file is 46144f19cab7SFlorent Revest * not a socket. 461534b2021cSJesper Dangaard Brouer * 461634b2021cSJesper Dangaard Brouer * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 461734b2021cSJesper Dangaard Brouer * Description 4618e5e35e75SJesper Dangaard Brouer * Check packet size against exceeding MTU of net device (based 461934b2021cSJesper Dangaard Brouer * on *ifindex*). This helper will likely be used in combination 462034b2021cSJesper Dangaard Brouer * with helpers that adjust/change the packet size. 462134b2021cSJesper Dangaard Brouer * 462234b2021cSJesper Dangaard Brouer * The argument *len_diff* can be used for querying with a planned 462334b2021cSJesper Dangaard Brouer * size change. This allows to check MTU prior to changing packet 462434b2021cSJesper Dangaard Brouer * ctx. Providing an *len_diff* adjustment that is larger than the 462534b2021cSJesper Dangaard Brouer * actual packet size (resulting in negative packet size) will in 462634b2021cSJesper Dangaard Brouer * principle not exceed the MTU, why it is not considered a 462734b2021cSJesper Dangaard Brouer * failure. Other BPF-helpers are needed for performing the 462834b2021cSJesper Dangaard Brouer * planned size change, why the responsability for catch a negative 462934b2021cSJesper Dangaard Brouer * packet size belong in those helpers. 463034b2021cSJesper Dangaard Brouer * 463134b2021cSJesper Dangaard Brouer * Specifying *ifindex* zero means the MTU check is performed 463234b2021cSJesper Dangaard Brouer * against the current net device. This is practical if this isn't 463334b2021cSJesper Dangaard Brouer * used prior to redirect. 463434b2021cSJesper Dangaard Brouer * 4635e5e35e75SJesper Dangaard Brouer * On input *mtu_len* must be a valid pointer, else verifier will 4636e5e35e75SJesper Dangaard Brouer * reject BPF program. If the value *mtu_len* is initialized to 4637e5e35e75SJesper Dangaard Brouer * zero then the ctx packet size is use. When value *mtu_len* is 4638e5e35e75SJesper Dangaard Brouer * provided as input this specify the L3 length that the MTU check 4639e5e35e75SJesper Dangaard Brouer * is done against. Remember XDP and TC length operate at L2, but 4640e5e35e75SJesper Dangaard Brouer * this value is L3 as this correlate to MTU and IP-header tot_len 4641e5e35e75SJesper Dangaard Brouer * values which are L3 (similar behavior as bpf_fib_lookup). 4642e5e35e75SJesper Dangaard Brouer * 464334b2021cSJesper Dangaard Brouer * The Linux kernel route table can configure MTUs on a more 464434b2021cSJesper Dangaard Brouer * specific per route level, which is not provided by this helper. 464534b2021cSJesper Dangaard Brouer * For route level MTU checks use the **bpf_fib_lookup**\ () 464634b2021cSJesper Dangaard Brouer * helper. 464734b2021cSJesper Dangaard Brouer * 464834b2021cSJesper Dangaard Brouer * *ctx* is either **struct xdp_md** for XDP programs or 464934b2021cSJesper Dangaard Brouer * **struct sk_buff** for tc cls_act programs. 465034b2021cSJesper Dangaard Brouer * 465134b2021cSJesper Dangaard Brouer * The *flags* argument can be a combination of one or more of the 465234b2021cSJesper Dangaard Brouer * following values: 465334b2021cSJesper Dangaard Brouer * 465434b2021cSJesper Dangaard Brouer * **BPF_MTU_CHK_SEGS** 465534b2021cSJesper Dangaard Brouer * This flag will only works for *ctx* **struct sk_buff**. 465634b2021cSJesper Dangaard Brouer * If packet context contains extra packet segment buffers 465734b2021cSJesper Dangaard Brouer * (often knows as GSO skb), then MTU check is harder to 465834b2021cSJesper Dangaard Brouer * check at this point, because in transmit path it is 465934b2021cSJesper Dangaard Brouer * possible for the skb packet to get re-segmented 466034b2021cSJesper Dangaard Brouer * (depending on net device features). This could still be 466134b2021cSJesper Dangaard Brouer * a MTU violation, so this flag enables performing MTU 466234b2021cSJesper Dangaard Brouer * check against segments, with a different violation 466334b2021cSJesper Dangaard Brouer * return code to tell it apart. Check cannot use len_diff. 466434b2021cSJesper Dangaard Brouer * 466534b2021cSJesper Dangaard Brouer * On return *mtu_len* pointer contains the MTU value of the net 466634b2021cSJesper Dangaard Brouer * device. Remember the net device configured MTU is the L3 size, 4667e5e35e75SJesper Dangaard Brouer * which is returned here and XDP and TC length operate at L2. 466834b2021cSJesper Dangaard Brouer * Helper take this into account for you, but remember when using 4669e5e35e75SJesper Dangaard Brouer * MTU value in your BPF-code. 467034b2021cSJesper Dangaard Brouer * 467134b2021cSJesper Dangaard Brouer * Return 467234b2021cSJesper Dangaard Brouer * * 0 on success, and populate MTU value in *mtu_len* pointer. 467334b2021cSJesper Dangaard Brouer * 467434b2021cSJesper Dangaard Brouer * * < 0 if any input argument is invalid (*mtu_len* not updated) 467534b2021cSJesper Dangaard Brouer * 467634b2021cSJesper Dangaard Brouer * MTU violations return positive values, but also populate MTU 467734b2021cSJesper Dangaard Brouer * value in *mtu_len* pointer, as this can be needed for 467834b2021cSJesper Dangaard Brouer * implementing PMTU handing: 467934b2021cSJesper Dangaard Brouer * 468034b2021cSJesper Dangaard Brouer * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 468134b2021cSJesper Dangaard Brouer * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 468234b2021cSJesper Dangaard Brouer * 468369c087baSYonghong Song * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 468469c087baSYonghong Song * Description 468569c087baSYonghong Song * For each element in **map**, call **callback_fn** function with 468669c087baSYonghong Song * **map**, **callback_ctx** and other map-specific parameters. 468769c087baSYonghong Song * The **callback_fn** should be a static function and 468869c087baSYonghong Song * the **callback_ctx** should be a pointer to the stack. 468969c087baSYonghong Song * The **flags** is used to control certain aspects of the helper. 469069c087baSYonghong Song * Currently, the **flags** must be 0. 469169c087baSYonghong Song * 469269c087baSYonghong Song * The following are a list of supported map types and their 469369c087baSYonghong Song * respective expected callback signatures: 469469c087baSYonghong Song * 469569c087baSYonghong Song * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 469669c087baSYonghong Song * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 469769c087baSYonghong Song * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 469869c087baSYonghong Song * 469969c087baSYonghong Song * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 470069c087baSYonghong Song * 470169c087baSYonghong Song * For per_cpu maps, the map_value is the value on the cpu where the 470269c087baSYonghong Song * bpf_prog is running. 470369c087baSYonghong Song * 470469c087baSYonghong Song * If **callback_fn** return 0, the helper will continue to the next 470569c087baSYonghong Song * element. If return value is 1, the helper will skip the rest of 470669c087baSYonghong Song * elements and return. Other return values are not used now. 470769c087baSYonghong Song * 470869c087baSYonghong Song * Return 470969c087baSYonghong Song * The number of traversed map elements for success, **-EINVAL** for 471069c087baSYonghong Song * invalid **flags**. 4711*7b15523aSFlorent Revest * 4712*7b15523aSFlorent Revest * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 4713*7b15523aSFlorent Revest * Description 4714*7b15523aSFlorent Revest * Outputs a string into the **str** buffer of size **str_size** 4715*7b15523aSFlorent Revest * based on a format string stored in a read-only map pointed by 4716*7b15523aSFlorent Revest * **fmt**. 4717*7b15523aSFlorent Revest * 4718*7b15523aSFlorent Revest * Each format specifier in **fmt** corresponds to one u64 element 4719*7b15523aSFlorent Revest * in the **data** array. For strings and pointers where pointees 4720*7b15523aSFlorent Revest * are accessed, only the pointer values are stored in the *data* 4721*7b15523aSFlorent Revest * array. The *data_len* is the size of *data* in bytes. 4722*7b15523aSFlorent Revest * 4723*7b15523aSFlorent Revest * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 4724*7b15523aSFlorent Revest * memory. Reading kernel memory may fail due to either invalid 4725*7b15523aSFlorent Revest * address or valid address but requiring a major memory fault. If 4726*7b15523aSFlorent Revest * reading kernel memory fails, the string for **%s** will be an 4727*7b15523aSFlorent Revest * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 4728*7b15523aSFlorent Revest * Not returning error to bpf program is consistent with what 4729*7b15523aSFlorent Revest * **bpf_trace_printk**\ () does for now. 4730*7b15523aSFlorent Revest * 4731*7b15523aSFlorent Revest * Return 4732*7b15523aSFlorent Revest * The strictly positive length of the formatted string, including 4733*7b15523aSFlorent Revest * the trailing zero character. If the return value is greater than 4734*7b15523aSFlorent Revest * **str_size**, **str** contains a truncated string, guaranteed to 4735*7b15523aSFlorent Revest * be zero-terminated except when **str_size** is 0. 