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 */ 823024d95aSDaniel Borkmann __u8 data[0]; /* Arbitrary size */ 83b95a5c4dSDaniel Mack }; 84b95a5c4dSDaniel Mack 85de9cbbaaSRoman Gushchin struct bpf_cgroup_storage_key { 86de9cbbaaSRoman Gushchin __u64 cgroup_inode_id; /* cgroup inode id */ 876fc88c35SDave Marchevsky __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 88de9cbbaaSRoman Gushchin }; 89de9cbbaaSRoman Gushchin 90d4ccaf58SHao Luo enum bpf_cgroup_iter_order { 91d4ffb6f3SHao Luo BPF_CGROUP_ITER_ORDER_UNSPEC = 0, 92d4ffb6f3SHao Luo BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */ 93d4ffb6f3SHao Luo BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */ 94d4ffb6f3SHao Luo BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */ 95d4ffb6f3SHao Luo BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */ 96d4ccaf58SHao Luo }; 97d4ccaf58SHao Luo 985e7b3020SYonghong Song union bpf_iter_link_info { 995e7b3020SYonghong Song struct { 1005e7b3020SYonghong Song __u32 map_fd; 1015e7b3020SYonghong Song } map; 102d4ccaf58SHao Luo struct { 103d4ccaf58SHao Luo enum bpf_cgroup_iter_order order; 104d4ccaf58SHao Luo 105d4ccaf58SHao Luo /* At most one of cgroup_fd and cgroup_id can be non-zero. If 106d4ccaf58SHao Luo * both are zero, the walk starts from the default cgroup v2 107d4ccaf58SHao Luo * root. For walking v1 hierarchy, one should always explicitly 108d4ccaf58SHao Luo * specify cgroup_fd. 109d4ccaf58SHao Luo */ 110d4ccaf58SHao Luo __u32 cgroup_fd; 111d4ccaf58SHao Luo __u64 cgroup_id; 112d4ccaf58SHao Luo } cgroup; 113f0d74c4dSKui-Feng Lee /* Parameters of task iterators. */ 114f0d74c4dSKui-Feng Lee struct { 115f0d74c4dSKui-Feng Lee __u32 tid; 116f0d74c4dSKui-Feng Lee __u32 pid; 117f0d74c4dSKui-Feng Lee __u32 pid_fd; 118f0d74c4dSKui-Feng Lee } task; 1195e7b3020SYonghong Song }; 1205e7b3020SYonghong Song 1217799e4d9SJoe Stringer /* BPF syscall commands, see bpf(2) man-page for more details. */ 1227799e4d9SJoe Stringer /** 1237799e4d9SJoe Stringer * DOC: eBPF Syscall Preamble 1247799e4d9SJoe Stringer * 1257799e4d9SJoe Stringer * The operation to be performed by the **bpf**\ () system call is determined 1267799e4d9SJoe Stringer * by the *cmd* argument. Each operation takes an accompanying argument, 1277799e4d9SJoe Stringer * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 1287799e4d9SJoe Stringer * below). The size argument is the size of the union pointed to by *attr*. 1297799e4d9SJoe Stringer */ 1307799e4d9SJoe Stringer /** 1317799e4d9SJoe Stringer * DOC: eBPF Syscall Commands 1327799e4d9SJoe Stringer * 1337799e4d9SJoe Stringer * BPF_MAP_CREATE 1347799e4d9SJoe Stringer * Description 1357799e4d9SJoe Stringer * Create a map and return a file descriptor that refers to the 1367799e4d9SJoe Stringer * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 1377799e4d9SJoe Stringer * is automatically enabled for the new file descriptor. 1387799e4d9SJoe Stringer * 1397799e4d9SJoe Stringer * Applying **close**\ (2) to the file descriptor returned by 1407799e4d9SJoe Stringer * **BPF_MAP_CREATE** will delete the map (but see NOTES). 1417799e4d9SJoe Stringer * 1427799e4d9SJoe Stringer * Return 1437799e4d9SJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 1447799e4d9SJoe Stringer * error occurred (in which case, *errno* is set appropriately). 1457799e4d9SJoe Stringer * 1467799e4d9SJoe Stringer * BPF_MAP_LOOKUP_ELEM 1477799e4d9SJoe Stringer * Description 1487799e4d9SJoe Stringer * Look up an element with a given *key* in the map referred to 1497799e4d9SJoe Stringer * by the file descriptor *map_fd*. 1507799e4d9SJoe Stringer * 1516690523bSJoe Stringer * The *flags* argument may be specified as one of the 1526690523bSJoe Stringer * following: 1536690523bSJoe Stringer * 1546690523bSJoe Stringer * **BPF_F_LOCK** 1556690523bSJoe Stringer * Look up the value of a spin-locked map without 1566690523bSJoe Stringer * returning the lock. This must be specified if the 1576690523bSJoe Stringer * elements contain a spinlock. 1586690523bSJoe Stringer * 1597799e4d9SJoe Stringer * Return 1607799e4d9SJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 1617799e4d9SJoe Stringer * is set appropriately. 1627799e4d9SJoe Stringer * 1637799e4d9SJoe Stringer * BPF_MAP_UPDATE_ELEM 1647799e4d9SJoe Stringer * Description 1657799e4d9SJoe Stringer * Create or update an element (key/value pair) in a specified map. 1667799e4d9SJoe Stringer * 1677799e4d9SJoe Stringer * The *flags* argument should be specified as one of the 1687799e4d9SJoe Stringer * following: 1697799e4d9SJoe Stringer * 1707799e4d9SJoe Stringer * **BPF_ANY** 1717799e4d9SJoe Stringer * Create a new element or update an existing element. 1727799e4d9SJoe Stringer * **BPF_NOEXIST** 1737799e4d9SJoe Stringer * Create a new element only if it did not exist. 1747799e4d9SJoe Stringer * **BPF_EXIST** 1757799e4d9SJoe Stringer * Update an existing element. 1766690523bSJoe Stringer * **BPF_F_LOCK** 1776690523bSJoe Stringer * Update a spin_lock-ed map element. 1787799e4d9SJoe Stringer * 1797799e4d9SJoe Stringer * Return 1807799e4d9SJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 1817799e4d9SJoe Stringer * is set appropriately. 1827799e4d9SJoe Stringer * 1837799e4d9SJoe Stringer * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 1847799e4d9SJoe Stringer * **E2BIG**, **EEXIST**, or **ENOENT**. 1857799e4d9SJoe Stringer * 1867799e4d9SJoe Stringer * **E2BIG** 1877799e4d9SJoe Stringer * The number of elements in the map reached the 1887799e4d9SJoe Stringer * *max_entries* limit specified at map creation time. 1897799e4d9SJoe Stringer * **EEXIST** 1907799e4d9SJoe Stringer * If *flags* specifies **BPF_NOEXIST** and the element 1917799e4d9SJoe Stringer * with *key* already exists in the map. 1927799e4d9SJoe Stringer * **ENOENT** 1937799e4d9SJoe Stringer * If *flags* specifies **BPF_EXIST** and the element with 1947799e4d9SJoe Stringer * *key* does not exist in the map. 1957799e4d9SJoe Stringer * 1967799e4d9SJoe Stringer * BPF_MAP_DELETE_ELEM 1977799e4d9SJoe Stringer * Description 1987799e4d9SJoe Stringer * Look up and delete an element by key in a specified map. 1997799e4d9SJoe Stringer * 2007799e4d9SJoe Stringer * Return 2017799e4d9SJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 2027799e4d9SJoe Stringer * is set appropriately. 2037799e4d9SJoe Stringer * 2047799e4d9SJoe Stringer * BPF_MAP_GET_NEXT_KEY 2057799e4d9SJoe Stringer * Description 2067799e4d9SJoe Stringer * Look up an element by key in a specified map and return the key 2077799e4d9SJoe Stringer * of the next element. Can be used to iterate over all elements 2087799e4d9SJoe Stringer * in the map. 2097799e4d9SJoe Stringer * 2107799e4d9SJoe Stringer * Return 2117799e4d9SJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 2127799e4d9SJoe Stringer * is set appropriately. 2137799e4d9SJoe Stringer * 2147799e4d9SJoe Stringer * The following cases can be used to iterate over all elements of 2157799e4d9SJoe Stringer * the map: 2167799e4d9SJoe Stringer * 2177799e4d9SJoe Stringer * * If *key* is not found, the operation returns zero and sets 2187799e4d9SJoe Stringer * the *next_key* pointer to the key of the first element. 2197799e4d9SJoe Stringer * * If *key* is found, the operation returns zero and sets the 2207799e4d9SJoe Stringer * *next_key* pointer to the key of the next element. 2217799e4d9SJoe Stringer * * If *key* is the last element, returns -1 and *errno* is set 2227799e4d9SJoe Stringer * to **ENOENT**. 2237799e4d9SJoe Stringer * 2247799e4d9SJoe Stringer * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 2257799e4d9SJoe Stringer * **EINVAL** on error. 2267799e4d9SJoe Stringer * 2277799e4d9SJoe Stringer * BPF_PROG_LOAD 2287799e4d9SJoe Stringer * Description 2297799e4d9SJoe Stringer * Verify and load an eBPF program, returning a new file 2307799e4d9SJoe Stringer * descriptor associated with the program. 2317799e4d9SJoe Stringer * 2327799e4d9SJoe Stringer * Applying **close**\ (2) to the file descriptor returned by 2337799e4d9SJoe Stringer * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 2347799e4d9SJoe Stringer * 2357799e4d9SJoe Stringer * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 2367799e4d9SJoe Stringer * automatically enabled for the new file descriptor. 2377799e4d9SJoe Stringer * 2387799e4d9SJoe Stringer * Return 2397799e4d9SJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 2407799e4d9SJoe Stringer * error occurred (in which case, *errno* is set appropriately). 2417799e4d9SJoe Stringer * 242f67c9cbfSJoe Stringer * BPF_OBJ_PIN 243f67c9cbfSJoe Stringer * Description 244f67c9cbfSJoe Stringer * Pin an eBPF program or map referred by the specified *bpf_fd* 245f67c9cbfSJoe Stringer * to the provided *pathname* on the filesystem. 246f67c9cbfSJoe Stringer * 2478aacb3c8SJoe Stringer * The *pathname* argument must not contain a dot ("."). 2488aacb3c8SJoe Stringer * 2498aacb3c8SJoe Stringer * On success, *pathname* retains a reference to the eBPF object, 2508aacb3c8SJoe Stringer * preventing deallocation of the object when the original 2518aacb3c8SJoe Stringer * *bpf_fd* is closed. This allow the eBPF object to live beyond 2528aacb3c8SJoe Stringer * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 2538aacb3c8SJoe Stringer * process. 2548aacb3c8SJoe Stringer * 2558aacb3c8SJoe Stringer * Applying **unlink**\ (2) or similar calls to the *pathname* 2568aacb3c8SJoe Stringer * unpins the object from the filesystem, removing the reference. 2578aacb3c8SJoe Stringer * If no other file descriptors or filesystem nodes refer to the 2588aacb3c8SJoe Stringer * same object, it will be deallocated (see NOTES). 2598aacb3c8SJoe Stringer * 2608aacb3c8SJoe Stringer * The filesystem type for the parent directory of *pathname* must 2618aacb3c8SJoe Stringer * be **BPF_FS_MAGIC**. 2628aacb3c8SJoe Stringer * 263f67c9cbfSJoe Stringer * Return 264f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 265f67c9cbfSJoe Stringer * is set appropriately. 266f67c9cbfSJoe Stringer * 267f67c9cbfSJoe Stringer * BPF_OBJ_GET 268f67c9cbfSJoe Stringer * Description 269f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF object pinned to the 270f67c9cbfSJoe Stringer * specified *pathname*. 271f67c9cbfSJoe Stringer * 272f67c9cbfSJoe Stringer * Return 273f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 274f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 275f67c9cbfSJoe Stringer * 276f67c9cbfSJoe Stringer * BPF_PROG_ATTACH 277f67c9cbfSJoe Stringer * Description 278f67c9cbfSJoe Stringer * Attach an eBPF program to a *target_fd* at the specified 279f67c9cbfSJoe Stringer * *attach_type* hook. 280f67c9cbfSJoe Stringer * 28132e76b18SJoe Stringer * The *attach_type* specifies the eBPF attachment point to 28232e76b18SJoe Stringer * attach the program to, and must be one of *bpf_attach_type* 28332e76b18SJoe Stringer * (see below). 