4736*7b15523aSFlorent Revest * 4737*7b15523aSFlorent Revest * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 47387a4b28c6SDaniel Borkmann */ 4739ebb676daSThomas Graf #define __BPF_FUNC_MAPPER(FN) \ 4740ebb676daSThomas Graf FN(unspec), \ 4741ebb676daSThomas Graf FN(map_lookup_elem), \ 4742ebb676daSThomas Graf FN(map_update_elem), \ 4743ebb676daSThomas Graf FN(map_delete_elem), \ 4744ebb676daSThomas Graf FN(probe_read), \ 4745ebb676daSThomas Graf FN(ktime_get_ns), \ 4746ebb676daSThomas Graf FN(trace_printk), \ 4747ebb676daSThomas Graf FN(get_prandom_u32), \ 4748ebb676daSThomas Graf FN(get_smp_processor_id), \ 4749ebb676daSThomas Graf FN(skb_store_bytes), \ 4750ebb676daSThomas Graf FN(l3_csum_replace), \ 4751ebb676daSThomas Graf FN(l4_csum_replace), \ 4752ebb676daSThomas Graf FN(tail_call), \ 4753ebb676daSThomas Graf FN(clone_redirect), \ 4754ebb676daSThomas Graf FN(get_current_pid_tgid), \ 4755ebb676daSThomas Graf FN(get_current_uid_gid), \ 4756ebb676daSThomas Graf FN(get_current_comm), \ 4757ebb676daSThomas Graf FN(get_cgroup_classid), \ 4758ebb676daSThomas Graf FN(skb_vlan_push), \ 4759ebb676daSThomas Graf FN(skb_vlan_pop), \ 4760ebb676daSThomas Graf FN(skb_get_tunnel_key), \ 4761ebb676daSThomas Graf FN(skb_set_tunnel_key), \ 4762ebb676daSThomas Graf FN(perf_event_read), \ 4763ebb676daSThomas Graf FN(redirect), \ 4764ebb676daSThomas Graf FN(get_route_realm), \ 4765ebb676daSThomas Graf FN(perf_event_output), \ 4766ebb676daSThomas Graf FN(skb_load_bytes), \ 4767ebb676daSThomas Graf FN(get_stackid), \ 4768ebb676daSThomas Graf FN(csum_diff), \ 4769ebb676daSThomas Graf FN(skb_get_tunnel_opt), \ 4770ebb676daSThomas Graf FN(skb_set_tunnel_opt), \ 4771ebb676daSThomas Graf FN(skb_change_proto), \ 4772ebb676daSThomas Graf FN(skb_change_type), \ 4773ebb676daSThomas Graf FN(skb_under_cgroup), \ 4774ebb676daSThomas Graf FN(get_hash_recalc), \ 4775ebb676daSThomas Graf FN(get_current_task), \ 4776ebb676daSThomas Graf FN(probe_write_user), \ 4777ebb676daSThomas Graf FN(current_task_under_cgroup), \ 4778ebb676daSThomas Graf FN(skb_change_tail), \ 4779ebb676daSThomas Graf FN(skb_pull_data), \ 4780ebb676daSThomas Graf FN(csum_update), \ 4781ebb676daSThomas Graf FN(set_hash_invalid), \ 47823a0af8fdSThomas Graf FN(get_numa_node_id), \ 478317bedab2SMartin KaFai Lau FN(skb_change_head), \ 4784a5e8c070SGianluca Borello FN(xdp_adjust_head), \ 478591b8270fSChenbo Feng FN(probe_read_str), \ 47866acc5c29SChenbo Feng FN(get_socket_cookie), \ 4787ded092cdSDaniel Borkmann FN(get_socket_uid), \ 47888c4b4c7eSLawrence Brakmo FN(set_hash), \ 47892be7e212SDaniel Borkmann FN(setsockopt), \ 479097f91a7cSJohn Fastabend FN(skb_adjust_room), \ 4791174a79ffSJohn Fastabend FN(redirect_map), \ 4792174a79ffSJohn Fastabend FN(sk_redirect_map), \ 4793174a79ffSJohn Fastabend FN(sock_map_update), \ 4794908432caSYonghong Song FN(xdp_adjust_meta), \ 47954bebdc7aSYonghong Song FN(perf_event_read_value), \ 4796cd86d1fdSLawrence Brakmo FN(perf_prog_read_value), \ 47979802d865SJosef Bacik FN(getsockopt), \ 4798b13d8807SLawrence Brakmo FN(override_return), \ 47994f738adbSJohn Fastabend FN(sock_ops_cb_flags_set), \ 48002a100317SJohn Fastabend FN(msg_redirect_map), \ 480191843d54SJohn Fastabend FN(msg_apply_bytes), \ 4802015632bbSJohn Fastabend FN(msg_cork_bytes), \ 4803d74bad4eSAndrey Ignatov FN(msg_pull_data), \ 4804b32cc5b9SNikita V. Shirokov FN(bind), \ 480512bed760SEyal Birger FN(xdp_adjust_tail), \ 4806c195651eSYonghong Song FN(skb_get_xfrm_state), \ 48074e1ec56cSDaniel Borkmann FN(get_stack), \ 480887f5fc7eSDavid Ahern FN(skb_load_bytes_relative), \ 480981110384SJohn Fastabend FN(fib_lookup), \ 481081110384SJohn Fastabend FN(sock_hash_update), \ 481181110384SJohn Fastabend FN(msg_redirect_hash), \ 4812fe94cc29SMathieu Xhonneux FN(sk_redirect_hash), \ 4813fe94cc29SMathieu Xhonneux FN(lwt_push_encap), \ 4814fe94cc29SMathieu Xhonneux FN(lwt_seg6_store_bytes), \ 4815fe94cc29SMathieu Xhonneux FN(lwt_seg6_adjust_srh), \ 4816f4364dcfSSean Young FN(lwt_seg6_action), \ 4817f4364dcfSSean Young FN(rc_repeat), \ 4818cb20b08eSDaniel Borkmann FN(rc_keydown), \ 4819bf6fa2c8SYonghong Song FN(skb_cgroup_id), \ 4820cd339431SRoman Gushchin FN(get_current_cgroup_id), \ 48212dbb9b9eSMartin KaFai Lau FN(get_local_storage), \ 482277236281SAndrey Ignatov FN(sk_select_reuseport), \ 48236acc9b43SJoe Stringer FN(skb_ancestor_cgroup_id), \ 48246acc9b43SJoe Stringer FN(sk_lookup_tcp), \ 48256acc9b43SJoe Stringer FN(sk_lookup_udp), \ 4826f1a2e44aSMauricio Vasquez B FN(sk_release), \ 4827f1a2e44aSMauricio Vasquez B FN(map_push_elem), \ 4828f1a2e44aSMauricio Vasquez B FN(map_pop_elem), \ 48296fff607eSJohn Fastabend FN(map_peek_elem), \ 48307246d8edSJohn Fastabend FN(msg_push_data), \ 483101d3240aSSean Young FN(msg_pop_data), \ 4832d83525caSAlexei Starovoitov FN(rc_pointer_rel), \ 4833d83525caSAlexei Starovoitov FN(spin_lock), \ 483446f8bc92SMartin KaFai Lau FN(spin_unlock), \ 4835655a51e5SMartin KaFai Lau FN(sk_fullsock), \ 4836f7c917baSbrakmo FN(tcp_sock), \ 4837dbafd7ddSMartin KaFai Lau FN(skb_ecn_set_ce), \ 4838edbf8c01SLorenz Bauer FN(get_listener_sock), \ 483939904084SLorenz Bauer FN(skc_lookup_tcp), \ 4840808649fbSAndrey Ignatov FN(tcp_check_syncookie), \ 48411d11b301SAndrey Ignatov FN(sysctl_get_name), \ 48424e63acdfSAndrey Ignatov FN(sysctl_get_current_value), \ 48434e63acdfSAndrey Ignatov FN(sysctl_get_new_value), \ 4844d7a4cb9bSAndrey Ignatov FN(sysctl_set_new_value), \ 4845d7a4cb9bSAndrey Ignatov FN(strtol), \ 48466ac99e8fSMartin KaFai Lau FN(strtoul), \ 48476ac99e8fSMartin KaFai Lau FN(sk_storage_get), \ 48488b401f9eSYonghong Song FN(sk_storage_delete), \ 484970d66244SPetar Penkov FN(send_signal), \ 4850a7658e1aSAlexei Starovoitov FN(tcp_gen_syncookie), \ 48516ae08ae3SDaniel Borkmann FN(skb_output), \ 48526ae08ae3SDaniel Borkmann FN(probe_read_user), \ 48536ae08ae3SDaniel Borkmann FN(probe_read_kernel), \ 48546ae08ae3SDaniel Borkmann FN(probe_read_user_str), \ 4855206057feSMartin KaFai Lau FN(probe_read_kernel_str), \ 48568482941fSYonghong Song FN(tcp_send_ack), \ 48575576b991SMartin KaFai Lau FN(send_signal_thread), \ 4858fff7b643SDaniel Xu FN(jiffies64), \ 4859b4490c5cSCarlos Neira FN(read_branch_records), \ 4860d831ee84SEelco Chaudron FN(get_ns_current_pid_tgid), \ 4861f318903cSDaniel Borkmann FN(xdp_output), \ 48620f09abd1SDaniel Borkmann FN(get_netns_cookie), \ 4863cf7fbe66SJoe Stringer FN(get_current_ancestor_cgroup_id), \ 486471d19214SMaciej Żenczykowski FN(sk_assign), \ 4865492e639fSYonghong Song FN(ktime_get_boot_ns), \ 4866492e639fSYonghong Song FN(seq_printf), \ 4867f307fa2cSAndrey Ignatov FN(seq_write), \ 4868f307fa2cSAndrey Ignatov FN(sk_cgroup_id), \ 4869457f4436SAndrii Nakryiko FN(sk_ancestor_cgroup_id), \ 4870457f4436SAndrii Nakryiko FN(ringbuf_output), \ 4871457f4436SAndrii Nakryiko FN(ringbuf_reserve), \ 4872457f4436SAndrii Nakryiko FN(ringbuf_submit), \ 4873457f4436SAndrii Nakryiko FN(ringbuf_discard), \ 48747cdec54fSDaniel Borkmann FN(ringbuf_query), \ 4875af7ec138SYonghong Song FN(csum_level), \ 4876478cfbdfSYonghong Song FN(skc_to_tcp6_sock), \ 4877478cfbdfSYonghong Song FN(skc_to_tcp_sock), \ 4878478cfbdfSYonghong Song FN(skc_to_tcp_timewait_sock), \ 48790d4fad3eSYonghong Song FN(skc_to_tcp_request_sock), \ 4880fa28dcb8SSong Liu FN(skc_to_udp6_sock), \ 4881fa28dcb8SSong Liu FN(get_task_stack), \ 48820813a841SMartin KaFai Lau FN(load_hdr_opt), \ 48830813a841SMartin KaFai Lau FN(store_hdr_opt), \ 48848ea63684SKP Singh FN(reserve_hdr_opt), \ 48858ea63684SKP Singh FN(inode_storage_get), \ 48868ea63684SKP Singh FN(inode_storage_delete), \ 48876e22ab9dSJiri Olsa FN(d_path), \ 488807be4c4aSAlexei Starovoitov FN(copy_from_user), \ 4889c4d0bfb4SAlan Maguire FN(snprintf_btf), \ 4890eb411377SAlan Maguire FN(seq_printf_btf), \ 4891b426ce83SDaniel Borkmann FN(skb_cgroup_classid), \ 4892b4ab3141SDaniel Borkmann FN(redirect_neigh), \ 4893b7906b70SAndrii Nakryiko FN(per_cpu_ptr), \ 4894b7906b70SAndrii Nakryiko FN(this_cpu_ptr), \ 48959aa1206eSDaniel Borkmann FN(redirect_peer), \ 48964cf1bc1fSKP Singh FN(task_storage_get), \ 48974cf1bc1fSKP Singh FN(task_storage_delete), \ 48983ca1032aSKP Singh FN(get_current_task_btf), \ 48993f6719c7SKP Singh FN(bprm_opts_set), \ 4900d0551261SDmitrii Banshchikov FN(ktime_get_coarse_ns), \ 490127672f0dSKP Singh FN(ima_inode_hash), \ 49024f19cab7SFlorent Revest FN(sock_from_file), \ 490334b2021cSJesper Dangaard Brouer FN(check_mtu), \ 490469c087baSYonghong Song FN(for_each_map_elem), \ 4905*7b15523aSFlorent Revest FN(snprintf), \ 4906fa28dcb8SSong Liu /* */ 49077a4b28c6SDaniel Borkmann 4908ebb676daSThomas Graf /* integer value in 'imm' field of BPF_CALL instruction selects which helper 4909ebb676daSThomas Graf * function eBPF program intends to call 49102d0e30c3SDaniel Borkmann */ 4911ebb676daSThomas Graf #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 4912ebb676daSThomas Graf enum bpf_func_id { 4913ebb676daSThomas Graf __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 491409756af4SAlexei Starovoitov __BPF_FUNC_MAX_ID, 491509756af4SAlexei Starovoitov }; 4916ebb676daSThomas Graf #undef __BPF_ENUM_FN 491709756af4SAlexei Starovoitov 4918781c53bcSDaniel Borkmann /* All flags used by eBPF helper functions, placed here. */ 4919781c53bcSDaniel Borkmann 4920781c53bcSDaniel Borkmann /* BPF_FUNC_skb_store_bytes flags. */ 49211aae4bddSAndrii Nakryiko enum { 49221aae4bddSAndrii Nakryiko BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 49231aae4bddSAndrii Nakryiko BPF_F_INVALIDATE_HASH = (1ULL << 1), 49241aae4bddSAndrii Nakryiko }; 4925781c53bcSDaniel Borkmann 4926781c53bcSDaniel Borkmann /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 4927781c53bcSDaniel Borkmann * First 4 bits are for passing the header field size. 4928781c53bcSDaniel Borkmann */ 49291aae4bddSAndrii Nakryiko enum { 49301aae4bddSAndrii Nakryiko BPF_F_HDR_FIELD_MASK = 0xfULL, 49311aae4bddSAndrii Nakryiko }; 4932781c53bcSDaniel Borkmann 4933781c53bcSDaniel Borkmann /* BPF_FUNC_l4_csum_replace flags. */ 49341aae4bddSAndrii Nakryiko enum { 49351aae4bddSAndrii Nakryiko BPF_F_PSEUDO_HDR = (1ULL << 4), 49361aae4bddSAndrii Nakryiko BPF_F_MARK_MANGLED_0 = (1ULL << 5), 49371aae4bddSAndrii Nakryiko BPF_F_MARK_ENFORCE = (1ULL << 6), 49381aae4bddSAndrii Nakryiko }; 4939781c53bcSDaniel Borkmann 4940781c53bcSDaniel Borkmann /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 49411aae4bddSAndrii Nakryiko enum { 49421aae4bddSAndrii Nakryiko BPF_F_INGRESS = (1ULL << 0), 49431aae4bddSAndrii Nakryiko }; 4944781c53bcSDaniel Borkmann 4945c6c33454SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 49461aae4bddSAndrii Nakryiko enum { 49471aae4bddSAndrii Nakryiko BPF_F_TUNINFO_IPV6 = (1ULL << 0), 49481aae4bddSAndrii Nakryiko }; 4949c6c33454SDaniel Borkmann 4950c195651eSYonghong Song /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 49511aae4bddSAndrii Nakryiko enum { 49521aae4bddSAndrii Nakryiko BPF_F_SKIP_FIELD_MASK = 0xffULL, 49531aae4bddSAndrii Nakryiko BPF_F_USER_STACK = (1ULL << 8), 4954c195651eSYonghong Song /* flags used by BPF_FUNC_get_stackid only. */ 49551aae4bddSAndrii Nakryiko BPF_F_FAST_STACK_CMP = (1ULL << 9), 49561aae4bddSAndrii Nakryiko BPF_F_REUSE_STACKID = (1ULL << 10), 4957c195651eSYonghong Song /* flags used by BPF_FUNC_get_stack only. */ 49581aae4bddSAndrii Nakryiko BPF_F_USER_BUILD_ID = (1ULL << 11), 49591aae4bddSAndrii Nakryiko }; 4960d5a3b1f6SAlexei Starovoitov 49612da897e5SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key flags. */ 49621aae4bddSAndrii Nakryiko enum { 49631aae4bddSAndrii Nakryiko BPF_F_ZERO_CSUM_TX = (1ULL << 1), 49641aae4bddSAndrii Nakryiko BPF_F_DONT_FRAGMENT = (1ULL << 2), 49651aae4bddSAndrii Nakryiko BPF_F_SEQ_NUMBER = (1ULL << 3), 49661aae4bddSAndrii Nakryiko }; 49672da897e5SDaniel Borkmann 4968908432caSYonghong Song /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 4969908432caSYonghong Song * BPF_FUNC_perf_event_read_value flags. 