28432e76b18SJoe Stringer * 28532e76b18SJoe Stringer * The *attach_bpf_fd* must be a valid file descriptor for a 28632e76b18SJoe Stringer * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 28732e76b18SJoe Stringer * or sock_ops type corresponding to the specified *attach_type*. 28832e76b18SJoe Stringer * 28932e76b18SJoe Stringer * The *target_fd* must be a valid file descriptor for a kernel 29032e76b18SJoe Stringer * object which depends on the attach type of *attach_bpf_fd*: 29132e76b18SJoe Stringer * 29232e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_DEVICE**, 29332e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SKB**, 29432e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK**, 29532e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 29632e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 29732e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 29832e76b18SJoe Stringer * **BPF_PROG_TYPE_SOCK_OPS** 29932e76b18SJoe Stringer * 30032e76b18SJoe Stringer * Control Group v2 hierarchy with the eBPF controller 30132e76b18SJoe Stringer * enabled. Requires the kernel to be compiled with 30232e76b18SJoe Stringer * **CONFIG_CGROUP_BPF**. 30332e76b18SJoe Stringer * 30432e76b18SJoe Stringer * **BPF_PROG_TYPE_FLOW_DISSECTOR** 30532e76b18SJoe Stringer * 30632e76b18SJoe Stringer * Network namespace (eg /proc/self/ns/net). 30732e76b18SJoe Stringer * 30832e76b18SJoe Stringer * **BPF_PROG_TYPE_LIRC_MODE2** 30932e76b18SJoe Stringer * 31032e76b18SJoe Stringer * LIRC device path (eg /dev/lircN). Requires the kernel 31132e76b18SJoe Stringer * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 31232e76b18SJoe Stringer * 31332e76b18SJoe Stringer * **BPF_PROG_TYPE_SK_SKB**, 31432e76b18SJoe Stringer * **BPF_PROG_TYPE_SK_MSG** 31532e76b18SJoe Stringer * 31632e76b18SJoe Stringer * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 31732e76b18SJoe Stringer * 318f67c9cbfSJoe Stringer * Return 319f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 320f67c9cbfSJoe Stringer * is set appropriately. 321f67c9cbfSJoe Stringer * 322f67c9cbfSJoe Stringer * BPF_PROG_DETACH 323f67c9cbfSJoe Stringer * Description 324f67c9cbfSJoe Stringer * Detach the eBPF program associated with the *target_fd* at the 325f67c9cbfSJoe Stringer * hook specified by *attach_type*. The program must have been 326f67c9cbfSJoe Stringer * previously attached using **BPF_PROG_ATTACH**. 327f67c9cbfSJoe Stringer * 328f67c9cbfSJoe Stringer * Return 329f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 330f67c9cbfSJoe Stringer * is set appropriately. 331f67c9cbfSJoe Stringer * 332f67c9cbfSJoe Stringer * BPF_PROG_TEST_RUN 333f67c9cbfSJoe Stringer * Description 3342a3fdca4SJoe Stringer * Run the eBPF program associated with the *prog_fd* a *repeat* 3352a3fdca4SJoe Stringer * number of times against a provided program context *ctx_in* and 3362a3fdca4SJoe Stringer * data *data_in*, and return the modified program context 3372a3fdca4SJoe Stringer * *ctx_out*, *data_out* (for example, packet data), result of the 3382a3fdca4SJoe Stringer * execution *retval*, and *duration* of the test run. 339f67c9cbfSJoe Stringer * 340f3c45326SJoe Stringer * The sizes of the buffers provided as input and output 341f3c45326SJoe Stringer * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 342f3c45326SJoe Stringer * be provided in the corresponding variables *ctx_size_in*, 343f3c45326SJoe Stringer * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 344f3c45326SJoe Stringer * of these parameters are not provided (ie set to NULL), the 345f3c45326SJoe Stringer * corresponding size field must be zero. 346f3c45326SJoe Stringer * 347f3c45326SJoe Stringer * Some program types have particular requirements: 348f3c45326SJoe Stringer * 349f3c45326SJoe Stringer * **BPF_PROG_TYPE_SK_LOOKUP** 350f3c45326SJoe Stringer * *data_in* and *data_out* must be NULL. 351f3c45326SJoe Stringer * 352f3c45326SJoe Stringer * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 353f3c45326SJoe Stringer * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 354f3c45326SJoe Stringer * 355f3c45326SJoe Stringer * *ctx_out*, *data_in* and *data_out* must be NULL. 356f3c45326SJoe Stringer * *repeat* must be zero. 357f3c45326SJoe Stringer * 358e40fbbf0SUsama Arif * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 359e40fbbf0SUsama Arif * 360f67c9cbfSJoe Stringer * Return 361f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 362f67c9cbfSJoe Stringer * is set appropriately. 363f67c9cbfSJoe Stringer * 3642a3fdca4SJoe Stringer * **ENOSPC** 3652a3fdca4SJoe Stringer * Either *data_size_out* or *ctx_size_out* is too small. 3662a3fdca4SJoe Stringer * **ENOTSUPP** 3672a3fdca4SJoe Stringer * This command is not supported by the program type of 3682a3fdca4SJoe Stringer * the program referred to by *prog_fd*. 3692a3fdca4SJoe Stringer * 370f67c9cbfSJoe Stringer * BPF_PROG_GET_NEXT_ID 371f67c9cbfSJoe Stringer * Description 372f67c9cbfSJoe Stringer * Fetch the next eBPF program currently loaded into the kernel. 373f67c9cbfSJoe Stringer * 374f67c9cbfSJoe Stringer * Looks for the eBPF program with an id greater than *start_id* 375f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF programs 376f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 377f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 378f67c9cbfSJoe Stringer * 379f67c9cbfSJoe Stringer * Return 380f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 381f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 382f67c9cbfSJoe Stringer * 383f67c9cbfSJoe Stringer * BPF_MAP_GET_NEXT_ID 384f67c9cbfSJoe Stringer * Description 385f67c9cbfSJoe Stringer * Fetch the next eBPF map currently loaded into the kernel. 386f67c9cbfSJoe Stringer * 387f67c9cbfSJoe Stringer * Looks for the eBPF map with an id greater than *start_id* 388f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF maps 389f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 390f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 391f67c9cbfSJoe Stringer * 392f67c9cbfSJoe Stringer * Return 393f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 394f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 395f67c9cbfSJoe Stringer * 396f67c9cbfSJoe Stringer * BPF_PROG_GET_FD_BY_ID 397f67c9cbfSJoe Stringer * Description 398f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF program corresponding to 399f67c9cbfSJoe Stringer * *prog_id*. 400f67c9cbfSJoe Stringer * 401f67c9cbfSJoe Stringer * Return 402f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 403f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 404f67c9cbfSJoe Stringer * 405f67c9cbfSJoe Stringer * BPF_MAP_GET_FD_BY_ID 406f67c9cbfSJoe Stringer * Description 407f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF map corresponding to 408f67c9cbfSJoe Stringer * *map_id*. 409f67c9cbfSJoe Stringer * 410f67c9cbfSJoe Stringer * Return 411f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 412f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 413f67c9cbfSJoe Stringer * 414f67c9cbfSJoe Stringer * BPF_OBJ_GET_INFO_BY_FD 415f67c9cbfSJoe Stringer * Description 416f67c9cbfSJoe Stringer * Obtain information about the eBPF object corresponding to 417f67c9cbfSJoe Stringer * *bpf_fd*. 418f67c9cbfSJoe Stringer * 419f67c9cbfSJoe Stringer * Populates up to *info_len* bytes of *info*, which will be in 420f67c9cbfSJoe Stringer * one of the following formats depending on the eBPF object type 421f67c9cbfSJoe Stringer * of *bpf_fd*: 422f67c9cbfSJoe Stringer * 423f67c9cbfSJoe Stringer * * **struct bpf_prog_info** 424f67c9cbfSJoe Stringer * * **struct bpf_map_info** 425f67c9cbfSJoe Stringer * * **struct bpf_btf_info** 426f67c9cbfSJoe Stringer * * **struct bpf_link_info** 427f67c9cbfSJoe Stringer * 428f67c9cbfSJoe Stringer * Return 429f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 430f67c9cbfSJoe Stringer * is set appropriately. 431f67c9cbfSJoe Stringer * 432f67c9cbfSJoe Stringer * BPF_PROG_QUERY 433f67c9cbfSJoe Stringer * Description 434f67c9cbfSJoe Stringer * Obtain information about eBPF programs associated with the 435f67c9cbfSJoe Stringer * specified *attach_type* hook. 436f67c9cbfSJoe Stringer * 4375d999994SJoe Stringer * The *target_fd* must be a valid file descriptor for a kernel 4385d999994SJoe Stringer * object which depends on the attach type of *attach_bpf_fd*: 4395d999994SJoe Stringer * 4405d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_DEVICE**, 4415d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SKB**, 4425d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK**, 4435d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 4445d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 4455d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 4465d999994SJoe Stringer * **BPF_PROG_TYPE_SOCK_OPS** 4475d999994SJoe Stringer * 4485d999994SJoe Stringer * Control Group v2 hierarchy with the eBPF controller 4495d999994SJoe Stringer * enabled. Requires the kernel to be compiled with 4505d999994SJoe Stringer * **CONFIG_CGROUP_BPF**. 4515d999994SJoe Stringer * 4525d999994SJoe Stringer * **BPF_PROG_TYPE_FLOW_DISSECTOR** 4535d999994SJoe Stringer * 4545d999994SJoe Stringer * Network namespace (eg /proc/self/ns/net). 4555d999994SJoe Stringer * 4565d999994SJoe Stringer * **BPF_PROG_TYPE_LIRC_MODE2** 4575d999994SJoe Stringer * 4585d999994SJoe Stringer * LIRC device path (eg /dev/lircN). Requires the kernel 4595d999994SJoe Stringer * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 4605d999994SJoe Stringer * 4615d999994SJoe Stringer * **BPF_PROG_QUERY** always fetches the number of programs 4625d999994SJoe Stringer * attached and the *attach_flags* which were used to attach those 4635d999994SJoe Stringer * programs. Additionally, if *prog_ids* is nonzero and the number 4645d999994SJoe Stringer * of attached programs is less than *prog_cnt*, populates 4655d999994SJoe Stringer * *prog_ids* with the eBPF program ids of the programs attached 4665d999994SJoe Stringer * at *target_fd*. 4675d999994SJoe Stringer * 4685d999994SJoe Stringer * The following flags may alter the result: 4695d999994SJoe Stringer * 4705d999994SJoe Stringer * **BPF_F_QUERY_EFFECTIVE** 4715d999994SJoe Stringer * Only return information regarding programs which are 4725d999994SJoe Stringer * currently effective at the specified *target_fd*. 4735d999994SJoe Stringer * 474f67c9cbfSJoe Stringer * Return 475f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 476f67c9cbfSJoe Stringer * is set appropriately. 477f67c9cbfSJoe Stringer * 478f67c9cbfSJoe Stringer * BPF_RAW_TRACEPOINT_OPEN 479f67c9cbfSJoe Stringer * Description 480f67c9cbfSJoe Stringer * Attach an eBPF program to a tracepoint *name* to access kernel 481f67c9cbfSJoe Stringer * internal arguments of the tracepoint in their raw form. 482f67c9cbfSJoe Stringer * 483f67c9cbfSJoe Stringer * The *prog_fd* must be a valid file descriptor associated with 484f67c9cbfSJoe Stringer * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 485f67c9cbfSJoe Stringer * 486f67c9cbfSJoe Stringer * No ABI guarantees are made about the content of tracepoint 487f67c9cbfSJoe Stringer * arguments exposed to the corresponding eBPF program. 488f67c9cbfSJoe Stringer * 489f67c9cbfSJoe Stringer * Applying **close**\ (2) to the file descriptor returned by 490f67c9cbfSJoe Stringer * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 491f67c9cbfSJoe Stringer * 492f67c9cbfSJoe Stringer * Return 493f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 494f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 495f67c9cbfSJoe Stringer * 496f67c9cbfSJoe Stringer * BPF_BTF_LOAD 497f67c9cbfSJoe Stringer * Description 498f67c9cbfSJoe Stringer * Verify and load BPF Type Format (BTF) metadata into the kernel, 499f67c9cbfSJoe Stringer * returning a new file descriptor associated with the metadata. 500f67c9cbfSJoe Stringer * BTF is described in more detail at 501f67c9cbfSJoe Stringer * https://www.kernel.org/doc/html/latest/bpf/btf.html. 502f67c9cbfSJoe Stringer * 503f67c9cbfSJoe Stringer * The *btf* parameter must point to valid memory providing 504f67c9cbfSJoe Stringer * *btf_size* bytes of BTF binary metadata. 505f67c9cbfSJoe Stringer * 506f67c9cbfSJoe Stringer * The returned file descriptor can be passed to other **bpf**\ () 507f67c9cbfSJoe Stringer * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 508f67c9cbfSJoe Stringer * associate the BTF with those objects. 509f67c9cbfSJoe Stringer * 510f67c9cbfSJoe Stringer * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 511f67c9cbfSJoe Stringer * parameters to specify a *btf_log_buf*, *btf_log_size* and 512f67c9cbfSJoe Stringer * *btf_log_level* which allow the kernel to return freeform log 513f67c9cbfSJoe Stringer * output regarding the BTF verification process. 514f67c9cbfSJoe Stringer * 515f67c9cbfSJoe Stringer * Return 516f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 517f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 518f67c9cbfSJoe Stringer * 519f67c9cbfSJoe Stringer * BPF_BTF_GET_FD_BY_ID 520f67c9cbfSJoe Stringer * Description 521f67c9cbfSJoe Stringer * Open a file descriptor for the BPF Type Format (BTF) 522f67c9cbfSJoe Stringer * corresponding to *btf_id*. 523f67c9cbfSJoe Stringer * 524f67c9cbfSJoe Stringer * Return 525f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 526f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 527f67c9cbfSJoe Stringer * 528f67c9cbfSJoe Stringer * BPF_TASK_FD_QUERY 529f67c9cbfSJoe Stringer * Description 530f67c9cbfSJoe Stringer * Obtain information about eBPF programs associated with the 531f67c9cbfSJoe Stringer * target process identified by *pid* and *fd*. 532f67c9cbfSJoe Stringer * 533f67c9cbfSJoe Stringer * If the *pid* and *fd* are associated with a tracepoint, kprobe 534f67c9cbfSJoe Stringer * or uprobe perf event, then the *prog_id* and *fd_type* will 535f67c9cbfSJoe Stringer * be populated with the eBPF program id and file descriptor type 536f67c9cbfSJoe Stringer * of type **bpf_task_fd_type**. If associated with a kprobe or 537f67c9cbfSJoe Stringer * uprobe, the *probe_offset* and *probe_addr* will also be 538f67c9cbfSJoe Stringer * populated. Optionally, if *buf* is provided, then up to 539f67c9cbfSJoe Stringer * *buf_len* bytes of *buf* will be populated with the name of 540f67c9cbfSJoe Stringer * the tracepoint, kprobe or uprobe. 541f67c9cbfSJoe Stringer * 542f67c9cbfSJoe Stringer * The resulting *prog_id* may be introspected in deeper detail 543f67c9cbfSJoe Stringer * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 544f67c9cbfSJoe Stringer * 545f67c9cbfSJoe Stringer * Return 546f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 547f67c9cbfSJoe Stringer * is set appropriately. 548f67c9cbfSJoe Stringer * 549f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_AND_DELETE_ELEM 550f67c9cbfSJoe Stringer * Description 551f67c9cbfSJoe Stringer * Look up an element with the given *key* in the map referred to 552f67c9cbfSJoe Stringer * by the file descriptor *fd*, and if found, delete the element. 553f67c9cbfSJoe Stringer * 5543e87f192SDenis Salopek * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 5553e87f192SDenis Salopek * types, the *flags* argument needs to be set to 0, but for other 5563e87f192SDenis Salopek * map types, it may be specified as: 5573e87f192SDenis Salopek * 5583e87f192SDenis Salopek * **BPF_F_LOCK** 5593e87f192SDenis Salopek * Look up and delete the value of a spin-locked map 5603e87f192SDenis Salopek * without returning the lock. This must be specified if 5613e87f192SDenis Salopek * the elements contain a spinlock. 5623e87f192SDenis Salopek * 563f67c9cbfSJoe Stringer * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 564f67c9cbfSJoe Stringer * implement this command as a "pop" operation, deleting the top 565f67c9cbfSJoe Stringer * element rather than one corresponding to *key*. 566f67c9cbfSJoe Stringer * The *key* and *key_len* parameters should be zeroed when 567f67c9cbfSJoe Stringer * issuing this operation for these map types. 568f67c9cbfSJoe Stringer * 569f67c9cbfSJoe Stringer * This command is only valid for the following map types: 570f67c9cbfSJoe Stringer * * **BPF_MAP_TYPE_QUEUE** 571f67c9cbfSJoe Stringer * * **BPF_MAP_TYPE_STACK** 5723e87f192SDenis Salopek * * **BPF_MAP_TYPE_HASH** 5733e87f192SDenis Salopek * * **BPF_MAP_TYPE_PERCPU_HASH** 5743e87f192SDenis Salopek * * **BPF_MAP_TYPE_LRU_HASH** 5753e87f192SDenis Salopek * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 576f67c9cbfSJoe Stringer * 577f67c9cbfSJoe Stringer * Return 578f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 579f67c9cbfSJoe Stringer * is set appropriately. 580f67c9cbfSJoe Stringer * 581f67c9cbfSJoe Stringer * BPF_MAP_FREEZE 582f67c9cbfSJoe Stringer * Description 583f67c9cbfSJoe Stringer * Freeze the permissions of the specified map. 584f67c9cbfSJoe Stringer * 585f67c9cbfSJoe Stringer * Write permissions may be frozen by passing zero *flags*. 586f67c9cbfSJoe Stringer * Upon success, no future syscall invocations may alter the 587f67c9cbfSJoe Stringer * map state of *map_fd*. Write operations from eBPF programs 588f67c9cbfSJoe Stringer * are still possible for a frozen map. 589f67c9cbfSJoe Stringer * 590f67c9cbfSJoe Stringer * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 591f67c9cbfSJoe Stringer * 592f67c9cbfSJoe Stringer * Return 593f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 594f67c9cbfSJoe Stringer * is set appropriately. 595f67c9cbfSJoe Stringer * 596f67c9cbfSJoe Stringer * BPF_BTF_GET_NEXT_ID 597f67c9cbfSJoe Stringer * Description 598f67c9cbfSJoe Stringer * Fetch the next BPF Type Format (BTF) object currently loaded 599f67c9cbfSJoe Stringer * into the kernel. 600f67c9cbfSJoe Stringer * 601f67c9cbfSJoe Stringer * Looks for the BTF object with an id greater than *start_id* 602f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other BTF objects 603f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 604f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 605f67c9cbfSJoe Stringer * 606f67c9cbfSJoe Stringer * Return 607f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 608f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 609f67c9cbfSJoe Stringer * 610f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_BATCH 611f67c9cbfSJoe Stringer * Description 612f67c9cbfSJoe Stringer * Iterate and fetch multiple elements in a map. 613f67c9cbfSJoe Stringer * 6140cb80454SJoe Stringer * Two opaque values are used to manage batch operations, 6150cb80454SJoe Stringer * *in_batch* and *out_batch*. Initially, *in_batch* must be set 6160cb80454SJoe Stringer * to NULL to begin the batched operation. After each subsequent 6170cb80454SJoe Stringer * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 6180cb80454SJoe Stringer * *out_batch* as the *in_batch* for the next operation to 6190cb80454SJoe Stringer * continue iteration from the current point. 6200cb80454SJoe Stringer * 6210cb80454SJoe Stringer * The *keys* and *values* are output parameters which must point 6220cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 6230cb80454SJoe Stringer * and value size of the map *map_fd*. The *keys* buffer must be 6240cb80454SJoe Stringer * of *key_size* * *count*. The *values* buffer must be of 6250cb80454SJoe Stringer * *value_size* * *count*. 6260cb80454SJoe Stringer * 6270cb80454SJoe Stringer * The *elem_flags* argument may be specified as one of the 6280cb80454SJoe Stringer * following: 6290cb80454SJoe Stringer * 6300cb80454SJoe Stringer * **BPF_F_LOCK** 6310cb80454SJoe Stringer * Look up the value of a spin-locked map without 6320cb80454SJoe Stringer * returning the lock. This must be specified if the 6330cb80454SJoe Stringer * elements contain a spinlock. 6340cb80454SJoe Stringer * 6350cb80454SJoe Stringer * On success, *count* elements from the map are copied into the 6360cb80454SJoe Stringer * user buffer, with the keys copied into *keys* and the values 6370cb80454SJoe Stringer * copied into the corresponding indices in *values*. 6380cb80454SJoe Stringer * 6390cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 6400cb80454SJoe Stringer * is set to the number of successfully processed elements. 6410cb80454SJoe Stringer * 642f67c9cbfSJoe Stringer * Return 643f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 644f67c9cbfSJoe Stringer * is set appropriately. 645f67c9cbfSJoe Stringer * 6460cb80454SJoe Stringer * May set *errno* to **ENOSPC** to indicate that *keys* or 6470cb80454SJoe Stringer * *values* is too small to dump an entire bucket during 6480cb80454SJoe Stringer * iteration of a hash-based map type. 6490cb80454SJoe Stringer * 650f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_AND_DELETE_BATCH 651f67c9cbfSJoe Stringer * Description 6520cb80454SJoe Stringer * Iterate and delete all elements in a map. 6530cb80454SJoe Stringer * 6540cb80454SJoe Stringer * This operation has the same behavior as 6550cb80454SJoe Stringer * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 6560cb80454SJoe Stringer * 6570cb80454SJoe Stringer * * Every element that is successfully returned is also deleted 6580cb80454SJoe Stringer * from the map. This is at least *count* elements. Note that 6590cb80454SJoe Stringer * *count* is both an input and an output parameter. 6600cb80454SJoe Stringer * * Upon returning with *errno* set to **EFAULT**, up to 6610cb80454SJoe Stringer * *count* elements may be deleted without returning the keys 6620cb80454SJoe Stringer * and values of the deleted elements. 