4970908432caSYonghong Song */ 49711aae4bddSAndrii Nakryiko enum { 49721aae4bddSAndrii Nakryiko BPF_F_INDEX_MASK = 0xffffffffULL, 49731aae4bddSAndrii Nakryiko BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 4974555c8a86SDaniel Borkmann /* BPF_FUNC_perf_event_output for sk_buff input context. */ 49751aae4bddSAndrii Nakryiko BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 49761aae4bddSAndrii Nakryiko }; 49771e33759cSDaniel Borkmann 4978f71c6143SJoe Stringer /* Current network namespace */ 49791aae4bddSAndrii Nakryiko enum { 49801aae4bddSAndrii Nakryiko BPF_F_CURRENT_NETNS = (-1L), 49811aae4bddSAndrii Nakryiko }; 4982f71c6143SJoe Stringer 49837cdec54fSDaniel Borkmann /* BPF_FUNC_csum_level level values. */ 49847cdec54fSDaniel Borkmann enum { 49857cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_QUERY, 49867cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_INC, 49877cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_DEC, 49887cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_RESET, 49897cdec54fSDaniel Borkmann }; 49907cdec54fSDaniel Borkmann 49912278f6ccSWillem de Bruijn /* BPF_FUNC_skb_adjust_room flags. */ 49921aae4bddSAndrii Nakryiko enum { 49931aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 49941aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 49951aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 49961aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 49971aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 4998836e66c2SDaniel Borkmann BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 4999d01b59c9SXuesen Huang BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 50001aae4bddSAndrii Nakryiko }; 50012278f6ccSWillem de Bruijn 50021aae4bddSAndrii Nakryiko enum { 50031aae4bddSAndrii Nakryiko BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 50041aae4bddSAndrii Nakryiko BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 50051aae4bddSAndrii Nakryiko }; 500658dfc900SAlan Maguire 500758dfc900SAlan Maguire #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 500858dfc900SAlan Maguire BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 500958dfc900SAlan Maguire << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 5010868d5235SWillem de Bruijn 5011808649fbSAndrey Ignatov /* BPF_FUNC_sysctl_get_name flags. */ 50121aae4bddSAndrii Nakryiko enum { 50131aae4bddSAndrii Nakryiko BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 50141aae4bddSAndrii Nakryiko }; 5015808649fbSAndrey Ignatov 5016f836a56eSKP Singh /* BPF_FUNC_<kernel_obj>_storage_get flags */ 50171aae4bddSAndrii Nakryiko enum { 5018f836a56eSKP Singh BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 5019f836a56eSKP Singh /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 5020f836a56eSKP Singh * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 5021f836a56eSKP Singh */ 5022f836a56eSKP Singh BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 50231aae4bddSAndrii Nakryiko }; 50246ac99e8fSMartin KaFai Lau 5025fff7b643SDaniel Xu /* BPF_FUNC_read_branch_records flags. */ 50261aae4bddSAndrii Nakryiko enum { 50271aae4bddSAndrii Nakryiko BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 50281aae4bddSAndrii Nakryiko }; 5029fff7b643SDaniel Xu 5030457f4436SAndrii Nakryiko /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 5031457f4436SAndrii Nakryiko * BPF_FUNC_bpf_ringbuf_output flags. 5032457f4436SAndrii Nakryiko */ 5033457f4436SAndrii Nakryiko enum { 5034457f4436SAndrii Nakryiko BPF_RB_NO_WAKEUP = (1ULL << 0), 5035457f4436SAndrii Nakryiko BPF_RB_FORCE_WAKEUP = (1ULL << 1), 5036457f4436SAndrii Nakryiko }; 5037457f4436SAndrii Nakryiko 5038457f4436SAndrii Nakryiko /* BPF_FUNC_bpf_ringbuf_query flags */ 5039457f4436SAndrii Nakryiko enum { 5040457f4436SAndrii Nakryiko BPF_RB_AVAIL_DATA = 0, 5041457f4436SAndrii Nakryiko BPF_RB_RING_SIZE = 1, 5042457f4436SAndrii Nakryiko BPF_RB_CONS_POS = 2, 5043457f4436SAndrii Nakryiko BPF_RB_PROD_POS = 3, 5044457f4436SAndrii Nakryiko }; 5045457f4436SAndrii Nakryiko 5046457f4436SAndrii Nakryiko /* BPF ring buffer constants */ 5047457f4436SAndrii Nakryiko enum { 5048457f4436SAndrii Nakryiko BPF_RINGBUF_BUSY_BIT = (1U << 31), 5049457f4436SAndrii Nakryiko BPF_RINGBUF_DISCARD_BIT = (1U << 30), 5050457f4436SAndrii Nakryiko BPF_RINGBUF_HDR_SZ = 8, 5051457f4436SAndrii Nakryiko }; 5052457f4436SAndrii Nakryiko 5053e9ddbb77SJakub Sitnicki /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 5054e9ddbb77SJakub Sitnicki enum { 5055e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 5056e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 5057e9ddbb77SJakub Sitnicki }; 5058e9ddbb77SJakub Sitnicki 50592be7e212SDaniel Borkmann /* Mode for BPF_FUNC_skb_adjust_room helper. */ 50602be7e212SDaniel Borkmann enum bpf_adj_room_mode { 50612be7e212SDaniel Borkmann BPF_ADJ_ROOM_NET, 506214aa3192SWillem de Bruijn BPF_ADJ_ROOM_MAC, 50632be7e212SDaniel Borkmann }; 50642be7e212SDaniel Borkmann 50654e1ec56cSDaniel Borkmann /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 50664e1ec56cSDaniel Borkmann enum bpf_hdr_start_off { 50674e1ec56cSDaniel Borkmann BPF_HDR_START_MAC, 50684e1ec56cSDaniel Borkmann BPF_HDR_START_NET, 50694e1ec56cSDaniel Borkmann }; 50704e1ec56cSDaniel Borkmann 5071fe94cc29SMathieu Xhonneux /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 5072fe94cc29SMathieu Xhonneux enum bpf_lwt_encap_mode { 5073fe94cc29SMathieu Xhonneux BPF_LWT_ENCAP_SEG6, 50743e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_SEG6_INLINE, 50753e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_IP, 5076fe94cc29SMathieu Xhonneux }; 5077fe94cc29SMathieu Xhonneux 50783f6719c7SKP Singh /* Flags for bpf_bprm_opts_set helper */ 50793f6719c7SKP Singh enum { 50803f6719c7SKP Singh BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 50813f6719c7SKP Singh }; 50823f6719c7SKP Singh 5083b7df9adaSDaniel Borkmann #define __bpf_md_ptr(type, name) \ 5084b7df9adaSDaniel Borkmann union { \ 5085b7df9adaSDaniel Borkmann type name; \ 5086b7df9adaSDaniel Borkmann __u64 :64; \ 5087b7df9adaSDaniel Borkmann } __attribute__((aligned(8))) 5088b7df9adaSDaniel Borkmann 50899bac3d6dSAlexei Starovoitov /* user accessible mirror of in-kernel sk_buff. 50909bac3d6dSAlexei Starovoitov * new fields can only be added to the end of this structure 50919bac3d6dSAlexei Starovoitov */ 50929bac3d6dSAlexei Starovoitov struct __sk_buff { 50939bac3d6dSAlexei Starovoitov __u32 len; 50949bac3d6dSAlexei Starovoitov __u32 pkt_type; 50959bac3d6dSAlexei Starovoitov __u32 mark; 50969bac3d6dSAlexei Starovoitov __u32 queue_mapping; 5097c2497395SAlexei Starovoitov __u32 protocol; 5098c2497395SAlexei Starovoitov __u32 vlan_present; 5099c2497395SAlexei Starovoitov __u32 vlan_tci; 510027cd5452SMichal Sekletar __u32 vlan_proto; 5101bcad5718SDaniel Borkmann __u32 priority; 510237e82c2fSAlexei Starovoitov __u32 ingress_ifindex; 510337e82c2fSAlexei Starovoitov __u32 ifindex; 5104d691f9e8SAlexei Starovoitov __u32 tc_index; 5105d691f9e8SAlexei Starovoitov __u32 cb[5]; 5106ba7591d8SDaniel Borkmann __u32 hash; 5107045efa82SDaniel Borkmann __u32 tc_classid; 5108969bf05eSAlexei Starovoitov __u32 data; 5109969bf05eSAlexei Starovoitov __u32 data_end; 5110b1d9fc41SDaniel Borkmann __u32 napi_id; 51118a31db56SJohn Fastabend 5112de8f3a83SDaniel Borkmann /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 51138a31db56SJohn Fastabend __u32 family; 51148a31db56SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 51158a31db56SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 51168a31db56SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 51178a31db56SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 51188a31db56SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 51198a31db56SJohn Fastabend __u32 local_port; /* stored in host byte order */ 5120de8f3a83SDaniel Borkmann /* ... here. */ 5121de8f3a83SDaniel Borkmann 5122de8f3a83SDaniel Borkmann __u32 data_meta; 5123b7df9adaSDaniel Borkmann __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 5124f11216b2SVlad Dumitrescu __u64 tstamp; 5125e3da08d0SPetar Penkov __u32 wire_len; 5126d9ff286aSEric Dumazet __u32 gso_segs; 512746f8bc92SMartin KaFai Lau __bpf_md_ptr(struct bpf_sock *, sk); 5128cf62089bSWillem de Bruijn __u32 gso_size; 51299bac3d6dSAlexei Starovoitov }; 51309bac3d6dSAlexei Starovoitov 5131d3aa45ceSAlexei Starovoitov struct bpf_tunnel_key { 5132d3aa45ceSAlexei Starovoitov __u32 tunnel_id; 5133c6c33454SDaniel Borkmann union { 5134d3aa45ceSAlexei Starovoitov __u32 remote_ipv4; 5135c6c33454SDaniel Borkmann __u32 remote_ipv6[4]; 5136c6c33454SDaniel Borkmann }; 5137c6c33454SDaniel Borkmann __u8 tunnel_tos; 5138c6c33454SDaniel Borkmann __u8 tunnel_ttl; 51391fbc2e0cSDaniel Borkmann __u16 tunnel_ext; /* Padding, future use. */ 51404018ab18SDaniel Borkmann __u32 tunnel_label; 5141d3aa45ceSAlexei Starovoitov }; 5142d3aa45ceSAlexei Starovoitov 514312bed760SEyal Birger /* user accessible mirror of in-kernel xfrm_state. 514412bed760SEyal Birger * new fields can only be added to the end of this structure 514512bed760SEyal Birger */ 514612bed760SEyal Birger struct bpf_xfrm_state { 514712bed760SEyal Birger __u32 reqid; 514812bed760SEyal Birger __u32 spi; /* Stored in network byte order */ 514912bed760SEyal Birger __u16 family; 51501fbc2e0cSDaniel Borkmann __u16 ext; /* Padding, future use. */ 515112bed760SEyal Birger union { 515212bed760SEyal Birger __u32 remote_ipv4; /* Stored in network byte order */ 515312bed760SEyal Birger __u32 remote_ipv6[4]; /* Stored in network byte order */ 515412bed760SEyal Birger }; 515512bed760SEyal Birger }; 515612bed760SEyal Birger 51573a0af8fdSThomas Graf /* Generic BPF return codes which all BPF program types may support. 51583a0af8fdSThomas Graf * The values are binary compatible with their TC_ACT_* counter-part to 51593a0af8fdSThomas Graf * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 51603a0af8fdSThomas Graf * programs. 51613a0af8fdSThomas Graf * 51623a0af8fdSThomas Graf * XDP is handled seprately, see XDP_*. 51633a0af8fdSThomas Graf */ 51643a0af8fdSThomas Graf enum bpf_ret_code { 51653a0af8fdSThomas Graf BPF_OK = 0, 51663a0af8fdSThomas Graf /* 1 reserved */ 51673a0af8fdSThomas Graf BPF_DROP = 2, 51683a0af8fdSThomas Graf /* 3-6 reserved */ 51693a0af8fdSThomas Graf BPF_REDIRECT = 7, 51703e0bd37cSPeter Oskolkov /* >127 are reserved for prog type specific return codes. 51713e0bd37cSPeter Oskolkov * 51723e0bd37cSPeter Oskolkov * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 51733e0bd37cSPeter Oskolkov * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 51743e0bd37cSPeter Oskolkov * changed and should be routed based on its new L3 header. 51753e0bd37cSPeter Oskolkov * (This is an L3 redirect, as opposed to L2 redirect 51763e0bd37cSPeter Oskolkov * represented by BPF_REDIRECT above). 51773e0bd37cSPeter Oskolkov */ 51783e0bd37cSPeter Oskolkov BPF_LWT_REROUTE = 128, 51793a0af8fdSThomas Graf }; 51803a0af8fdSThomas Graf 518161023658SDavid Ahern struct bpf_sock { 518261023658SDavid Ahern __u32 bound_dev_if; 5183aa4c1037SDavid Ahern __u32 family; 5184aa4c1037SDavid Ahern __u32 type; 5185aa4c1037SDavid Ahern __u32 protocol; 5186482dca93SDavid Ahern __u32 mark; 5187482dca93SDavid Ahern __u32 priority; 5188aa65d696SMartin KaFai Lau /* IP address also allows 1 and 2 bytes access */ 5189aa65d696SMartin KaFai Lau __u32 src_ip4; 5190aa65d696SMartin KaFai Lau __u32 src_ip6[4]; 5191aa65d696SMartin KaFai Lau __u32 src_port; /* host byte order */ 5192aa65d696SMartin KaFai Lau __u32 dst_port; /* network byte order */ 5193aa65d696SMartin KaFai Lau __u32 dst_ip4; 5194aa65d696SMartin KaFai Lau __u32 dst_ip6[4]; 5195aa65d696SMartin KaFai Lau __u32 state; 5196c3c16f2eSAmritha Nambiar __s32 rx_queue_mapping; 519761023658SDavid Ahern }; 519861023658SDavid Ahern 5199655a51e5SMartin KaFai Lau struct bpf_tcp_sock { 5200655a51e5SMartin KaFai Lau __u32 snd_cwnd; /* Sending congestion window */ 5201655a51e5SMartin KaFai Lau __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 5202655a51e5SMartin KaFai Lau __u32 rtt_min; 5203655a51e5SMartin KaFai Lau __u32 snd_ssthresh; /* Slow start size threshold */ 5204655a51e5SMartin KaFai Lau __u32 rcv_nxt; /* What we want to receive next */ 5205655a51e5SMartin KaFai Lau __u32 snd_nxt; /* Next sequence we send */ 5206655a51e5SMartin KaFai Lau __u32 snd_una; /* First byte we want an ack for */ 5207655a51e5SMartin KaFai Lau __u32 mss_cache; /* Cached effective mss, not including SACKS */ 5208655a51e5SMartin KaFai Lau __u32 ecn_flags; /* ECN status bits. */ 5209655a51e5SMartin KaFai Lau __u32 rate_delivered; /* saved rate sample: packets delivered */ 5210655a51e5SMartin KaFai Lau __u32 rate_interval_us; /* saved rate sample: time elapsed */ 5211655a51e5SMartin KaFai Lau __u32 packets_out; /* Packets which are "in flight" */ 5212655a51e5SMartin KaFai Lau __u32 retrans_out; /* Retransmitted packets out */ 5213655a51e5SMartin KaFai Lau __u32 total_retrans; /* Total retransmits for entire connection */ 5214655a51e5SMartin KaFai Lau __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 5215655a51e5SMartin KaFai Lau * total number of segments in. 5216655a51e5SMartin KaFai Lau */ 5217655a51e5SMartin KaFai Lau __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 5218655a51e5SMartin KaFai Lau * total number of data segments in. 5219655a51e5SMartin KaFai Lau */ 5220655a51e5SMartin KaFai Lau __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 5221655a51e5SMartin KaFai Lau * The total number of segments sent. 5222655a51e5SMartin KaFai Lau */ 5223655a51e5SMartin KaFai Lau __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 5224655a51e5SMartin KaFai Lau * total number of data segments sent. 5225655a51e5SMartin KaFai Lau */ 5226655a51e5SMartin KaFai Lau __u32 lost_out; /* Lost packets */ 5227655a51e5SMartin KaFai Lau __u32 sacked_out; /* SACK'd packets */ 5228655a51e5SMartin KaFai Lau __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 5229655a51e5SMartin KaFai Lau * sum(delta(rcv_nxt)), or how many bytes 5230655a51e5SMartin KaFai Lau * were acked. 5231655a51e5SMartin KaFai Lau */ 5232655a51e5SMartin KaFai Lau __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 5233655a51e5SMartin KaFai Lau * sum(delta(snd_una)), or how many bytes 5234655a51e5SMartin KaFai Lau * were acked. 5235655a51e5SMartin KaFai Lau */ 52360357746dSStanislav Fomichev __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 52370357746dSStanislav Fomichev * total number of DSACK blocks received 52380357746dSStanislav Fomichev */ 52390357746dSStanislav Fomichev __u32 delivered; /* Total data packets delivered incl. rexmits */ 52400357746dSStanislav Fomichev __u32 delivered_ce; /* Like the above but only ECE marked packets */ 5241c2cb5e82SStanislav Fomichev __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 5242655a51e5SMartin KaFai Lau }; 5243655a51e5SMartin KaFai Lau 52446acc9b43SJoe Stringer struct bpf_sock_tuple { 52456acc9b43SJoe Stringer union { 52466acc9b43SJoe Stringer struct { 52476acc9b43SJoe Stringer __be32 saddr; 52486acc9b43SJoe Stringer __be32 daddr; 52496acc9b43SJoe Stringer __be16 sport; 52506acc9b43SJoe Stringer __be16 dport; 52516acc9b43SJoe Stringer } ipv4; 52526acc9b43SJoe Stringer struct { 52536acc9b43SJoe Stringer __be32 saddr[4]; 52546acc9b43SJoe Stringer __be32 daddr[4]; 52556acc9b43SJoe Stringer __be16 sport; 52566acc9b43SJoe Stringer __be16 dport; 52576acc9b43SJoe Stringer } ipv6; 52586acc9b43SJoe Stringer }; 52596acc9b43SJoe Stringer }; 52606acc9b43SJoe Stringer 5261fada7fdcSJonathan Lemon struct bpf_xdp_sock { 5262fada7fdcSJonathan Lemon __u32 queue_id; 5263fada7fdcSJonathan Lemon }; 5264fada7fdcSJonathan Lemon 526517bedab2SMartin KaFai Lau #define XDP_PACKET_HEADROOM 256 526617bedab2SMartin KaFai Lau 52676a773a15SBrenden Blanco /* User return codes for XDP prog type. 52686a773a15SBrenden Blanco * A valid XDP program must return one of these defined values. All other 52699beb8bedSDaniel Borkmann * return codes are reserved for future use. Unknown return codes will 52709beb8bedSDaniel Borkmann * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 52716a773a15SBrenden Blanco */ 52726a773a15SBrenden Blanco enum xdp_action { 52736a773a15SBrenden Blanco XDP_ABORTED = 0, 52746a773a15SBrenden Blanco XDP_DROP, 52756a773a15SBrenden Blanco XDP_PASS, 52766ce96ca3SBrenden Blanco XDP_TX, 5277814abfabSJohn Fastabend XDP_REDIRECT, 52786a773a15SBrenden Blanco }; 52796a773a15SBrenden Blanco 52806a773a15SBrenden Blanco /* user accessible metadata for XDP packet hook 52816a773a15SBrenden Blanco * new fields must be added to the end of this structure 52826a773a15SBrenden Blanco */ 52836a773a15SBrenden Blanco struct xdp_md { 52846a773a15SBrenden Blanco __u32 data; 52856a773a15SBrenden Blanco __u32 data_end; 5286de8f3a83SDaniel Borkmann __u32 data_meta; 5287daaf24c6SJesper Dangaard Brouer /* Below access go through struct xdp_rxq_info */ 528802dd3291SJesper Dangaard Brouer __u32 ingress_ifindex; /* rxq->dev->ifindex */ 528902dd3291SJesper Dangaard Brouer __u32 rx_queue_index; /* rxq->queue_index */ 529064b59025SDavid Ahern 529164b59025SDavid Ahern __u32 egress_ifindex; /* txq->dev->ifindex */ 52926a773a15SBrenden Blanco }; 52936a773a15SBrenden Blanco 5294281920b7SJesper Dangaard Brouer /* DEVMAP map-value layout 5295281920b7SJesper Dangaard Brouer * 5296281920b7SJesper Dangaard Brouer * The struct data-layout of map-value is a configuration interface. 5297281920b7SJesper Dangaard Brouer * New members can only be added to the end of this structure. 5298281920b7SJesper Dangaard Brouer */ 5299281920b7SJesper Dangaard Brouer struct bpf_devmap_val { 5300281920b7SJesper Dangaard Brouer __u32 ifindex; /* device index */ 5301281920b7SJesper Dangaard Brouer union { 5302281920b7SJesper Dangaard Brouer int fd; /* prog fd on map write */ 5303281920b7SJesper Dangaard Brouer __u32 id; /* prog id on map read */ 5304281920b7SJesper Dangaard Brouer } bpf_prog; 5305281920b7SJesper Dangaard Brouer }; 5306281920b7SJesper Dangaard Brouer 5307644bfe51SLorenzo Bianconi /* CPUMAP map-value layout 5308644bfe51SLorenzo Bianconi * 5309644bfe51SLorenzo Bianconi * The struct data-layout of map-value is a configuration interface. 5310644bfe51SLorenzo Bianconi * New members can only be added to the end of this structure. 5311644bfe51SLorenzo Bianconi */ 5312644bfe51SLorenzo Bianconi struct bpf_cpumap_val { 5313644bfe51SLorenzo Bianconi __u32 qsize; /* queue size to remote target CPU */ 531492164774SLorenzo Bianconi union { 531592164774SLorenzo Bianconi int fd; /* prog fd on map write */ 531692164774SLorenzo Bianconi __u32 id; /* prog id on map read */ 531792164774SLorenzo Bianconi } bpf_prog; 5318644bfe51SLorenzo Bianconi }; 5319644bfe51SLorenzo Bianconi 5320174a79ffSJohn Fastabend enum sk_action { 5321bfa64075SJohn Fastabend SK_DROP = 0, 5322bfa64075SJohn Fastabend SK_PASS, 5323174a79ffSJohn Fastabend }; 5324174a79ffSJohn Fastabend 53254f738adbSJohn Fastabend /* user accessible metadata for SK_MSG packet hook, new fields must 53264f738adbSJohn Fastabend * be added to the end of this structure 53274f738adbSJohn Fastabend */ 53284f738adbSJohn Fastabend struct sk_msg_md { 5329b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 5330b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 5331303def35SJohn Fastabend 5332303def35SJohn Fastabend __u32 family; 5333303def35SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 5334303def35SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 5335303def35SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 5336303def35SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 5337303def35SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 5338303def35SJohn Fastabend __u32 local_port; /* stored in host byte order */ 53393bdbd022SJohn Fastabend __u32 size; /* Total size of sk_msg */ 534013d70f5aSJohn Fastabend 534113d70f5aSJohn Fastabend __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 53424f738adbSJohn Fastabend }; 53434f738adbSJohn Fastabend 53442dbb9b9eSMartin KaFai Lau struct sk_reuseport_md { 53452dbb9b9eSMartin KaFai Lau /* 53462dbb9b9eSMartin KaFai Lau * Start of directly accessible data. It begins from 53472dbb9b9eSMartin KaFai Lau * the tcp/udp header. 