663f67c9cbfSJoe Stringer * 664f67c9cbfSJoe Stringer * Return 665f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 666f67c9cbfSJoe Stringer * is set appropriately. 667f67c9cbfSJoe Stringer * 668f67c9cbfSJoe Stringer * BPF_MAP_UPDATE_BATCH 669f67c9cbfSJoe Stringer * Description 6700cb80454SJoe Stringer * Update multiple elements in a map by *key*. 6710cb80454SJoe Stringer * 6720cb80454SJoe Stringer * The *keys* and *values* are input parameters which must point 6730cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 6740cb80454SJoe Stringer * and value size of the map *map_fd*. The *keys* buffer must be 6750cb80454SJoe Stringer * of *key_size* * *count*. The *values* buffer must be of 6760cb80454SJoe Stringer * *value_size* * *count*. 6770cb80454SJoe Stringer * 6780cb80454SJoe Stringer * Each element specified in *keys* is sequentially updated to the 6790cb80454SJoe Stringer * value in the corresponding index in *values*. The *in_batch* 6800cb80454SJoe Stringer * and *out_batch* parameters are ignored and should be zeroed. 6810cb80454SJoe Stringer * 6820cb80454SJoe Stringer * The *elem_flags* argument should be specified as one of the 6830cb80454SJoe Stringer * following: 6840cb80454SJoe Stringer * 6850cb80454SJoe Stringer * **BPF_ANY** 6860cb80454SJoe Stringer * Create new elements or update a existing elements. 6870cb80454SJoe Stringer * **BPF_NOEXIST** 6880cb80454SJoe Stringer * Create new elements only if they do not exist. 6890cb80454SJoe Stringer * **BPF_EXIST** 6900cb80454SJoe Stringer * Update existing elements. 6910cb80454SJoe Stringer * **BPF_F_LOCK** 6920cb80454SJoe Stringer * Update spin_lock-ed map elements. This must be 6930cb80454SJoe Stringer * specified if the map value contains a spinlock. 6940cb80454SJoe Stringer * 6950cb80454SJoe Stringer * On success, *count* elements from the map are updated. 6960cb80454SJoe Stringer * 6970cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 6980cb80454SJoe Stringer * is set to the number of successfully processed elements. 699f67c9cbfSJoe Stringer * 700f67c9cbfSJoe Stringer * Return 701f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 702f67c9cbfSJoe Stringer * is set appropriately. 703f67c9cbfSJoe Stringer * 7040cb80454SJoe Stringer * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 7050cb80454SJoe Stringer * **E2BIG**. **E2BIG** indicates that the number of elements in 7060cb80454SJoe Stringer * the map reached the *max_entries* limit specified at map 7070cb80454SJoe Stringer * creation time. 7080cb80454SJoe Stringer * 7090cb80454SJoe Stringer * May set *errno* to one of the following error codes under 7100cb80454SJoe Stringer * specific circumstances: 7110cb80454SJoe Stringer * 7120cb80454SJoe Stringer * **EEXIST** 7130cb80454SJoe Stringer * If *flags* specifies **BPF_NOEXIST** and the element 7140cb80454SJoe Stringer * with *key* already exists in the map. 7150cb80454SJoe Stringer * **ENOENT** 7160cb80454SJoe Stringer * If *flags* specifies **BPF_EXIST** and the element with 7170cb80454SJoe Stringer * *key* does not exist in the map. 7180cb80454SJoe Stringer * 719f67c9cbfSJoe Stringer * BPF_MAP_DELETE_BATCH 720f67c9cbfSJoe Stringer * Description 7210cb80454SJoe Stringer * Delete multiple elements in a map by *key*. 7220cb80454SJoe Stringer * 7230cb80454SJoe Stringer * The *keys* parameter is an input parameter which must point 7240cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 7250cb80454SJoe Stringer * size of the map *map_fd*, that is, *key_size* * *count*. 7260cb80454SJoe Stringer * 7270cb80454SJoe Stringer * Each element specified in *keys* is sequentially deleted. The 7280cb80454SJoe Stringer * *in_batch*, *out_batch*, and *values* parameters are ignored 7290cb80454SJoe Stringer * and should be zeroed. 7300cb80454SJoe Stringer * 7310cb80454SJoe Stringer * The *elem_flags* argument may be specified as one of the 7320cb80454SJoe Stringer * following: 7330cb80454SJoe Stringer * 7340cb80454SJoe Stringer * **BPF_F_LOCK** 7350cb80454SJoe Stringer * Look up the value of a spin-locked map without 7360cb80454SJoe Stringer * returning the lock. This must be specified if the 7370cb80454SJoe Stringer * elements contain a spinlock. 7380cb80454SJoe Stringer * 7390cb80454SJoe Stringer * On success, *count* elements from the map are updated. 7400cb80454SJoe Stringer * 7410cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 7420cb80454SJoe Stringer * is set to the number of successfully processed elements. If 7430cb80454SJoe Stringer * *errno* is **EFAULT**, up to *count* elements may be been 7440cb80454SJoe Stringer * deleted. 745f67c9cbfSJoe Stringer * 746f67c9cbfSJoe Stringer * Return 747f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 748f67c9cbfSJoe Stringer * is set appropriately. 749f67c9cbfSJoe Stringer * 750f67c9cbfSJoe Stringer * BPF_LINK_CREATE 751f67c9cbfSJoe Stringer * Description 752f67c9cbfSJoe Stringer * Attach an eBPF program to a *target_fd* at the specified 753f67c9cbfSJoe Stringer * *attach_type* hook and return a file descriptor handle for 754f67c9cbfSJoe Stringer * managing the link. 755f67c9cbfSJoe Stringer * 756f67c9cbfSJoe Stringer * Return 757f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 758f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 759f67c9cbfSJoe Stringer * 760f67c9cbfSJoe Stringer * BPF_LINK_UPDATE 761f67c9cbfSJoe Stringer * Description 762f67c9cbfSJoe Stringer * Update the eBPF program in the specified *link_fd* to 763f67c9cbfSJoe Stringer * *new_prog_fd*. 764f67c9cbfSJoe Stringer * 765f67c9cbfSJoe Stringer * Return 766f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 767f67c9cbfSJoe Stringer * is set appropriately. 768f67c9cbfSJoe Stringer * 769f67c9cbfSJoe Stringer * BPF_LINK_GET_FD_BY_ID 770f67c9cbfSJoe Stringer * Description 771f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF Link corresponding to 772f67c9cbfSJoe Stringer * *link_id*. 773f67c9cbfSJoe Stringer * 774f67c9cbfSJoe Stringer * Return 775f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 776f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 777f67c9cbfSJoe Stringer * 778f67c9cbfSJoe Stringer * BPF_LINK_GET_NEXT_ID 779f67c9cbfSJoe Stringer * Description 780f67c9cbfSJoe Stringer * Fetch the next eBPF link currently loaded into the kernel. 781f67c9cbfSJoe Stringer * 782f67c9cbfSJoe Stringer * Looks for the eBPF link with an id greater than *start_id* 783f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF links 784f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 785f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 786f67c9cbfSJoe Stringer * 787f67c9cbfSJoe Stringer * Return 788f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 789f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 790f67c9cbfSJoe Stringer * 791f67c9cbfSJoe Stringer * BPF_ENABLE_STATS 792f67c9cbfSJoe Stringer * Description 793f67c9cbfSJoe Stringer * Enable eBPF runtime statistics gathering. 794f67c9cbfSJoe Stringer * 795f67c9cbfSJoe Stringer * Runtime statistics gathering for the eBPF runtime is disabled 796f67c9cbfSJoe Stringer * by default to minimize the corresponding performance overhead. 797f67c9cbfSJoe Stringer * This command enables statistics globally. 798f67c9cbfSJoe Stringer * 799f67c9cbfSJoe Stringer * Multiple programs may independently enable statistics. 800f67c9cbfSJoe Stringer * After gathering the desired statistics, eBPF runtime statistics 801f67c9cbfSJoe Stringer * may be disabled again by calling **close**\ (2) for the file 802f67c9cbfSJoe Stringer * descriptor returned by this function. Statistics will only be 803f67c9cbfSJoe Stringer * disabled system-wide when all outstanding file descriptors 804f67c9cbfSJoe Stringer * returned by prior calls for this subcommand are closed. 805f67c9cbfSJoe Stringer * 806f67c9cbfSJoe Stringer * Return 807f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 808f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 809f67c9cbfSJoe Stringer * 810f67c9cbfSJoe Stringer * BPF_ITER_CREATE 811f67c9cbfSJoe Stringer * Description 812f67c9cbfSJoe Stringer * Create an iterator on top of the specified *link_fd* (as 813f67c9cbfSJoe Stringer * previously created using **BPF_LINK_CREATE**) and return a 814f67c9cbfSJoe Stringer * file descriptor that can be used to trigger the iteration. 815f67c9cbfSJoe Stringer * 816f67c9cbfSJoe Stringer * If the resulting file descriptor is pinned to the filesystem 817f67c9cbfSJoe Stringer * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 818f67c9cbfSJoe Stringer * for that path will trigger the iterator to read kernel state 819f67c9cbfSJoe Stringer * using the eBPF program attached to *link_fd*. 820f67c9cbfSJoe Stringer * 821f67c9cbfSJoe Stringer * Return 822f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 823f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 824f67c9cbfSJoe Stringer * 825f67c9cbfSJoe Stringer * BPF_LINK_DETACH 826f67c9cbfSJoe Stringer * Description 827f67c9cbfSJoe Stringer * Forcefully detach the specified *link_fd* from its 828f67c9cbfSJoe Stringer * corresponding attachment point. 829f67c9cbfSJoe Stringer * 830f67c9cbfSJoe Stringer * Return 831f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 832f67c9cbfSJoe Stringer * is set appropriately. 833f67c9cbfSJoe Stringer * 834f67c9cbfSJoe Stringer * BPF_PROG_BIND_MAP 835f67c9cbfSJoe Stringer * Description 836f67c9cbfSJoe Stringer * Bind a map to the lifetime of an eBPF program. 837f67c9cbfSJoe Stringer * 838f67c9cbfSJoe Stringer * The map identified by *map_fd* is bound to the program 839f67c9cbfSJoe Stringer * identified by *prog_fd* and only released when *prog_fd* is 840f67c9cbfSJoe Stringer * released. This may be used in cases where metadata should be 841f67c9cbfSJoe Stringer * associated with a program which otherwise does not contain any 842f67c9cbfSJoe Stringer * references to the map (for example, embedded in the eBPF 843f67c9cbfSJoe Stringer * program instructions). 844f67c9cbfSJoe Stringer * 845f67c9cbfSJoe Stringer * Return 846f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 847f67c9cbfSJoe Stringer * is set appropriately. 848f67c9cbfSJoe Stringer * 8497799e4d9SJoe Stringer * NOTES 8507799e4d9SJoe Stringer * eBPF objects (maps and programs) can be shared between processes. 8518aacb3c8SJoe Stringer * 8528aacb3c8SJoe Stringer * * After **fork**\ (2), the child inherits file descriptors 8538aacb3c8SJoe Stringer * referring to the same eBPF objects. 8548aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be transferred over 8558aacb3c8SJoe Stringer * **unix**\ (7) domain sockets. 8568aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be duplicated in the 8578aacb3c8SJoe Stringer * usual way, using **dup**\ (2) and similar calls. 