53482dbb9b9eSMartin KaFai Lau */ 5349b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 5350b7df9adaSDaniel Borkmann /* End of directly accessible data */ 5351b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 53522dbb9b9eSMartin KaFai Lau /* 53532dbb9b9eSMartin KaFai Lau * Total length of packet (starting from the tcp/udp header). 53542dbb9b9eSMartin KaFai Lau * Note that the directly accessible bytes (data_end - data) 53552dbb9b9eSMartin KaFai Lau * could be less than this "len". Those bytes could be 53562dbb9b9eSMartin KaFai Lau * indirectly read by a helper "bpf_skb_load_bytes()". 53572dbb9b9eSMartin KaFai Lau */ 53582dbb9b9eSMartin KaFai Lau __u32 len; 53592dbb9b9eSMartin KaFai Lau /* 53602dbb9b9eSMartin KaFai Lau * Eth protocol in the mac header (network byte order). e.g. 53612dbb9b9eSMartin KaFai Lau * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 53622dbb9b9eSMartin KaFai Lau */ 53632dbb9b9eSMartin KaFai Lau __u32 eth_protocol; 53642dbb9b9eSMartin KaFai Lau __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 53652dbb9b9eSMartin KaFai Lau __u32 bind_inany; /* Is sock bound to an INANY address? */ 53662dbb9b9eSMartin KaFai Lau __u32 hash; /* A hash of the packet 4 tuples */ 53672dbb9b9eSMartin KaFai Lau }; 53682dbb9b9eSMartin KaFai Lau 53691e270976SMartin KaFai Lau #define BPF_TAG_SIZE 8 53701e270976SMartin KaFai Lau 53711e270976SMartin KaFai Lau struct bpf_prog_info { 53721e270976SMartin KaFai Lau __u32 type; 53731e270976SMartin KaFai Lau __u32 id; 53741e270976SMartin KaFai Lau __u8 tag[BPF_TAG_SIZE]; 53751e270976SMartin KaFai Lau __u32 jited_prog_len; 53761e270976SMartin KaFai Lau __u32 xlated_prog_len; 53771e270976SMartin KaFai Lau __aligned_u64 jited_prog_insns; 53781e270976SMartin KaFai Lau __aligned_u64 xlated_prog_insns; 5379cb4d2b3fSMartin KaFai Lau __u64 load_time; /* ns since boottime */ 5380cb4d2b3fSMartin KaFai Lau __u32 created_by_uid; 5381cb4d2b3fSMartin KaFai Lau __u32 nr_map_ids; 5382cb4d2b3fSMartin KaFai Lau __aligned_u64 map_ids; 5383067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 5384675fc275SJakub Kicinski __u32 ifindex; 5385b85fab0eSJiri Olsa __u32 gpl_compatible:1; 53860472301aSBaruch Siach __u32 :31; /* alignment pad */ 5387675fc275SJakub Kicinski __u64 netns_dev; 5388675fc275SJakub Kicinski __u64 netns_ino; 5389dbecd738SSandipan Das __u32 nr_jited_ksyms; 5390815581c1SSandipan Das __u32 nr_jited_func_lens; 5391dbecd738SSandipan Das __aligned_u64 jited_ksyms; 5392815581c1SSandipan Das __aligned_u64 jited_func_lens; 5393838e9690SYonghong Song __u32 btf_id; 5394838e9690SYonghong Song __u32 func_info_rec_size; 5395838e9690SYonghong Song __aligned_u64 func_info; 539611d8b82dSYonghong Song __u32 nr_func_info; 539711d8b82dSYonghong Song __u32 nr_line_info; 5398c454a46bSMartin KaFai Lau __aligned_u64 line_info; 5399c454a46bSMartin KaFai Lau __aligned_u64 jited_line_info; 540011d8b82dSYonghong Song __u32 nr_jited_line_info; 5401c454a46bSMartin KaFai Lau __u32 line_info_rec_size; 5402c454a46bSMartin KaFai Lau __u32 jited_line_info_rec_size; 5403c872bdb3SSong Liu __u32 nr_prog_tags; 5404c872bdb3SSong Liu __aligned_u64 prog_tags; 54055f8f8b93SAlexei Starovoitov __u64 run_time_ns; 54065f8f8b93SAlexei Starovoitov __u64 run_cnt; 54079ed9e9baSAlexei Starovoitov __u64 recursion_misses; 54081e270976SMartin KaFai Lau } __attribute__((aligned(8))); 54091e270976SMartin KaFai Lau 54101e270976SMartin KaFai Lau struct bpf_map_info { 54111e270976SMartin KaFai Lau __u32 type; 54121e270976SMartin KaFai Lau __u32 id; 54131e270976SMartin KaFai Lau __u32 key_size; 54141e270976SMartin KaFai Lau __u32 value_size; 54151e270976SMartin KaFai Lau __u32 max_entries; 54161e270976SMartin KaFai Lau __u32 map_flags; 5417067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 541852775b33SJakub Kicinski __u32 ifindex; 541985d33df3SMartin KaFai Lau __u32 btf_vmlinux_value_type_id; 542052775b33SJakub Kicinski __u64 netns_dev; 542152775b33SJakub Kicinski __u64 netns_ino; 542278958fcaSMartin KaFai Lau __u32 btf_id; 54239b2cf328SMartin KaFai Lau __u32 btf_key_type_id; 54249b2cf328SMartin KaFai Lau __u32 btf_value_type_id; 54251e270976SMartin KaFai Lau } __attribute__((aligned(8))); 54261e270976SMartin KaFai Lau 542762dab84cSMartin KaFai Lau struct bpf_btf_info { 542862dab84cSMartin KaFai Lau __aligned_u64 btf; 542962dab84cSMartin KaFai Lau __u32 btf_size; 543062dab84cSMartin KaFai Lau __u32 id; 543153297220SAndrii Nakryiko __aligned_u64 name; 543253297220SAndrii Nakryiko __u32 name_len; 543353297220SAndrii Nakryiko __u32 kernel_btf; 543462dab84cSMartin KaFai Lau } __attribute__((aligned(8))); 543562dab84cSMartin KaFai Lau 5436f2e10bffSAndrii Nakryiko struct bpf_link_info { 5437f2e10bffSAndrii Nakryiko __u32 type; 5438f2e10bffSAndrii Nakryiko __u32 id; 5439f2e10bffSAndrii Nakryiko __u32 prog_id; 5440f2e10bffSAndrii Nakryiko union { 5441f2e10bffSAndrii Nakryiko struct { 5442f2e10bffSAndrii Nakryiko __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 5443f2e10bffSAndrii Nakryiko __u32 tp_name_len; /* in/out: tp_name buffer len */ 5444f2e10bffSAndrii Nakryiko } raw_tracepoint; 5445f2e10bffSAndrii Nakryiko struct { 5446f2e10bffSAndrii Nakryiko __u32 attach_type; 5447441e8c66SToke Høiland-Jørgensen __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 5448441e8c66SToke Høiland-Jørgensen __u32 target_btf_id; /* BTF type id inside the object */ 5449f2e10bffSAndrii Nakryiko } tracing; 5450f2e10bffSAndrii Nakryiko struct { 5451f2e10bffSAndrii Nakryiko __u64 cgroup_id; 5452f2e10bffSAndrii Nakryiko __u32 attach_type; 5453f2e10bffSAndrii Nakryiko } cgroup; 54547f045a49SJakub Sitnicki struct { 54556b0a249aSYonghong Song __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 54566b0a249aSYonghong Song __u32 target_name_len; /* in/out: target_name buffer len */ 54576b0a249aSYonghong Song union { 5458b0c9eb37SYonghong Song struct { 54596b0a249aSYonghong Song __u32 map_id; 54606b0a249aSYonghong Song } map; 5461b0c9eb37SYonghong Song }; 54626b0a249aSYonghong Song } iter; 54636b0a249aSYonghong Song struct { 54647f045a49SJakub Sitnicki __u32 netns_ino; 54657f045a49SJakub Sitnicki __u32 attach_type; 54667f045a49SJakub Sitnicki } netns; 5467c1931c97SAndrii Nakryiko struct { 5468c1931c97SAndrii Nakryiko __u32 ifindex; 5469c1931c97SAndrii Nakryiko } xdp; 5470f2e10bffSAndrii Nakryiko }; 5471f2e10bffSAndrii Nakryiko } __attribute__((aligned(8))); 5472f2e10bffSAndrii Nakryiko 54734fbac77dSAndrey Ignatov /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 54744fbac77dSAndrey Ignatov * by user and intended to be used by socket (e.g. to bind to, depends on 5475bfdfa517SRandy Dunlap * attach type). 54764fbac77dSAndrey Ignatov */ 54774fbac77dSAndrey Ignatov struct bpf_sock_addr { 54784fbac77dSAndrey Ignatov __u32 user_family; /* Allows 4-byte read, but no write. */ 54794fbac77dSAndrey Ignatov __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 54804fbac77dSAndrey Ignatov * Stored in network byte order. 54814fbac77dSAndrey Ignatov */ 5482d4ecfeb1SStanislav Fomichev __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 54834fbac77dSAndrey Ignatov * Stored in network byte order. 54844fbac77dSAndrey Ignatov */ 54857aebfa1bSAndrey Ignatov __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 54864fbac77dSAndrey Ignatov * Stored in network byte order 54874fbac77dSAndrey Ignatov */ 54884fbac77dSAndrey Ignatov __u32 family; /* Allows 4-byte read, but no write */ 54894fbac77dSAndrey Ignatov __u32 type; /* Allows 4-byte read, but no write */ 54904fbac77dSAndrey Ignatov __u32 protocol; /* Allows 4-byte read, but no write */ 5491600c70baSStanislav Fomichev __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 54921cedee13SAndrey Ignatov * Stored in network byte order. 54931cedee13SAndrey Ignatov */ 5494d4ecfeb1SStanislav Fomichev __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 54951cedee13SAndrey Ignatov * Stored in network byte order. 54961cedee13SAndrey Ignatov */ 5497fb85c4a7SStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 54984fbac77dSAndrey Ignatov }; 54994fbac77dSAndrey Ignatov 550040304b2aSLawrence Brakmo /* User bpf_sock_ops struct to access socket values and specify request ops 550140304b2aSLawrence Brakmo * and their replies. 550240304b2aSLawrence Brakmo * Some of this fields are in network (bigendian) byte order and may need 550340304b2aSLawrence Brakmo * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 550440304b2aSLawrence Brakmo * New fields can only be added at the end of this structure 550540304b2aSLawrence Brakmo */ 550640304b2aSLawrence Brakmo struct bpf_sock_ops { 550740304b2aSLawrence Brakmo __u32 op; 550840304b2aSLawrence Brakmo union { 5509de525be2SLawrence Brakmo __u32 args[4]; /* Optionally passed to bpf program */ 5510de525be2SLawrence Brakmo __u32 reply; /* Returned by bpf program */ 5511de525be2SLawrence Brakmo __u32 replylong[4]; /* Optionally returned by bpf prog */ 551240304b2aSLawrence Brakmo }; 551340304b2aSLawrence Brakmo __u32 family; 551440304b2aSLawrence Brakmo __u32 remote_ip4; /* Stored in network byte order */ 551540304b2aSLawrence Brakmo __u32 local_ip4; /* Stored in network byte order */ 551640304b2aSLawrence Brakmo __u32 remote_ip6[4]; /* Stored in network byte order */ 551740304b2aSLawrence Brakmo __u32 local_ip6[4]; /* Stored in network byte order */ 551840304b2aSLawrence Brakmo __u32 remote_port; /* Stored in network byte order */ 551940304b2aSLawrence Brakmo __u32 local_port; /* stored in host byte order */ 5520f19397a5SLawrence Brakmo __u32 is_fullsock; /* Some TCP fields are only valid if 5521f19397a5SLawrence Brakmo * there is a full socket. If not, the 5522f19397a5SLawrence Brakmo * fields read as zero. 5523f19397a5SLawrence Brakmo */ 5524f19397a5SLawrence Brakmo __u32 snd_cwnd; 5525f19397a5SLawrence Brakmo __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 5526b13d8807SLawrence Brakmo __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 552744f0e430SLawrence Brakmo __u32 state; 552844f0e430SLawrence Brakmo __u32 rtt_min; 552944f0e430SLawrence Brakmo __u32 snd_ssthresh; 553044f0e430SLawrence Brakmo __u32 rcv_nxt; 553144f0e430SLawrence Brakmo __u32 snd_nxt; 553244f0e430SLawrence Brakmo __u32 snd_una; 553344f0e430SLawrence Brakmo __u32 mss_cache; 553444f0e430SLawrence Brakmo __u32 ecn_flags; 553544f0e430SLawrence Brakmo __u32 rate_delivered; 553644f0e430SLawrence Brakmo __u32 rate_interval_us; 553744f0e430SLawrence Brakmo __u32 packets_out; 553844f0e430SLawrence Brakmo __u32 retrans_out; 553944f0e430SLawrence Brakmo __u32 total_retrans; 554044f0e430SLawrence Brakmo __u32 segs_in; 554144f0e430SLawrence Brakmo __u32 data_segs_in; 554244f0e430SLawrence Brakmo __u32 segs_out; 554344f0e430SLawrence Brakmo __u32 data_segs_out; 554444f0e430SLawrence Brakmo __u32 lost_out; 554544f0e430SLawrence Brakmo __u32 sacked_out; 554644f0e430SLawrence Brakmo __u32 sk_txhash; 554744f0e430SLawrence Brakmo __u64 bytes_received; 554844f0e430SLawrence Brakmo __u64 bytes_acked; 55491314ef56SStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 55500813a841SMartin KaFai Lau /* [skb_data, skb_data_end) covers the whole TCP header. 