8588aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be pinned to the 8598aacb3c8SJoe Stringer * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 8608aacb3c8SJoe Stringer * 8618aacb3c8SJoe Stringer * An eBPF object is deallocated only after all file descriptors referring 8628aacb3c8SJoe Stringer * to the object have been closed and no references remain pinned to the 8638aacb3c8SJoe Stringer * filesystem or attached (for example, bound to a program or device). 8647799e4d9SJoe Stringer */ 86599c55f7dSAlexei Starovoitov enum bpf_cmd { 86699c55f7dSAlexei Starovoitov BPF_MAP_CREATE, 867db20fd2bSAlexei Starovoitov BPF_MAP_LOOKUP_ELEM, 868db20fd2bSAlexei Starovoitov BPF_MAP_UPDATE_ELEM, 869db20fd2bSAlexei Starovoitov BPF_MAP_DELETE_ELEM, 870db20fd2bSAlexei Starovoitov BPF_MAP_GET_NEXT_KEY, 87109756af4SAlexei Starovoitov BPF_PROG_LOAD, 872b2197755SDaniel Borkmann BPF_OBJ_PIN, 873b2197755SDaniel Borkmann BPF_OBJ_GET, 874f4324551SDaniel Mack BPF_PROG_ATTACH, 875f4324551SDaniel Mack BPF_PROG_DETACH, 8761cf1cae9SAlexei Starovoitov BPF_PROG_TEST_RUN, 8775d67f349SAlexei Starovoitov BPF_PROG_RUN = BPF_PROG_TEST_RUN, 87834ad5580SMartin KaFai Lau BPF_PROG_GET_NEXT_ID, 87934ad5580SMartin KaFai Lau BPF_MAP_GET_NEXT_ID, 880b16d9aa4SMartin KaFai Lau BPF_PROG_GET_FD_BY_ID, 881bd5f5f4eSMartin KaFai Lau BPF_MAP_GET_FD_BY_ID, 8821e270976SMartin KaFai Lau BPF_OBJ_GET_INFO_BY_FD, 883468e2f64SAlexei Starovoitov BPF_PROG_QUERY, 884c4f6699dSAlexei Starovoitov BPF_RAW_TRACEPOINT_OPEN, 885f56a653cSMartin KaFai Lau BPF_BTF_LOAD, 88678958fcaSMartin KaFai Lau BPF_BTF_GET_FD_BY_ID, 88741bdc4b4SYonghong Song BPF_TASK_FD_QUERY, 888bd513cd0SMauricio Vasquez B BPF_MAP_LOOKUP_AND_DELETE_ELEM, 88987df15deSDaniel Borkmann BPF_MAP_FREEZE, 8901b9ed84eSQuentin Monnet BPF_BTF_GET_NEXT_ID, 891cb4d03abSBrian Vazquez BPF_MAP_LOOKUP_BATCH, 89205799638SYonghong Song BPF_MAP_LOOKUP_AND_DELETE_BATCH, 893aa2e93b8SBrian Vazquez BPF_MAP_UPDATE_BATCH, 894aa2e93b8SBrian Vazquez BPF_MAP_DELETE_BATCH, 895af6eea57SAndrii Nakryiko BPF_LINK_CREATE, 8960c991ebcSAndrii Nakryiko BPF_LINK_UPDATE, 8972d602c8cSAndrii Nakryiko BPF_LINK_GET_FD_BY_ID, 8982d602c8cSAndrii Nakryiko BPF_LINK_GET_NEXT_ID, 899d46edd67SSong Liu BPF_ENABLE_STATS, 900ac51d99bSYonghong Song BPF_ITER_CREATE, 90173b11c2aSAndrii Nakryiko BPF_LINK_DETACH, 902ef15314aSYiFei Zhu BPF_PROG_BIND_MAP, 90399c55f7dSAlexei Starovoitov }; 90499c55f7dSAlexei Starovoitov 90599c55f7dSAlexei Starovoitov enum bpf_map_type { 90699c55f7dSAlexei Starovoitov BPF_MAP_TYPE_UNSPEC, 9070f8e4bd8SAlexei Starovoitov BPF_MAP_TYPE_HASH, 90828fbcfa0SAlexei Starovoitov BPF_MAP_TYPE_ARRAY, 90904fd61abSAlexei Starovoitov BPF_MAP_TYPE_PROG_ARRAY, 910ea317b26SKaixu Xia BPF_MAP_TYPE_PERF_EVENT_ARRAY, 911824bd0ceSAlexei Starovoitov BPF_MAP_TYPE_PERCPU_HASH, 912a10423b8SAlexei Starovoitov BPF_MAP_TYPE_PERCPU_ARRAY, 913d5a3b1f6SAlexei Starovoitov BPF_MAP_TYPE_STACK_TRACE, 9144ed8ec52SMartin KaFai Lau BPF_MAP_TYPE_CGROUP_ARRAY, 91529ba732aSMartin KaFai Lau BPF_MAP_TYPE_LRU_HASH, 9168f844938SMartin KaFai Lau BPF_MAP_TYPE_LRU_PERCPU_HASH, 917b95a5c4dSDaniel Mack BPF_MAP_TYPE_LPM_TRIE, 91856f668dfSMartin KaFai Lau BPF_MAP_TYPE_ARRAY_OF_MAPS, 919bcc6b1b7SMartin KaFai Lau BPF_MAP_TYPE_HASH_OF_MAPS, 920546ac1ffSJohn Fastabend BPF_MAP_TYPE_DEVMAP, 921174a79ffSJohn Fastabend BPF_MAP_TYPE_SOCKMAP, 9226710e112SJesper Dangaard Brouer BPF_MAP_TYPE_CPUMAP, 923fbfc504aSBjörn Töpel BPF_MAP_TYPE_XSKMAP, 92481110384SJohn Fastabend BPF_MAP_TYPE_SOCKHASH, 925c4bcfb38SYonghong Song BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 926c4bcfb38SYonghong Song /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching 927c4bcfb38SYonghong Song * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to 928c4bcfb38SYonghong Song * both cgroup-attached and other progs and supports all functionality 929c4bcfb38SYonghong Song * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark 930c4bcfb38SYonghong Song * BPF_MAP_TYPE_CGROUP_STORAGE deprecated. 931c4bcfb38SYonghong Song */ 932c4bcfb38SYonghong Song BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 9335dc4c4b7SMartin KaFai Lau BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 934b741f163SRoman Gushchin BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 935f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_QUEUE, 936f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_STACK, 9376ac99e8fSMartin KaFai Lau BPF_MAP_TYPE_SK_STORAGE, 9386f9d451aSToke Høiland-Jørgensen BPF_MAP_TYPE_DEVMAP_HASH, 93985d33df3SMartin KaFai Lau BPF_MAP_TYPE_STRUCT_OPS, 940457f4436SAndrii Nakryiko BPF_MAP_TYPE_RINGBUF, 9418ea63684SKP Singh BPF_MAP_TYPE_INODE_STORAGE, 9424cf1bc1fSKP Singh BPF_MAP_TYPE_TASK_STORAGE, 9439330986cSJoanne Koong BPF_MAP_TYPE_BLOOM_FILTER, 944583c1f42SDavid Vernet BPF_MAP_TYPE_USER_RINGBUF, 945c4bcfb38SYonghong Song BPF_MAP_TYPE_CGRP_STORAGE, 94699c55f7dSAlexei Starovoitov }; 94799c55f7dSAlexei Starovoitov 9486c4fc209SDaniel Borkmann /* Note that tracing related programs such as 9496c4fc209SDaniel Borkmann * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 9506c4fc209SDaniel Borkmann * are not subject to a stable API since kernel internal data 9516c4fc209SDaniel Borkmann * structures can change from release to release and may 9526c4fc209SDaniel Borkmann * therefore break existing tracing BPF programs. Tracing BPF 9536c4fc209SDaniel Borkmann * programs correspond to /a/ specific kernel which is to be 9546c4fc209SDaniel Borkmann * analyzed, and not /a/ specific kernel /and/ all future ones. 9556c4fc209SDaniel Borkmann */ 95609756af4SAlexei Starovoitov enum bpf_prog_type { 95709756af4SAlexei Starovoitov BPF_PROG_TYPE_UNSPEC, 958ddd872bcSAlexei Starovoitov BPF_PROG_TYPE_SOCKET_FILTER, 9592541517cSAlexei Starovoitov BPF_PROG_TYPE_KPROBE, 96096be4325SDaniel Borkmann BPF_PROG_TYPE_SCHED_CLS, 96194caee8cSDaniel Borkmann BPF_PROG_TYPE_SCHED_ACT, 96298b5c2c6SAlexei Starovoitov BPF_PROG_TYPE_TRACEPOINT, 9636a773a15SBrenden Blanco BPF_PROG_TYPE_XDP, 9640515e599SAlexei Starovoitov BPF_PROG_TYPE_PERF_EVENT, 9650e33661dSDaniel Mack BPF_PROG_TYPE_CGROUP_SKB, 96661023658SDavid Ahern BPF_PROG_TYPE_CGROUP_SOCK, 9673a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_IN, 9683a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_OUT, 9693a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_XMIT, 97040304b2aSLawrence Brakmo BPF_PROG_TYPE_SOCK_OPS, 971b005fd18SJohn Fastabend BPF_PROG_TYPE_SK_SKB, 972ebc614f6SRoman Gushchin BPF_PROG_TYPE_CGROUP_DEVICE, 9734f738adbSJohn Fastabend BPF_PROG_TYPE_SK_MSG, 974c4f6699dSAlexei Starovoitov BPF_PROG_TYPE_RAW_TRACEPOINT, 9754fbac77dSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 976004d4b27SMathieu Xhonneux BPF_PROG_TYPE_LWT_SEG6LOCAL, 977f4364dcfSSean Young BPF_PROG_TYPE_LIRC_MODE2, 9782dbb9b9eSMartin KaFai Lau BPF_PROG_TYPE_SK_REUSEPORT, 979d58e468bSPetar Penkov BPF_PROG_TYPE_FLOW_DISSECTOR, 9807b146cebSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SYSCTL, 9819df1c28bSMatt Mullins BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 9820d01da6aSStanislav Fomichev BPF_PROG_TYPE_CGROUP_SOCKOPT, 983f1b9509cSAlexei Starovoitov BPF_PROG_TYPE_TRACING, 98427ae7997SMartin KaFai Lau BPF_PROG_TYPE_STRUCT_OPS, 985be8704ffSAlexei Starovoitov BPF_PROG_TYPE_EXT, 986fc611f47SKP Singh BPF_PROG_TYPE_LSM, 987e9ddbb77SJakub Sitnicki BPF_PROG_TYPE_SK_LOOKUP, 98879a7f8bdSAlexei Starovoitov BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 98984601d6eSFlorian Westphal BPF_PROG_TYPE_NETFILTER, 99009756af4SAlexei Starovoitov }; 99109756af4SAlexei Starovoitov 9920e33661dSDaniel Mack enum bpf_attach_type { 9930e33661dSDaniel Mack BPF_CGROUP_INET_INGRESS, 9940e33661dSDaniel Mack BPF_CGROUP_INET_EGRESS, 99561023658SDavid Ahern BPF_CGROUP_INET_SOCK_CREATE, 99640304b2aSLawrence Brakmo BPF_CGROUP_SOCK_OPS, 997464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_PARSER, 998464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_VERDICT, 999ebc614f6SRoman Gushchin BPF_CGROUP_DEVICE, 10004f738adbSJohn Fastabend BPF_SK_MSG_VERDICT, 10014fbac77dSAndrey Ignatov BPF_CGROUP_INET4_BIND, 10024fbac77dSAndrey Ignatov BPF_CGROUP_INET6_BIND, 1003d74bad4eSAndrey Ignatov BPF_CGROUP_INET4_CONNECT, 1004d74bad4eSAndrey Ignatov BPF_CGROUP_INET6_CONNECT, 1005aac3fc32SAndrey Ignatov BPF_CGROUP_INET4_POST_BIND, 1006aac3fc32SAndrey Ignatov BPF_CGROUP_INET6_POST_BIND, 10071cedee13SAndrey Ignatov BPF_CGROUP_UDP4_SENDMSG, 10081cedee13SAndrey Ignatov BPF_CGROUP_UDP6_SENDMSG, 1009f4364dcfSSean Young BPF_LIRC_MODE2, 1010d58e468bSPetar Penkov BPF_FLOW_DISSECTOR, 10117b146cebSAndrey Ignatov BPF_CGROUP_SYSCTL, 1012983695faSDaniel Borkmann BPF_CGROUP_UDP4_RECVMSG, 1013983695faSDaniel Borkmann BPF_CGROUP_UDP6_RECVMSG, 10140d01da6aSStanislav Fomichev BPF_CGROUP_GETSOCKOPT, 10150d01da6aSStanislav Fomichev BPF_CGROUP_SETSOCKOPT, 1016f1b9509cSAlexei Starovoitov BPF_TRACE_RAW_TP, 1017fec56f58SAlexei Starovoitov BPF_TRACE_FENTRY, 1018fec56f58SAlexei Starovoitov BPF_TRACE_FEXIT, 1019ae240823SKP Singh BPF_MODIFY_RETURN, 1020fc611f47SKP Singh BPF_LSM_MAC, 102115d83c4dSYonghong Song BPF_TRACE_ITER, 10221b66d253SDaniel Borkmann BPF_CGROUP_INET4_GETPEERNAME, 10231b66d253SDaniel Borkmann BPF_CGROUP_INET6_GETPEERNAME, 10241b66d253SDaniel Borkmann BPF_CGROUP_INET4_GETSOCKNAME, 10251b66d253SDaniel Borkmann BPF_CGROUP_INET6_GETSOCKNAME, 1026fbee97feSDavid Ahern BPF_XDP_DEVMAP, 1027f5836749SStanislav Fomichev BPF_CGROUP_INET_SOCK_RELEASE, 102892164774SLorenzo Bianconi BPF_XDP_CPUMAP, 1029e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP, 1030aa8d3a71SAndrii Nakryiko BPF_XDP, 1031a7ba4558SCong Wang BPF_SK_SKB_VERDICT, 1032d5e4ddaeSKuniyuki Iwashima BPF_SK_REUSEPORT_SELECT, 1033d5e4ddaeSKuniyuki Iwashima BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 1034b89fbfbbSAndrii Nakryiko BPF_PERF_EVENT, 10350dcac272SJiri Olsa BPF_TRACE_KPROBE_MULTI, 103669fd337aSStanislav Fomichev BPF_LSM_CGROUP, 103768b04864SKui-Feng Lee BPF_STRUCT_OPS, 1038132328e8SFlorian Westphal BPF_NETFILTER, 10390e33661dSDaniel Mack __MAX_BPF_ATTACH_TYPE 10400e33661dSDaniel Mack }; 10410e33661dSDaniel Mack 10420e33661dSDaniel Mack #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 10430e33661dSDaniel Mack 1044f2e10bffSAndrii Nakryiko enum bpf_link_type { 1045f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_UNSPEC = 0, 1046f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1047f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_TRACING = 2, 1048f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_CGROUP = 3, 1049de4e05caSYonghong Song BPF_LINK_TYPE_ITER = 4, 10507f045a49SJakub Sitnicki BPF_LINK_TYPE_NETNS = 5, 1051aa8d3a71SAndrii Nakryiko BPF_LINK_TYPE_XDP = 6, 1052b89fbfbbSAndrii Nakryiko BPF_LINK_TYPE_PERF_EVENT = 7, 10530dcac272SJiri Olsa BPF_LINK_TYPE_KPROBE_MULTI = 8, 1054f7e0beafSKui-Feng Lee BPF_LINK_TYPE_STRUCT_OPS = 9, 105584601d6eSFlorian Westphal BPF_LINK_TYPE_NETFILTER = 10, 1056f2e10bffSAndrii Nakryiko 1057f2e10bffSAndrii Nakryiko MAX_BPF_LINK_TYPE, 1058f2e10bffSAndrii Nakryiko }; 1059f2e10bffSAndrii Nakryiko 1060*1b715e1bSYafang Shao enum bpf_perf_event_type { 1061*1b715e1bSYafang Shao BPF_PERF_EVENT_UNSPEC = 0, 1062*1b715e1bSYafang Shao BPF_PERF_EVENT_UPROBE = 1, 1063*1b715e1bSYafang Shao BPF_PERF_EVENT_URETPROBE = 2, 1064*1b715e1bSYafang Shao BPF_PERF_EVENT_KPROBE = 3, 1065*1b715e1bSYafang Shao BPF_PERF_EVENT_KRETPROBE = 4, 1066*1b715e1bSYafang Shao BPF_PERF_EVENT_TRACEPOINT = 5, 1067*1b715e1bSYafang Shao BPF_PERF_EVENT_EVENT = 6, 1068*1b715e1bSYafang Shao }; 1069*1b715e1bSYafang Shao 1070324bda9eSAlexei Starovoitov /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1071324bda9eSAlexei Starovoitov * 1072324bda9eSAlexei Starovoitov * NONE(default): No further bpf programs allowed in the subtree. 