55510813a841SMartin KaFai Lau * 55520813a841SMartin KaFai Lau * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 55530813a841SMartin KaFai Lau * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 55540813a841SMartin KaFai Lau * header has not been written. 55550813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 55560813a841SMartin KaFai Lau * been written so far. 55570813a841SMartin KaFai Lau * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 55580813a841SMartin KaFai Lau * the 3WHS. 55590813a841SMartin KaFai Lau * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 55600813a841SMartin KaFai Lau * the 3WHS. 55610813a841SMartin KaFai Lau * 55620813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also be used to read a particular option. 55630813a841SMartin KaFai Lau */ 55640813a841SMartin KaFai Lau __bpf_md_ptr(void *, skb_data); 55650813a841SMartin KaFai Lau __bpf_md_ptr(void *, skb_data_end); 55660813a841SMartin KaFai Lau __u32 skb_len; /* The total length of a packet. 55670813a841SMartin KaFai Lau * It includes the header, options, 55680813a841SMartin KaFai Lau * and payload. 55690813a841SMartin KaFai Lau */ 55700813a841SMartin KaFai Lau __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 55710813a841SMartin KaFai Lau * an easy way to check for tcp_flags 55720813a841SMartin KaFai Lau * without parsing skb_data. 55730813a841SMartin KaFai Lau * 55740813a841SMartin KaFai Lau * In particular, the skb_tcp_flags 55750813a841SMartin KaFai Lau * will still be available in 55760813a841SMartin KaFai Lau * BPF_SOCK_OPS_HDR_OPT_LEN even though 55770813a841SMartin KaFai Lau * the outgoing header has not 55780813a841SMartin KaFai Lau * been written yet. 55790813a841SMartin KaFai Lau */ 558040304b2aSLawrence Brakmo }; 558140304b2aSLawrence Brakmo 5582b13d8807SLawrence Brakmo /* Definitions for bpf_sock_ops_cb_flags */ 55831aae4bddSAndrii Nakryiko enum { 55841aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 55851aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 55861aae4bddSAndrii Nakryiko BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 55871aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 55880813a841SMartin KaFai Lau /* Call bpf for all received TCP headers. The bpf prog will be 55890813a841SMartin KaFai Lau * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 55900813a841SMartin KaFai Lau * 55910813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 55920813a841SMartin KaFai Lau * for the header option related helpers that will be useful 55930813a841SMartin KaFai Lau * to the bpf programs. 55940813a841SMartin KaFai Lau * 55950813a841SMartin KaFai Lau * It could be used at the client/active side (i.e. connect() side) 55960813a841SMartin KaFai Lau * when the server told it that the server was in syncookie 55970813a841SMartin KaFai Lau * mode and required the active side to resend the bpf-written 55980813a841SMartin KaFai Lau * options. The active side can keep writing the bpf-options until 55990813a841SMartin KaFai Lau * it received a valid packet from the server side to confirm 56000813a841SMartin KaFai Lau * the earlier packet (and options) has been received. The later 56010813a841SMartin KaFai Lau * example patch is using it like this at the active side when the 56020813a841SMartin KaFai Lau * server is in syncookie mode. 56030813a841SMartin KaFai Lau * 56040813a841SMartin KaFai Lau * The bpf prog will usually turn this off in the common cases. 56050813a841SMartin KaFai Lau */ 560600d211a4SMartin KaFai Lau BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 56070813a841SMartin KaFai Lau /* Call bpf when kernel has received a header option that 56080813a841SMartin KaFai Lau * the kernel cannot handle. The bpf prog will be called under 56090813a841SMartin KaFai Lau * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 56100813a841SMartin KaFai Lau * 56110813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 56120813a841SMartin KaFai Lau * for the header option related helpers that will be useful 56130813a841SMartin KaFai Lau * to the bpf programs. 56140813a841SMartin KaFai Lau */ 561500d211a4SMartin KaFai Lau BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 56160813a841SMartin KaFai Lau /* Call bpf when the kernel is writing header options for the 56170813a841SMartin KaFai Lau * outgoing packet. The bpf prog will first be called 56180813a841SMartin KaFai Lau * to reserve space in a skb under 56190813a841SMartin KaFai Lau * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 56200813a841SMartin KaFai Lau * the bpf prog will be called to write the header option(s) 56210813a841SMartin KaFai Lau * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 56220813a841SMartin KaFai Lau * 56230813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 56240813a841SMartin KaFai Lau * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 56250813a841SMartin KaFai Lau * related helpers that will be useful to the bpf programs. 56260813a841SMartin KaFai Lau * 56270813a841SMartin KaFai Lau * The kernel gets its chance to reserve space and write 56280813a841SMartin KaFai Lau * options first before the BPF program does. 56290813a841SMartin KaFai Lau */ 5630331fca43SMartin KaFai Lau BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 56311aae4bddSAndrii Nakryiko /* Mask of all currently supported cb flags */ 5632331fca43SMartin KaFai Lau BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 56331aae4bddSAndrii Nakryiko }; 5634b13d8807SLawrence Brakmo 563540304b2aSLawrence Brakmo /* List of known BPF sock_ops operators. 563640304b2aSLawrence Brakmo * New entries can only be added at the end 563740304b2aSLawrence Brakmo */ 563840304b2aSLawrence Brakmo enum { 563940304b2aSLawrence Brakmo BPF_SOCK_OPS_VOID, 56408550f328SLawrence Brakmo BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 56418550f328SLawrence Brakmo * -1 if default value should be used 56428550f328SLawrence Brakmo */ 564313d3b1ebSLawrence Brakmo BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 564413d3b1ebSLawrence Brakmo * window (in packets) or -1 if default 564513d3b1ebSLawrence Brakmo * value should be used 564613d3b1ebSLawrence Brakmo */ 56479872a4bdSLawrence Brakmo BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 56489872a4bdSLawrence Brakmo * active connection is initialized 56499872a4bdSLawrence Brakmo */ 56509872a4bdSLawrence Brakmo BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 56519872a4bdSLawrence Brakmo * active connection is 56529872a4bdSLawrence Brakmo * established 56539872a4bdSLawrence Brakmo */ 56549872a4bdSLawrence Brakmo BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 56559872a4bdSLawrence Brakmo * passive connection is 56569872a4bdSLawrence Brakmo * established 56579872a4bdSLawrence Brakmo */ 565891b5b21cSLawrence Brakmo BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 565991b5b21cSLawrence Brakmo * needs ECN 566091b5b21cSLawrence Brakmo */ 5661e6546ef6SLawrence Brakmo BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 5662e6546ef6SLawrence Brakmo * based on the path and may be 5663e6546ef6SLawrence Brakmo * dependent on the congestion control 5664e6546ef6SLawrence Brakmo * algorithm. In general it indicates 5665e6546ef6SLawrence Brakmo * a congestion threshold. RTTs above 5666e6546ef6SLawrence Brakmo * this indicate congestion 5667e6546ef6SLawrence Brakmo */ 5668f89013f6SLawrence Brakmo BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 5669f89013f6SLawrence Brakmo * Arg1: value of icsk_retransmits 5670f89013f6SLawrence Brakmo * Arg2: value of icsk_rto 5671f89013f6SLawrence Brakmo * Arg3: whether RTO has expired 5672f89013f6SLawrence Brakmo */ 5673a31ad29eSLawrence Brakmo BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 5674a31ad29eSLawrence Brakmo * Arg1: sequence number of 1st byte 5675a31ad29eSLawrence Brakmo * Arg2: # segments 5676a31ad29eSLawrence Brakmo * Arg3: return value of 5677a31ad29eSLawrence Brakmo * tcp_transmit_skb (0 => success) 5678a31ad29eSLawrence Brakmo */ 5679d4487491SLawrence Brakmo BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 5680d4487491SLawrence Brakmo * Arg1: old_state 5681d4487491SLawrence Brakmo * Arg2: new_state 5682d4487491SLawrence Brakmo */ 5683f333ee0cSAndrey Ignatov BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 5684f333ee0cSAndrey Ignatov * socket transition to LISTEN state. 5685f333ee0cSAndrey Ignatov */ 568623729ff2SStanislav Fomichev BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 568723729ff2SStanislav Fomichev */ 56880813a841SMartin KaFai Lau BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 56890813a841SMartin KaFai Lau * It will be called to handle 56900813a841SMartin KaFai Lau * the packets received at 56910813a841SMartin KaFai Lau * an already established 56920813a841SMartin KaFai Lau * connection. 56930813a841SMartin KaFai Lau * 56940813a841SMartin KaFai Lau * sock_ops->skb_data: 56950813a841SMartin KaFai Lau * Referring to the received skb. 56960813a841SMartin KaFai Lau * It covers the TCP header only. 56970813a841SMartin KaFai Lau * 56980813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also 56990813a841SMartin KaFai Lau * be used to search for a 57000813a841SMartin KaFai Lau * particular option. 