1073324bda9eSAlexei Starovoitov * 1074324bda9eSAlexei Starovoitov * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1075324bda9eSAlexei Starovoitov * the program in this cgroup yields to sub-cgroup program. 1076324bda9eSAlexei Starovoitov * 1077324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1078324bda9eSAlexei Starovoitov * that cgroup program gets run in addition to the program in this cgroup. 1079324bda9eSAlexei Starovoitov * 1080324bda9eSAlexei Starovoitov * Only one program is allowed to be attached to a cgroup with 1081324bda9eSAlexei Starovoitov * NONE or BPF_F_ALLOW_OVERRIDE flag. 1082324bda9eSAlexei Starovoitov * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1083324bda9eSAlexei Starovoitov * release old program and attach the new one. Attach flags has to match. 1084324bda9eSAlexei Starovoitov * 1085324bda9eSAlexei Starovoitov * Multiple programs are allowed to be attached to a cgroup with 1086324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1087324bda9eSAlexei Starovoitov * (those that were attached first, run first) 1088324bda9eSAlexei Starovoitov * The programs of sub-cgroup are executed first, then programs of 1089324bda9eSAlexei Starovoitov * this cgroup and then programs of parent cgroup. 1090324bda9eSAlexei Starovoitov * When children program makes decision (like picking TCP CA or sock bind) 1091324bda9eSAlexei Starovoitov * parent program has a chance to override it. 1092324bda9eSAlexei Starovoitov * 10937dd68b32SAndrey Ignatov * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 10947dd68b32SAndrey Ignatov * programs for a cgroup. Though it's possible to replace an old program at 10957dd68b32SAndrey Ignatov * any position by also specifying BPF_F_REPLACE flag and position itself in 10967dd68b32SAndrey Ignatov * replace_bpf_fd attribute. Old program at this position will be released. 10977dd68b32SAndrey Ignatov * 1098324bda9eSAlexei Starovoitov * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1099324bda9eSAlexei Starovoitov * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1100324bda9eSAlexei Starovoitov * Ex1: 1101324bda9eSAlexei Starovoitov * cgrp1 (MULTI progs A, B) -> 1102324bda9eSAlexei Starovoitov * cgrp2 (OVERRIDE prog C) -> 1103324bda9eSAlexei Starovoitov * cgrp3 (MULTI prog D) -> 1104324bda9eSAlexei Starovoitov * cgrp4 (OVERRIDE prog E) -> 1105324bda9eSAlexei Starovoitov * cgrp5 (NONE prog F) 1106324bda9eSAlexei Starovoitov * the event in cgrp5 triggers execution of F,D,A,B in that order. 1107324bda9eSAlexei Starovoitov * if prog F is detached, the execution is E,D,A,B 1108324bda9eSAlexei Starovoitov * if prog F and D are detached, the execution is E,A,B 1109324bda9eSAlexei Starovoitov * if prog F, E and D are detached, the execution is C,A,B 1110324bda9eSAlexei Starovoitov * 1111324bda9eSAlexei Starovoitov * All eligible programs are executed regardless of return code from 1112324bda9eSAlexei Starovoitov * earlier programs. 11137f677633SAlexei Starovoitov */ 11147f677633SAlexei Starovoitov #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1115324bda9eSAlexei Starovoitov #define BPF_F_ALLOW_MULTI (1U << 1) 11167dd68b32SAndrey Ignatov #define BPF_F_REPLACE (1U << 2) 11177f677633SAlexei Starovoitov 1118e07b98d9SDavid S. Miller /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1119e07b98d9SDavid S. Miller * verifier will perform strict alignment checking as if the kernel 1120e07b98d9SDavid S. Miller * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1121e07b98d9SDavid S. Miller * and NET_IP_ALIGN defined to 2. 1122e07b98d9SDavid S. Miller */ 1123e07b98d9SDavid S. Miller #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1124e07b98d9SDavid S. Miller 11255a70f4a6SMichael Weiß /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the 1126e9ee9efcSDavid Miller * verifier will allow any alignment whatsoever. On platforms 1127e9ee9efcSDavid Miller * with strict alignment requirements for loads ands stores (such 1128e9ee9efcSDavid Miller * as sparc and mips) the verifier validates that all loads and 1129e9ee9efcSDavid Miller * stores provably follow this requirement. This flag turns that 1130e9ee9efcSDavid Miller * checking and enforcement off. 1131e9ee9efcSDavid Miller * 1132e9ee9efcSDavid Miller * It is mostly used for testing when we want to validate the 1133e9ee9efcSDavid Miller * context and memory access aspects of the verifier, but because 1134e9ee9efcSDavid Miller * of an unaligned access the alignment check would trigger before 1135e9ee9efcSDavid Miller * the one we are interested in. 1136e9ee9efcSDavid Miller */ 1137e9ee9efcSDavid Miller #define BPF_F_ANY_ALIGNMENT (1U << 1) 1138e9ee9efcSDavid Miller 1139c240eff6SJiong Wang /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1140c240eff6SJiong Wang * Verifier does sub-register def/use analysis and identifies instructions whose 1141c240eff6SJiong Wang * def only matters for low 32-bit, high 32-bit is never referenced later 1142c240eff6SJiong Wang * through implicit zero extension. Therefore verifier notifies JIT back-ends 1143c240eff6SJiong Wang * that it is safe to ignore clearing high 32-bit for these instructions. This 1144c240eff6SJiong Wang * saves some back-ends a lot of code-gen. However such optimization is not 1145c240eff6SJiong Wang * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1146c240eff6SJiong Wang * hence hasn't used verifier's analysis result. But, we really want to have a 1147c240eff6SJiong Wang * way to be able to verify the correctness of the described optimization on 1148c240eff6SJiong Wang * x86_64 on which testsuites are frequently exercised. 1149c240eff6SJiong Wang * 1150c240eff6SJiong Wang * So, this flag is introduced. Once it is set, verifier will randomize high 1151c240eff6SJiong Wang * 32-bit for those instructions who has been identified as safe to ignore them. 1152c240eff6SJiong Wang * Then, if verifier is not doing correct analysis, such randomization will 1153c240eff6SJiong Wang * regress tests to expose bugs. 1154c240eff6SJiong Wang */ 1155c240eff6SJiong Wang #define BPF_F_TEST_RND_HI32 (1U << 2) 1156c240eff6SJiong Wang 115710d274e8SAlexei Starovoitov /* The verifier internal test flag. Behavior is undefined */ 115810d274e8SAlexei Starovoitov #define BPF_F_TEST_STATE_FREQ (1U << 3) 115910d274e8SAlexei Starovoitov 11601e6c62a8SAlexei Starovoitov /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 11611e6c62a8SAlexei Starovoitov * restrict map and helper usage for such programs. Sleepable BPF programs can 11621e6c62a8SAlexei Starovoitov * only be attached to hooks where kernel execution context allows sleeping. 11631e6c62a8SAlexei Starovoitov * Such programs are allowed to use helpers that may sleep like 11641e6c62a8SAlexei Starovoitov * bpf_copy_from_user(). 11651e6c62a8SAlexei Starovoitov */ 11661e6c62a8SAlexei Starovoitov #define BPF_F_SLEEPABLE (1U << 4) 11671e6c62a8SAlexei Starovoitov 1168c2f2cdbeSLorenzo Bianconi /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1169c2f2cdbeSLorenzo Bianconi * fully support xdp frags. 1170c2f2cdbeSLorenzo Bianconi */ 1171c2f2cdbeSLorenzo Bianconi #define BPF_F_XDP_HAS_FRAGS (1U << 5) 1172c2f2cdbeSLorenzo Bianconi 11732b3486bcSStanislav Fomichev /* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded 11742b3486bcSStanislav Fomichev * program becomes device-bound but can access XDP metadata. 11752b3486bcSStanislav Fomichev */ 11762b3486bcSStanislav Fomichev #define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6) 11772b3486bcSStanislav Fomichev 11780dcac272SJiri Olsa /* link_create.kprobe_multi.flags used in LINK_CREATE command for 11790dcac272SJiri Olsa * BPF_TRACE_KPROBE_MULTI attach type to create return probe. 11800dcac272SJiri Olsa */ 11810dcac272SJiri Olsa #define BPF_F_KPROBE_MULTI_RETURN (1U << 0) 11820dcac272SJiri Olsa 1183d8eca5bbSDaniel Borkmann /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 11844976b718SHao Luo * the following extensions: 1185d8eca5bbSDaniel Borkmann * 1186387544bfSAlexei Starovoitov * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1187387544bfSAlexei Starovoitov * insn[0].imm: map fd or fd_idx 11884976b718SHao Luo * insn[1].imm: 0 11894976b718SHao Luo * insn[0].off: 0 11904976b718SHao Luo * insn[1].off: 0 11914976b718SHao Luo * ldimm64 rewrite: address of map 11924976b718SHao Luo * verifier type: CONST_PTR_TO_MAP 1193d8eca5bbSDaniel Borkmann */ 1194f1a66f85SDaniel Borkmann #define BPF_PSEUDO_MAP_FD 1 1195387544bfSAlexei Starovoitov #define BPF_PSEUDO_MAP_IDX 5 1196387544bfSAlexei Starovoitov 1197387544bfSAlexei Starovoitov /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1198387544bfSAlexei Starovoitov * insn[0].imm: map fd or fd_idx 11994976b718SHao Luo * insn[1].imm: offset into value 12004976b718SHao Luo * insn[0].off: 0 12014976b718SHao Luo * insn[1].off: 0 12024976b718SHao Luo * ldimm64 rewrite: address of map[0]+offset 12034976b718SHao Luo * verifier type: PTR_TO_MAP_VALUE 12044976b718SHao Luo */ 1205d8eca5bbSDaniel Borkmann #define BPF_PSEUDO_MAP_VALUE 2 1206387544bfSAlexei Starovoitov #define BPF_PSEUDO_MAP_IDX_VALUE 6 1207387544bfSAlexei Starovoitov 12084976b718SHao Luo /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 12094976b718SHao Luo * insn[0].imm: kernel btd id of VAR 12104976b718SHao Luo * insn[1].imm: 0 12114976b718SHao Luo * insn[0].off: 0 12124976b718SHao Luo * insn[1].off: 0 12134976b718SHao Luo * ldimm64 rewrite: address of the kernel variable 12144976b718SHao Luo * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 12154976b718SHao Luo * is struct/union. 12164976b718SHao Luo */ 12174976b718SHao Luo #define BPF_PSEUDO_BTF_ID 3 121869c087baSYonghong Song /* insn[0].src_reg: BPF_PSEUDO_FUNC 121969c087baSYonghong Song * insn[0].imm: insn offset to the func 122069c087baSYonghong Song * insn[1].imm: 0 122169c087baSYonghong Song * insn[0].off: 0 122269c087baSYonghong Song * insn[1].off: 0 122369c087baSYonghong Song * ldimm64 rewrite: address of the function 122469c087baSYonghong Song * verifier type: PTR_TO_FUNC. 122569c087baSYonghong Song */ 122669c087baSYonghong Song #define BPF_PSEUDO_FUNC 4 1227f1a66f85SDaniel Borkmann 1228cc8b0b92SAlexei Starovoitov /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1229cc8b0b92SAlexei Starovoitov * offset to another bpf function 1230cc8b0b92SAlexei Starovoitov */ 1231cc8b0b92SAlexei Starovoitov #define BPF_PSEUDO_CALL 1 1232e6ac2450SMartin KaFai Lau /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1233e6ac2450SMartin KaFai Lau * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1234e6ac2450SMartin KaFai Lau */ 1235e6ac2450SMartin KaFai Lau #define BPF_PSEUDO_KFUNC_CALL 2 1236cc8b0b92SAlexei Starovoitov 12373274f520SAlexei Starovoitov /* flags for BPF_MAP_UPDATE_ELEM command */ 12381aae4bddSAndrii Nakryiko enum { 12391aae4bddSAndrii Nakryiko BPF_ANY = 0, /* create new element or update existing */ 12401aae4bddSAndrii Nakryiko BPF_NOEXIST = 1, /* create new element if it didn't exist */ 12411aae4bddSAndrii Nakryiko BPF_EXIST = 2, /* update existing element */ 12421aae4bddSAndrii Nakryiko BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 12431aae4bddSAndrii Nakryiko }; 12443274f520SAlexei Starovoitov 124596eabe7aSMartin KaFai Lau /* flags for BPF_MAP_CREATE command */ 12461aae4bddSAndrii Nakryiko enum { 12471aae4bddSAndrii Nakryiko BPF_F_NO_PREALLOC = (1U << 0), 124829ba732aSMartin KaFai Lau /* Instead of having one common LRU list in the 12498f844938SMartin KaFai Lau * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 125029ba732aSMartin KaFai Lau * which can scale and perform better. 125129ba732aSMartin KaFai Lau * Note, the LRU nodes (including free nodes) cannot be moved 125229ba732aSMartin KaFai Lau * across different LRU lists. 125329ba732aSMartin KaFai Lau */ 12541aae4bddSAndrii Nakryiko BPF_F_NO_COMMON_LRU = (1U << 1), 125596eabe7aSMartin KaFai Lau /* Specify numa node during map creation */ 12561aae4bddSAndrii Nakryiko BPF_F_NUMA_NODE = (1U << 2), 1257cb4d2b3fSMartin KaFai Lau 1258591fe988SDaniel Borkmann /* Flags for accessing BPF object from syscall side. */ 12591aae4bddSAndrii Nakryiko BPF_F_RDONLY = (1U << 3), 12601aae4bddSAndrii Nakryiko BPF_F_WRONLY = (1U << 4), 12616e71b04aSChenbo Feng 1262615755a7SSong Liu /* Flag for stack_map, store build_id+offset instead of pointer */ 12631aae4bddSAndrii Nakryiko BPF_F_STACK_BUILD_ID = (1U << 5), 1264615755a7SSong Liu 126596b3b6c9SLorenz Bauer /* Zero-initialize hash function seed. This should only be used for testing. */ 12661aae4bddSAndrii Nakryiko BPF_F_ZERO_SEED = (1U << 6), 126796b3b6c9SLorenz Bauer 1268591fe988SDaniel Borkmann /* Flags for accessing BPF object from program side. */ 12691aae4bddSAndrii Nakryiko BPF_F_RDONLY_PROG = (1U << 7), 12701aae4bddSAndrii Nakryiko BPF_F_WRONLY_PROG = (1U << 8), 1271591fe988SDaniel Borkmann 12728f51dfc7SStanislav Fomichev /* Clone map from listener for newly accepted socket */ 12731aae4bddSAndrii Nakryiko BPF_F_CLONE = (1U << 9), 12748f51dfc7SStanislav Fomichev 1275fc970227SAndrii Nakryiko /* Enable memory-mapping BPF map */ 12761aae4bddSAndrii Nakryiko BPF_F_MMAPABLE = (1U << 10), 1277792cacccSSong Liu 1278792cacccSSong Liu /* Share perf_event among processes */ 1279792cacccSSong Liu BPF_F_PRESERVE_ELEMS = (1U << 11), 12804a8f87e6SDaniel Borkmann 12814a8f87e6SDaniel Borkmann /* Create a map that is suitable to be an inner map with dynamic max entries */ 12824a8f87e6SDaniel Borkmann BPF_F_INNER_MAP = (1U << 12), 128368b04864SKui-Feng Lee 128468b04864SKui-Feng Lee /* Create a map that will be registered/unregesitered by the backed bpf_link */ 128568b04864SKui-Feng Lee BPF_F_LINK = (1U << 13), 1286cb8edce2SAndrii Nakryiko 1287cb8edce2SAndrii Nakryiko /* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */ 1288cb8edce2SAndrii Nakryiko BPF_F_PATH_FD = (1U << 14), 12891aae4bddSAndrii Nakryiko }; 1290fc970227SAndrii Nakryiko 1291f5bfcd95SAndrey Ignatov /* Flags for BPF_PROG_QUERY. */ 1292f5bfcd95SAndrey Ignatov 1293f5bfcd95SAndrey Ignatov /* Query effective (directly attached + inherited from ancestor cgroups) 1294f5bfcd95SAndrey Ignatov * programs that will be executed for events within a cgroup. 12950e426a3aSPu Lehui * attach_flags with this flag are always returned 0. 1296f5bfcd95SAndrey Ignatov */ 12972f183360SLorenz Bauer #define BPF_F_QUERY_EFFECTIVE (1U << 0) 12982f183360SLorenz Bauer 12991b4d60ecSSong Liu /* Flags for BPF_PROG_TEST_RUN */ 13001b4d60ecSSong Liu 13011b4d60ecSSong Liu /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 13021b4d60ecSSong Liu #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1303b530e9e1SToke Høiland-Jørgensen /* If set, XDP frames will be transmitted after processing */ 1304b530e9e1SToke Høiland-Jørgensen #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 13051b4d60ecSSong Liu 1306d46edd67SSong Liu /* type for BPF_ENABLE_STATS */ 1307d46edd67SSong Liu enum bpf_stats_type { 1308d46edd67SSong Liu /* enabled run_time_ns and run_cnt */ 1309d46edd67SSong Liu BPF_STATS_RUN_TIME = 0, 1310d46edd67SSong Liu }; 1311d46edd67SSong Liu 1312615755a7SSong Liu enum bpf_stack_build_id_status { 1313615755a7SSong Liu /* user space need an empty entry to identify end of a trace */ 1314615755a7SSong Liu BPF_STACK_BUILD_ID_EMPTY = 0, 1315615755a7SSong Liu /* with valid build_id and offset */ 1316615755a7SSong Liu BPF_STACK_BUILD_ID_VALID = 1, 1317615755a7SSong Liu /* couldn't get build_id, fallback to ip */ 1318615755a7SSong Liu BPF_STACK_BUILD_ID_IP = 2, 1319615755a7SSong Liu }; 1320615755a7SSong Liu 1321615755a7SSong Liu #define BPF_BUILD_ID_SIZE 20 1322615755a7SSong Liu struct bpf_stack_build_id { 1323615755a7SSong Liu __s32 status; 1324615755a7SSong Liu unsigned char build_id[BPF_BUILD_ID_SIZE]; 1325615755a7SSong Liu union { 1326615755a7SSong Liu __u64 offset; 1327615755a7SSong Liu __u64 ip; 1328615755a7SSong Liu }; 1329615755a7SSong Liu }; 1330615755a7SSong Liu 13311aae4bddSAndrii Nakryiko #define BPF_OBJ_NAME_LEN 16U 13321aae4bddSAndrii Nakryiko 133399c55f7dSAlexei Starovoitov union bpf_attr { 133499c55f7dSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_CREATE command */ 133599c55f7dSAlexei Starovoitov __u32 map_type; /* one of enum bpf_map_type */ 133699c55f7dSAlexei Starovoitov __u32 key_size; /* size of key in bytes */ 133799c55f7dSAlexei Starovoitov __u32 value_size; /* size of value in bytes */ 133899c55f7dSAlexei Starovoitov __u32 max_entries; /* max number of entries in a map */ 133996eabe7aSMartin KaFai Lau __u32 map_flags; /* BPF_MAP_CREATE related 134096eabe7aSMartin KaFai Lau * flags defined above. 134196eabe7aSMartin KaFai Lau */ 134256f668dfSMartin KaFai Lau __u32 inner_map_fd; /* fd pointing to the inner map */ 134396eabe7aSMartin KaFai Lau __u32 numa_node; /* numa node (effective only if 134496eabe7aSMartin KaFai Lau * BPF_F_NUMA_NODE is set). 134596eabe7aSMartin KaFai Lau */ 1346067cae47SMartin KaFai Lau char map_name[BPF_OBJ_NAME_LEN]; 1347a3884572SJakub Kicinski __u32 map_ifindex; /* ifindex of netdev to create on */ 1348a26ca7c9SMartin KaFai Lau __u32 btf_fd; /* fd pointing to a BTF type data */ 13499b2cf328SMartin KaFai Lau __u32 btf_key_type_id; /* BTF type_id of the key */ 13509b2cf328SMartin KaFai Lau __u32 btf_value_type_id; /* BTF type_id of the value */ 135185d33df3SMartin KaFai Lau __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 135285d33df3SMartin KaFai Lau * struct stored as the 135385d33df3SMartin KaFai Lau * map value 135485d33df3SMartin KaFai Lau */ 13559330986cSJoanne Koong /* Any per-map-type extra fields 13569330986cSJoanne Koong * 13579330986cSJoanne Koong * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 13589330986cSJoanne Koong * number of hash functions (if 0, the bloom filter will default 13599330986cSJoanne Koong * to using 5 hash functions). 13609330986cSJoanne Koong */ 13619330986cSJoanne Koong __u64 map_extra; 136299c55f7dSAlexei Starovoitov }; 1363db20fd2bSAlexei Starovoitov 1364db20fd2bSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1365db20fd2bSAlexei Starovoitov __u32 map_fd; 1366db20fd2bSAlexei Starovoitov __aligned_u64 key; 1367db20fd2bSAlexei Starovoitov union { 1368db20fd2bSAlexei Starovoitov __aligned_u64 value; 1369db20fd2bSAlexei Starovoitov __aligned_u64 next_key; 1370db20fd2bSAlexei Starovoitov }; 13713274f520SAlexei Starovoitov __u64 flags; 1372db20fd2bSAlexei Starovoitov }; 137309756af4SAlexei Starovoitov 1374cb4d03abSBrian Vazquez struct { /* struct used by BPF_MAP_*_BATCH commands */ 1375cb4d03abSBrian Vazquez __aligned_u64 in_batch; /* start batch, 1376cb4d03abSBrian Vazquez * NULL to start from beginning 1377cb4d03abSBrian Vazquez */ 1378cb4d03abSBrian Vazquez __aligned_u64 out_batch; /* output: next start batch */ 1379cb4d03abSBrian Vazquez __aligned_u64 keys; 1380cb4d03abSBrian Vazquez __aligned_u64 values; 1381cb4d03abSBrian Vazquez __u32 count; /* input/output: 1382cb4d03abSBrian Vazquez * input: # of key/value 1383cb4d03abSBrian Vazquez * elements 1384cb4d03abSBrian Vazquez * output: # of filled elements 1385cb4d03abSBrian Vazquez */ 1386cb4d03abSBrian Vazquez __u32 map_fd; 1387cb4d03abSBrian Vazquez __u64 elem_flags; 1388cb4d03abSBrian Vazquez __u64 flags; 1389cb4d03abSBrian Vazquez } batch; 1390cb4d03abSBrian Vazquez 139109756af4SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_LOAD command */ 139209756af4SAlexei Starovoitov __u32 prog_type; /* one of enum bpf_prog_type */ 139309756af4SAlexei Starovoitov __u32 insn_cnt; 139409756af4SAlexei Starovoitov __aligned_u64 insns; 139509756af4SAlexei Starovoitov __aligned_u64 license; 1396cbd35700SAlexei Starovoitov __u32 log_level; /* verbosity level of verifier */ 1397cbd35700SAlexei Starovoitov __u32 log_size; /* size of user buffer */ 1398cbd35700SAlexei Starovoitov __aligned_u64 log_buf; /* user supplied buffer */ 13996c4fc209SDaniel Borkmann __u32 kern_version; /* not used */ 1400e07b98d9SDavid S. Miller __u32 prog_flags; 1401067cae47SMartin KaFai Lau char prog_name[BPF_OBJ_NAME_LEN]; 14021f6f4cb7SJakub Kicinski __u32 prog_ifindex; /* ifindex of netdev to prep for */ 14035e43f899SAndrey Ignatov /* For some prog types expected attach type must be known at 14045e43f899SAndrey Ignatov * load time to verify attach type specific parts of prog 14055e43f899SAndrey Ignatov * (context accesses, allowed helpers, etc). 