57010813a841SMartin KaFai Lau */ 57020813a841SMartin KaFai Lau BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 57030813a841SMartin KaFai Lau * header option later in 57040813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 57050813a841SMartin KaFai Lau * Arg1: bool want_cookie. (in 57060813a841SMartin KaFai Lau * writing SYNACK only) 57070813a841SMartin KaFai Lau * 57080813a841SMartin KaFai Lau * sock_ops->skb_data: 57090813a841SMartin KaFai Lau * Not available because no header has 57100813a841SMartin KaFai Lau * been written yet. 57110813a841SMartin KaFai Lau * 57120813a841SMartin KaFai Lau * sock_ops->skb_tcp_flags: 57130813a841SMartin KaFai Lau * The tcp_flags of the 57140813a841SMartin KaFai Lau * outgoing skb. (e.g. SYN, ACK, FIN). 57150813a841SMartin KaFai Lau * 57160813a841SMartin KaFai Lau * bpf_reserve_hdr_opt() should 57170813a841SMartin KaFai Lau * be used to reserve space. 57180813a841SMartin KaFai Lau */ 57190813a841SMartin KaFai Lau BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 57200813a841SMartin KaFai Lau * Arg1: bool want_cookie. (in 57210813a841SMartin KaFai Lau * writing SYNACK only) 57220813a841SMartin KaFai Lau * 57230813a841SMartin KaFai Lau * sock_ops->skb_data: 57240813a841SMartin KaFai Lau * Referring to the outgoing skb. 57250813a841SMartin KaFai Lau * It covers the TCP header 57260813a841SMartin KaFai Lau * that has already been written 57270813a841SMartin KaFai Lau * by the kernel and the 57280813a841SMartin KaFai Lau * earlier bpf-progs. 57290813a841SMartin KaFai Lau * 57300813a841SMartin KaFai Lau * sock_ops->skb_tcp_flags: 57310813a841SMartin KaFai Lau * The tcp_flags of the outgoing 57320813a841SMartin KaFai Lau * skb. (e.g. SYN, ACK, FIN). 57330813a841SMartin KaFai Lau * 57340813a841SMartin KaFai Lau * bpf_store_hdr_opt() should 57350813a841SMartin KaFai Lau * be used to write the 57360813a841SMartin KaFai Lau * option. 57370813a841SMartin KaFai Lau * 57380813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also 57390813a841SMartin KaFai Lau * be used to search for a 57400813a841SMartin KaFai Lau * particular option that 57410813a841SMartin KaFai Lau * has already been written 57420813a841SMartin KaFai Lau * by the kernel or the 57430813a841SMartin KaFai Lau * earlier bpf-progs. 57440813a841SMartin KaFai Lau */ 5745d4487491SLawrence Brakmo }; 5746d4487491SLawrence Brakmo 5747d4487491SLawrence Brakmo /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 5748d4487491SLawrence Brakmo * changes between the TCP and BPF versions. Ideally this should never happen. 5749d4487491SLawrence Brakmo * If it does, we need to add code to convert them before calling 5750d4487491SLawrence Brakmo * the BPF sock_ops function. 5751d4487491SLawrence Brakmo */ 5752d4487491SLawrence Brakmo enum { 5753d4487491SLawrence Brakmo BPF_TCP_ESTABLISHED = 1, 5754d4487491SLawrence Brakmo BPF_TCP_SYN_SENT, 5755d4487491SLawrence Brakmo BPF_TCP_SYN_RECV, 5756d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT1, 5757d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT2, 5758d4487491SLawrence Brakmo BPF_TCP_TIME_WAIT, 5759d4487491SLawrence Brakmo BPF_TCP_CLOSE, 5760d4487491SLawrence Brakmo BPF_TCP_CLOSE_WAIT, 5761d4487491SLawrence Brakmo BPF_TCP_LAST_ACK, 5762d4487491SLawrence Brakmo BPF_TCP_LISTEN, 5763d4487491SLawrence Brakmo BPF_TCP_CLOSING, /* Now a valid state */ 5764d4487491SLawrence Brakmo BPF_TCP_NEW_SYN_RECV, 5765d4487491SLawrence Brakmo 5766d4487491SLawrence Brakmo BPF_TCP_MAX_STATES /* Leave at the end! */ 576740304b2aSLawrence Brakmo }; 576840304b2aSLawrence Brakmo 57691aae4bddSAndrii Nakryiko enum { 57701aae4bddSAndrii Nakryiko TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 57711aae4bddSAndrii Nakryiko TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 57722b8ee4f0SMartin KaFai Lau TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 5773ca584ba0SMartin KaFai Lau TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 57740813a841SMartin KaFai Lau /* Copy the SYN pkt to optval 57750813a841SMartin KaFai Lau * 57760813a841SMartin KaFai Lau * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 57770813a841SMartin KaFai Lau * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 57780813a841SMartin KaFai Lau * to only getting from the saved_syn. It can either get the 57790813a841SMartin KaFai Lau * syn packet from: 57800813a841SMartin KaFai Lau * 57810813a841SMartin KaFai Lau * 1. the just-received SYN packet (only available when writing the 57820813a841SMartin KaFai Lau * SYNACK). It will be useful when it is not necessary to 57830813a841SMartin KaFai Lau * save the SYN packet for latter use. It is also the only way 57840813a841SMartin KaFai Lau * to get the SYN during syncookie mode because the syn 57850813a841SMartin KaFai Lau * packet cannot be saved during syncookie. 57860813a841SMartin KaFai Lau * 57870813a841SMartin KaFai Lau * OR 57880813a841SMartin KaFai Lau * 57890813a841SMartin KaFai Lau * 2. the earlier saved syn which was done by 57900813a841SMartin KaFai Lau * bpf_setsockopt(TCP_SAVE_SYN). 57910813a841SMartin KaFai Lau * 57920813a841SMartin KaFai Lau * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 57930813a841SMartin KaFai Lau * SYN packet is obtained. 57940813a841SMartin KaFai Lau * 57950813a841SMartin KaFai Lau * If the bpf-prog does not need the IP[46] header, the 57960813a841SMartin KaFai Lau * bpf-prog can avoid parsing the IP header by using 57970813a841SMartin KaFai Lau * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 57980813a841SMartin KaFai Lau * IP[46] and TCP header by using TCP_BPF_SYN_IP. 57990813a841SMartin KaFai Lau * 58000813a841SMartin KaFai Lau * >0: Total number of bytes copied 58010813a841SMartin KaFai Lau * -ENOSPC: Not enough space in optval. Only optlen number of 58020813a841SMartin KaFai Lau * bytes is copied. 58030813a841SMartin KaFai Lau * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 58040813a841SMartin KaFai Lau * is not saved by setsockopt(TCP_SAVE_SYN). 58050813a841SMartin KaFai Lau */ 58060813a841SMartin KaFai Lau TCP_BPF_SYN = 1005, /* Copy the TCP header */ 58070813a841SMartin KaFai Lau TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 5808267cf9faSMartin KaFai Lau TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 58090813a841SMartin KaFai Lau }; 58100813a841SMartin KaFai Lau 58110813a841SMartin KaFai Lau enum { 58120813a841SMartin KaFai Lau BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 58130813a841SMartin KaFai Lau }; 58140813a841SMartin KaFai Lau 58150813a841SMartin KaFai Lau /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 58160813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 58170813a841SMartin KaFai Lau */ 58180813a841SMartin KaFai Lau enum { 58190813a841SMartin KaFai Lau BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 58200813a841SMartin KaFai Lau * total option spaces 58210813a841SMartin KaFai Lau * required for an established 58220813a841SMartin KaFai Lau * sk in order to calculate the 58230813a841SMartin KaFai Lau * MSS. No skb is actually 58240813a841SMartin KaFai Lau * sent. 58250813a841SMartin KaFai Lau */ 58260813a841SMartin KaFai Lau BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 58270813a841SMartin KaFai Lau * when sending a SYN. 58280813a841SMartin KaFai Lau */ 58291aae4bddSAndrii Nakryiko }; 5830fc747810SLawrence Brakmo 5831908432caSYonghong Song struct bpf_perf_event_value { 5832908432caSYonghong Song __u64 counter; 5833908432caSYonghong Song __u64 enabled; 5834908432caSYonghong Song __u64 running; 5835908432caSYonghong Song }; 5836908432caSYonghong Song 58371aae4bddSAndrii Nakryiko enum { 58381aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 58391aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_READ = (1ULL << 1), 58401aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_WRITE = (1ULL << 2), 58411aae4bddSAndrii Nakryiko }; 5842ebc614f6SRoman Gushchin 58431aae4bddSAndrii Nakryiko enum { 58441aae4bddSAndrii Nakryiko BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 58451aae4bddSAndrii Nakryiko BPF_DEVCG_DEV_CHAR = (1ULL << 1), 58461aae4bddSAndrii Nakryiko }; 5847ebc614f6SRoman Gushchin 5848ebc614f6SRoman Gushchin struct bpf_cgroup_dev_ctx { 584906ef0ccbSYonghong Song /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 585006ef0ccbSYonghong Song __u32 access_type; 5851ebc614f6SRoman Gushchin __u32 major; 5852ebc614f6SRoman Gushchin __u32 minor; 5853ebc614f6SRoman Gushchin }; 5854ebc614f6SRoman Gushchin 5855c4f6699dSAlexei Starovoitov struct bpf_raw_tracepoint_args { 5856c4f6699dSAlexei Starovoitov __u64 args[0]; 5857c4f6699dSAlexei Starovoitov }; 5858c4f6699dSAlexei Starovoitov 585987f5fc7eSDavid Ahern /* DIRECT: Skip the FIB rules and go to FIB table associated with device 586087f5fc7eSDavid Ahern * OUTPUT: Do lookup from egress perspective; default is ingress 586187f5fc7eSDavid Ahern */ 58621aae4bddSAndrii Nakryiko enum { 58631aae4bddSAndrii Nakryiko BPF_FIB_LOOKUP_DIRECT = (1U << 0), 58641aae4bddSAndrii Nakryiko BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 58651aae4bddSAndrii Nakryiko }; 586687f5fc7eSDavid Ahern 58674c79579bSDavid Ahern enum { 58684c79579bSDavid Ahern BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 58694c79579bSDavid Ahern BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 58704c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 58714c79579bSDavid Ahern BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 58724c79579bSDavid Ahern BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 58734c79579bSDavid Ahern BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 58744c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 58754c79579bSDavid Ahern BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 58764c79579bSDavid Ahern BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 58774c79579bSDavid Ahern }; 58784c79579bSDavid Ahern 587987f5fc7eSDavid Ahern struct bpf_fib_lookup { 5880fa898d76SDavid Ahern /* input: network family for lookup (AF_INET, AF_INET6) 5881fa898d76SDavid Ahern * output: network family of egress nexthop 5882fa898d76SDavid Ahern */ 5883fa898d76SDavid Ahern __u8 family; 588487f5fc7eSDavid Ahern 588587f5fc7eSDavid Ahern /* set if lookup is to consider L4 data - e.g., FIB rules */ 588687f5fc7eSDavid Ahern __u8 l4_protocol; 588787f5fc7eSDavid Ahern __be16 sport; 588887f5fc7eSDavid Ahern __be16 dport; 588987f5fc7eSDavid Ahern 5890e1850ea9SJesper Dangaard Brouer union { /* used for MTU check */ 5891e1850ea9SJesper Dangaard Brouer /* input to lookup */ 5892e1850ea9SJesper Dangaard Brouer __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 58934c79579bSDavid Ahern 5894e1850ea9SJesper Dangaard Brouer /* output: MTU value */ 5895e1850ea9SJesper Dangaard Brouer __u16 mtu_result; 5896e1850ea9SJesper Dangaard Brouer }; 58974c79579bSDavid Ahern /* input: L3 device index for lookup 58984c79579bSDavid Ahern * output: device index from FIB lookup 58994c79579bSDavid Ahern */ 59004c79579bSDavid Ahern __u32 ifindex; 590187f5fc7eSDavid Ahern 590287f5fc7eSDavid Ahern union { 590387f5fc7eSDavid Ahern /* inputs to lookup */ 590487f5fc7eSDavid Ahern __u8 tos; /* AF_INET */ 5905bd3a08aaSDavid Ahern __be32 flowinfo; /* AF_INET6, flow_label + priority */ 590687f5fc7eSDavid Ahern 5907fa898d76SDavid Ahern /* output: metric of fib result (IPv4/IPv6 only) */ 590887f5fc7eSDavid Ahern __u32 rt_metric; 590987f5fc7eSDavid Ahern }; 591087f5fc7eSDavid Ahern 591187f5fc7eSDavid Ahern union { 591287f5fc7eSDavid Ahern __be32 ipv4_src; 591387f5fc7eSDavid Ahern __u32 ipv6_src[4]; /* in6_addr; network order */ 591487f5fc7eSDavid Ahern }; 591587f5fc7eSDavid Ahern 5916fa898d76SDavid Ahern /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 5917fa898d76SDavid Ahern * network header. output: bpf_fib_lookup sets to gateway address 5918fa898d76SDavid Ahern * if FIB lookup returns gateway route 591987f5fc7eSDavid Ahern */ 592087f5fc7eSDavid Ahern union { 592187f5fc7eSDavid Ahern __be32 ipv4_dst; 592287f5fc7eSDavid Ahern __u32 ipv6_dst[4]; /* in6_addr; network order */ 592387f5fc7eSDavid Ahern }; 592487f5fc7eSDavid Ahern 592587f5fc7eSDavid Ahern /* output */ 592687f5fc7eSDavid Ahern __be16 h_vlan_proto; 592787f5fc7eSDavid Ahern __be16 h_vlan_TCI; 592887f5fc7eSDavid Ahern __u8 smac[6]; /* ETH_ALEN */ 592987f5fc7eSDavid Ahern __u8 dmac[6]; /* ETH_ALEN */ 593087f5fc7eSDavid Ahern }; 593187f5fc7eSDavid Ahern 5932ba452c9eSToke Høiland-Jørgensen struct bpf_redir_neigh { 5933ba452c9eSToke Høiland-Jørgensen /* network family for lookup (AF_INET, AF_INET6) */ 5934ba452c9eSToke Høiland-Jørgensen __u32 nh_family; 5935ba452c9eSToke Høiland-Jørgensen /* network address of nexthop; skips fib lookup to find gateway */ 5936ba452c9eSToke Høiland-Jørgensen union { 5937ba452c9eSToke Høiland-Jørgensen __be32 ipv4_nh; 5938ba452c9eSToke Høiland-Jørgensen __u32 ipv6_nh[4]; /* in6_addr; network order */ 5939ba452c9eSToke Høiland-Jørgensen }; 5940ba452c9eSToke Høiland-Jørgensen }; 5941ba452c9eSToke Høiland-Jørgensen 594234b2021cSJesper Dangaard Brouer /* bpf_check_mtu flags*/ 594334b2021cSJesper Dangaard Brouer enum bpf_check_mtu_flags { 594434b2021cSJesper Dangaard Brouer BPF_MTU_CHK_SEGS = (1U << 0), 594534b2021cSJesper Dangaard Brouer }; 594634b2021cSJesper Dangaard Brouer 594734b2021cSJesper Dangaard Brouer enum bpf_check_mtu_ret { 594834b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 594934b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 595034b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 595134b2021cSJesper Dangaard Brouer }; 595234b2021cSJesper Dangaard Brouer 595341bdc4b4SYonghong Song enum bpf_task_fd_type { 595441bdc4b4SYonghong Song BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 595541bdc4b4SYonghong Song BPF_FD_TYPE_TRACEPOINT, /* tp name */ 595641bdc4b4SYonghong Song BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 595741bdc4b4SYonghong Song BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 595841bdc4b4SYonghong Song BPF_FD_TYPE_UPROBE, /* filename + offset */ 595941bdc4b4SYonghong Song BPF_FD_TYPE_URETPROBE, /* filename + offset */ 596041bdc4b4SYonghong Song }; 596141bdc4b4SYonghong Song 59621aae4bddSAndrii Nakryiko enum { 59631aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 59641aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 59651aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 59661aae4bddSAndrii Nakryiko }; 5967086f9568SStanislav Fomichev 5968d58e468bSPetar Penkov struct bpf_flow_keys { 5969d58e468bSPetar Penkov __u16 nhoff; 5970d58e468bSPetar Penkov __u16 thoff; 5971d58e468bSPetar Penkov __u16 addr_proto; /* ETH_P_* of valid addrs */ 5972d58e468bSPetar Penkov __u8 is_frag; 5973d58e468bSPetar Penkov __u8 is_first_frag; 5974d58e468bSPetar Penkov __u8 is_encap; 5975d58e468bSPetar Penkov __u8 ip_proto; 5976d58e468bSPetar Penkov __be16 n_proto; 5977d58e468bSPetar Penkov __be16 sport; 5978d58e468bSPetar Penkov __be16 dport; 5979d58e468bSPetar Penkov union { 5980d58e468bSPetar Penkov struct { 5981d58e468bSPetar Penkov __be32 ipv4_src; 5982d58e468bSPetar Penkov __be32 ipv4_dst; 5983d58e468bSPetar Penkov }; 5984d58e468bSPetar Penkov struct { 5985d58e468bSPetar Penkov __u32 ipv6_src[4]; /* in6_addr; network order */ 5986d58e468bSPetar Penkov __u32 ipv6_dst[4]; /* in6_addr; network order */ 5987d58e468bSPetar Penkov }; 5988d58e468bSPetar Penkov }; 5989086f9568SStanislav Fomichev __u32 flags; 599071c99e32SStanislav Fomichev __be32 flow_label; 5991d58e468bSPetar Penkov }; 5992d58e468bSPetar Penkov 5993838e9690SYonghong Song struct bpf_func_info { 5994d30d42e0SMartin KaFai Lau __u32 insn_off; 5995838e9690SYonghong Song __u32 type_id; 5996838e9690SYonghong Song }; 5997838e9690SYonghong Song 5998c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 5999c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 6000c454a46bSMartin KaFai Lau 6001c454a46bSMartin KaFai Lau struct bpf_line_info { 6002c454a46bSMartin KaFai Lau __u32 insn_off; 6003c454a46bSMartin KaFai Lau __u32 file_name_off; 6004c454a46bSMartin KaFai Lau __u32 line_off; 6005c454a46bSMartin KaFai Lau __u32 line_col; 6006c454a46bSMartin KaFai Lau }; 6007c454a46bSMartin KaFai Lau 6008d83525caSAlexei Starovoitov struct bpf_spin_lock { 6009d83525caSAlexei Starovoitov __u32 val; 6010d83525caSAlexei Starovoitov }; 60117b146cebSAndrey Ignatov 60127b146cebSAndrey Ignatov struct bpf_sysctl { 60137b146cebSAndrey Ignatov __u32 write; /* Sysctl is being read (= 0) or written (= 1). 60147b146cebSAndrey Ignatov * Allows 1,2,4-byte read, but no write. 60157b146cebSAndrey Ignatov */ 6016e1550bfeSAndrey Ignatov __u32 file_pos; /* Sysctl file position to read from, write to. 6017e1550bfeSAndrey Ignatov * Allows 1,2,4-byte read an 4-byte write. 6018e1550bfeSAndrey Ignatov */ 60197b146cebSAndrey Ignatov }; 60207b146cebSAndrey Ignatov 60210d01da6aSStanislav Fomichev struct bpf_sockopt { 60220d01da6aSStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 60230d01da6aSStanislav Fomichev __bpf_md_ptr(void *, optval); 60240d01da6aSStanislav Fomichev __bpf_md_ptr(void *, optval_end); 60250d01da6aSStanislav Fomichev 60260d01da6aSStanislav Fomichev __s32 level; 60270d01da6aSStanislav Fomichev __s32 optname; 60280d01da6aSStanislav Fomichev __s32 optlen; 60290d01da6aSStanislav Fomichev __s32 retval; 60300d01da6aSStanislav Fomichev }; 60310d01da6aSStanislav Fomichev 6032b4490c5cSCarlos Neira struct bpf_pidns_info { 6033b4490c5cSCarlos Neira __u32 pid; 6034b4490c5cSCarlos Neira __u32 tgid; 6035b4490c5cSCarlos Neira }; 6036e9ddbb77SJakub Sitnicki 6037e9ddbb77SJakub Sitnicki /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 6038e9ddbb77SJakub Sitnicki struct bpf_sk_lookup { 60397c32e8f8SLorenz Bauer union { 6040e9ddbb77SJakub Sitnicki __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 60417c32e8f8SLorenz Bauer __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 60427c32e8f8SLorenz Bauer }; 6043e9ddbb77SJakub Sitnicki 6044e9ddbb77SJakub Sitnicki __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 6045e9ddbb77SJakub Sitnicki __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 6046e9ddbb77SJakub Sitnicki __u32 remote_ip4; /* Network byte order */ 6047e9ddbb77SJakub Sitnicki __u32 remote_ip6[4]; /* Network byte order */ 6048e9ddbb77SJakub Sitnicki __u32 remote_port; /* Network byte order */ 6049e9ddbb77SJakub Sitnicki __u32 local_ip4; /* Network byte order */ 6050e9ddbb77SJakub Sitnicki __u32 local_ip6[4]; /* Network byte order */ 6051e9ddbb77SJakub Sitnicki __u32 local_port; /* Host byte order */ 6052e9ddbb77SJakub Sitnicki }; 6053e9ddbb77SJakub Sitnicki 6054c4d0bfb4SAlan Maguire /* 6055c4d0bfb4SAlan Maguire * struct btf_ptr is used for typed pointer representation; the 6056c4d0bfb4SAlan Maguire * type id is used to render the pointer data as the appropriate type 6057c4d0bfb4SAlan Maguire * via the bpf_snprintf_btf() helper described above. A flags field - 6058c4d0bfb4SAlan Maguire * potentially to specify additional details about the BTF pointer 6059c4d0bfb4SAlan Maguire * (rather than its mode of display) - is included for future use. 6060c4d0bfb4SAlan Maguire * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 6061c4d0bfb4SAlan Maguire */ 6062c4d0bfb4SAlan Maguire struct btf_ptr { 6063c4d0bfb4SAlan Maguire void *ptr; 6064c4d0bfb4SAlan Maguire __u32 type_id; 6065c4d0bfb4SAlan Maguire __u32 flags; /* BTF ptr flags; unused at present. */ 6066c4d0bfb4SAlan Maguire }; 6067c4d0bfb4SAlan Maguire 6068c4d0bfb4SAlan Maguire /* 6069c4d0bfb4SAlan Maguire * Flags to control bpf_snprintf_btf() behaviour. 6070c4d0bfb4SAlan Maguire * - BTF_F_COMPACT: no formatting around type information 6071c4d0bfb4SAlan Maguire * - BTF_F_NONAME: no struct/union member names/types 6072c4d0bfb4SAlan Maguire * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 6073c4d0bfb4SAlan Maguire * equivalent to %px. 6074c4d0bfb4SAlan Maguire * - BTF_F_ZERO: show zero-valued struct/union members; they 6075c4d0bfb4SAlan Maguire * are not displayed by default 6076c4d0bfb4SAlan Maguire */ 6077c4d0bfb4SAlan Maguire enum { 6078c4d0bfb4SAlan Maguire BTF_F_COMPACT = (1ULL << 0), 6079c4d0bfb4SAlan Maguire BTF_F_NONAME = (1ULL << 1), 6080c4d0bfb4SAlan Maguire BTF_F_PTR_RAW = (1ULL << 2), 6081c4d0bfb4SAlan Maguire BTF_F_ZERO = (1ULL << 3), 6082c4d0bfb4SAlan Maguire }; 6083c4d0bfb4SAlan Maguire 6084daedfb22SAlexei Starovoitov #endif /* _UAPI__LINUX_BPF_H__ */ 6085