14065e43f899SAndrey Ignatov */ 14075e43f899SAndrey Ignatov __u32 expected_attach_type; 1408838e9690SYonghong Song __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1409838e9690SYonghong Song __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1410838e9690SYonghong Song __aligned_u64 func_info; /* func info */ 1411838e9690SYonghong Song __u32 func_info_cnt; /* number of bpf_func_info records */ 1412c454a46bSMartin KaFai Lau __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1413c454a46bSMartin KaFai Lau __aligned_u64 line_info; /* line info */ 1414c454a46bSMartin KaFai Lau __u32 line_info_cnt; /* number of bpf_line_info records */ 1415ccfe29ebSAlexei Starovoitov __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1416290248a5SAndrii Nakryiko union { 1417290248a5SAndrii Nakryiko /* valid prog_fd to attach to bpf prog */ 1418290248a5SAndrii Nakryiko __u32 attach_prog_fd; 1419290248a5SAndrii Nakryiko /* or valid module BTF object fd or 0 to attach to vmlinux */ 1420290248a5SAndrii Nakryiko __u32 attach_btf_obj_fd; 1421290248a5SAndrii Nakryiko }; 1422fbd94c7aSAlexei Starovoitov __u32 core_relo_cnt; /* number of bpf_core_relo */ 1423387544bfSAlexei Starovoitov __aligned_u64 fd_array; /* array of FDs */ 1424fbd94c7aSAlexei Starovoitov __aligned_u64 core_relos; 1425fbd94c7aSAlexei Starovoitov __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 142647a71c1fSAndrii Nakryiko /* output: actual total log contents size (including termintaing zero). 142747a71c1fSAndrii Nakryiko * It could be both larger than original log_size (if log was 142847a71c1fSAndrii Nakryiko * truncated), or smaller (if log buffer wasn't filled completely). 142947a71c1fSAndrii Nakryiko */ 143047a71c1fSAndrii Nakryiko __u32 log_true_size; 143109756af4SAlexei Starovoitov }; 1432b2197755SDaniel Borkmann 1433b2197755SDaniel Borkmann struct { /* anonymous struct used by BPF_OBJ_* commands */ 1434b2197755SDaniel Borkmann __aligned_u64 pathname; 1435b2197755SDaniel Borkmann __u32 bpf_fd; 14366e71b04aSChenbo Feng __u32 file_flags; 1437cb8edce2SAndrii Nakryiko /* Same as dirfd in openat() syscall; see openat(2) 1438cb8edce2SAndrii Nakryiko * manpage for details of path FD and pathname semantics; 1439cb8edce2SAndrii Nakryiko * path_fd should accompanied by BPF_F_PATH_FD flag set in 1440cb8edce2SAndrii Nakryiko * file_flags field, otherwise it should be set to zero; 1441cb8edce2SAndrii Nakryiko * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed. 1442cb8edce2SAndrii Nakryiko */ 1443cb8edce2SAndrii Nakryiko __s32 path_fd; 1444b2197755SDaniel Borkmann }; 1445f4324551SDaniel Mack 1446f4324551SDaniel Mack struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1447f4324551SDaniel Mack __u32 target_fd; /* container object to attach to */ 1448f4324551SDaniel Mack __u32 attach_bpf_fd; /* eBPF program to attach */ 1449f4324551SDaniel Mack __u32 attach_type; 14507f677633SAlexei Starovoitov __u32 attach_flags; 14517dd68b32SAndrey Ignatov __u32 replace_bpf_fd; /* previously attached eBPF 14527dd68b32SAndrey Ignatov * program to replace if 14537dd68b32SAndrey Ignatov * BPF_F_REPLACE is used 14547dd68b32SAndrey Ignatov */ 1455f4324551SDaniel Mack }; 14561cf1cae9SAlexei Starovoitov 14571cf1cae9SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 14581cf1cae9SAlexei Starovoitov __u32 prog_fd; 14591cf1cae9SAlexei Starovoitov __u32 retval; 1460b5a36b1eSLorenz Bauer __u32 data_size_in; /* input: len of data_in */ 1461b5a36b1eSLorenz Bauer __u32 data_size_out; /* input/output: len of data_out 1462b5a36b1eSLorenz Bauer * returns ENOSPC if data_out 1463b5a36b1eSLorenz Bauer * is too small. 1464b5a36b1eSLorenz Bauer */ 14651cf1cae9SAlexei Starovoitov __aligned_u64 data_in; 14661cf1cae9SAlexei Starovoitov __aligned_u64 data_out; 14671cf1cae9SAlexei Starovoitov __u32 repeat; 14681cf1cae9SAlexei Starovoitov __u32 duration; 1469b0b9395dSStanislav Fomichev __u32 ctx_size_in; /* input: len of ctx_in */ 1470b0b9395dSStanislav Fomichev __u32 ctx_size_out; /* input/output: len of ctx_out 1471b0b9395dSStanislav Fomichev * returns ENOSPC if ctx_out 1472b0b9395dSStanislav Fomichev * is too small. 1473b0b9395dSStanislav Fomichev */ 1474b0b9395dSStanislav Fomichev __aligned_u64 ctx_in; 1475b0b9395dSStanislav Fomichev __aligned_u64 ctx_out; 14761b4d60ecSSong Liu __u32 flags; 14771b4d60ecSSong Liu __u32 cpu; 1478b530e9e1SToke Høiland-Jørgensen __u32 batch_size; 14791cf1cae9SAlexei Starovoitov } test; 148034ad5580SMartin KaFai Lau 1481b16d9aa4SMartin KaFai Lau struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1482b16d9aa4SMartin KaFai Lau union { 148334ad5580SMartin KaFai Lau __u32 start_id; 1484b16d9aa4SMartin KaFai Lau __u32 prog_id; 1485bd5f5f4eSMartin KaFai Lau __u32 map_id; 148678958fcaSMartin KaFai Lau __u32 btf_id; 1487a3b80e10SAndrii Nakryiko __u32 link_id; 1488b16d9aa4SMartin KaFai Lau }; 148934ad5580SMartin KaFai Lau __u32 next_id; 14906e71b04aSChenbo Feng __u32 open_flags; 149134ad5580SMartin KaFai Lau }; 14921e270976SMartin KaFai Lau 14931e270976SMartin KaFai Lau struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 14941e270976SMartin KaFai Lau __u32 bpf_fd; 14951e270976SMartin KaFai Lau __u32 info_len; 14961e270976SMartin KaFai Lau __aligned_u64 info; 14971e270976SMartin KaFai Lau } info; 1498468e2f64SAlexei Starovoitov 1499468e2f64SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1500468e2f64SAlexei Starovoitov __u32 target_fd; /* container object to query */ 1501468e2f64SAlexei Starovoitov __u32 attach_type; 1502468e2f64SAlexei Starovoitov __u32 query_flags; 1503468e2f64SAlexei Starovoitov __u32 attach_flags; 1504468e2f64SAlexei Starovoitov __aligned_u64 prog_ids; 1505468e2f64SAlexei Starovoitov __u32 prog_cnt; 15060e426a3aSPu Lehui /* output: per-program attach_flags. 15070e426a3aSPu Lehui * not allowed to be set during effective query. 15080e426a3aSPu Lehui */ 15090e426a3aSPu Lehui __aligned_u64 prog_attach_flags; 1510468e2f64SAlexei Starovoitov } query; 1511c4f6699dSAlexei Starovoitov 1512af6eea57SAndrii Nakryiko struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1513c4f6699dSAlexei Starovoitov __u64 name; 1514c4f6699dSAlexei Starovoitov __u32 prog_fd; 1515c4f6699dSAlexei Starovoitov } raw_tracepoint; 1516f56a653cSMartin KaFai Lau 1517f56a653cSMartin KaFai Lau struct { /* anonymous struct for BPF_BTF_LOAD */ 1518f56a653cSMartin KaFai Lau __aligned_u64 btf; 1519f56a653cSMartin KaFai Lau __aligned_u64 btf_log_buf; 1520f56a653cSMartin KaFai Lau __u32 btf_size; 1521f56a653cSMartin KaFai Lau __u32 btf_log_size; 1522f56a653cSMartin KaFai Lau __u32 btf_log_level; 152347a71c1fSAndrii Nakryiko /* output: actual total log contents size (including termintaing zero). 152447a71c1fSAndrii Nakryiko * It could be both larger than original log_size (if log was 152547a71c1fSAndrii Nakryiko * truncated), or smaller (if log buffer wasn't filled completely). 152647a71c1fSAndrii Nakryiko */ 152747a71c1fSAndrii Nakryiko __u32 btf_log_true_size; 1528f56a653cSMartin KaFai Lau }; 152941bdc4b4SYonghong Song 153041bdc4b4SYonghong Song struct { 153141bdc4b4SYonghong Song __u32 pid; /* input: pid */ 153241bdc4b4SYonghong Song __u32 fd; /* input: fd */ 153341bdc4b4SYonghong Song __u32 flags; /* input: flags */ 153441bdc4b4SYonghong Song __u32 buf_len; /* input/output: buf len */ 153541bdc4b4SYonghong Song __aligned_u64 buf; /* input/output: 153641bdc4b4SYonghong Song * tp_name for tracepoint 153741bdc4b4SYonghong Song * symbol for kprobe 153841bdc4b4SYonghong Song * filename for uprobe 153941bdc4b4SYonghong Song */ 154041bdc4b4SYonghong Song __u32 prog_id; /* output: prod_id */ 154141bdc4b4SYonghong Song __u32 fd_type; /* output: BPF_FD_TYPE_* */ 154241bdc4b4SYonghong Song __u64 probe_offset; /* output: probe_offset */ 154341bdc4b4SYonghong Song __u64 probe_addr; /* output: probe_addr */ 154441bdc4b4SYonghong Song } task_fd_query; 1545af6eea57SAndrii Nakryiko 1546af6eea57SAndrii Nakryiko struct { /* struct used by BPF_LINK_CREATE command */ 154768b04864SKui-Feng Lee union { 1548af6eea57SAndrii Nakryiko __u32 prog_fd; /* eBPF program to attach */ 154968b04864SKui-Feng Lee __u32 map_fd; /* struct_ops to attach */ 155068b04864SKui-Feng Lee }; 1551aa8d3a71SAndrii Nakryiko union { 1552af6eea57SAndrii Nakryiko __u32 target_fd; /* object to attach to */ 1553aa8d3a71SAndrii Nakryiko __u32 target_ifindex; /* target ifindex */ 1554aa8d3a71SAndrii Nakryiko }; 1555af6eea57SAndrii Nakryiko __u32 attach_type; /* attach type */ 1556af6eea57SAndrii Nakryiko __u32 flags; /* extra flags */ 15574a1e7c0cSToke Høiland-Jørgensen union { 15584a1e7c0cSToke Høiland-Jørgensen __u32 target_btf_id; /* btf_id of target to attach to */ 15594a1e7c0cSToke Høiland-Jørgensen struct { 15605e7b3020SYonghong Song __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 15615e7b3020SYonghong Song __u32 iter_info_len; /* iter_info length */ 15624a1e7c0cSToke Høiland-Jørgensen }; 156382e6b1eeSAndrii Nakryiko struct { 156482e6b1eeSAndrii Nakryiko /* black box user-provided value passed through 156582e6b1eeSAndrii Nakryiko * to BPF program at the execution time and 156682e6b1eeSAndrii Nakryiko * accessible through bpf_get_attach_cookie() BPF helper 156782e6b1eeSAndrii Nakryiko */ 156882e6b1eeSAndrii Nakryiko __u64 bpf_cookie; 156982e6b1eeSAndrii Nakryiko } perf_event; 15700dcac272SJiri Olsa struct { 15710dcac272SJiri Olsa __u32 flags; 15720dcac272SJiri Olsa __u32 cnt; 15730dcac272SJiri Olsa __aligned_u64 syms; 15740dcac272SJiri Olsa __aligned_u64 addrs; 1575ca74823cSJiri Olsa __aligned_u64 cookies; 15760dcac272SJiri Olsa } kprobe_multi; 15772fcc8241SKui-Feng Lee struct { 15782fcc8241SKui-Feng Lee /* this is overlaid with the target_btf_id above. */ 15792fcc8241SKui-Feng Lee __u32 target_btf_id; 15802fcc8241SKui-Feng Lee /* black box user-provided value passed through 15812fcc8241SKui-Feng Lee * to BPF program at the execution time and 15822fcc8241SKui-Feng Lee * accessible through bpf_get_attach_cookie() BPF helper 15832fcc8241SKui-Feng Lee */ 15842fcc8241SKui-Feng Lee __u64 cookie; 15852fcc8241SKui-Feng Lee } tracing; 158684601d6eSFlorian Westphal struct { 158784601d6eSFlorian Westphal __u32 pf; 158884601d6eSFlorian Westphal __u32 hooknum; 158984601d6eSFlorian Westphal __s32 priority; 159084601d6eSFlorian Westphal __u32 flags; 159184601d6eSFlorian Westphal } netfilter; 15924a1e7c0cSToke Høiland-Jørgensen }; 1593af6eea57SAndrii Nakryiko } link_create; 15940c991ebcSAndrii Nakryiko 15950c991ebcSAndrii Nakryiko struct { /* struct used by BPF_LINK_UPDATE command */ 15960c991ebcSAndrii Nakryiko __u32 link_fd; /* link fd */ 1597aef56f2eSKui-Feng Lee union { 15980c991ebcSAndrii Nakryiko /* new program fd to update link with */ 15990c991ebcSAndrii Nakryiko __u32 new_prog_fd; 1600aef56f2eSKui-Feng Lee /* new struct_ops map fd to update link with */ 1601aef56f2eSKui-Feng Lee __u32 new_map_fd; 1602aef56f2eSKui-Feng Lee }; 16030c991ebcSAndrii Nakryiko __u32 flags; /* extra flags */ 1604aef56f2eSKui-Feng Lee union { 16050c991ebcSAndrii Nakryiko /* expected link's program fd; is specified only if 1606aef56f2eSKui-Feng Lee * BPF_F_REPLACE flag is set in flags. 1607aef56f2eSKui-Feng Lee */ 16080c991ebcSAndrii Nakryiko __u32 old_prog_fd; 1609aef56f2eSKui-Feng Lee /* expected link's map fd; is specified only 1610aef56f2eSKui-Feng Lee * if BPF_F_REPLACE flag is set. 1611aef56f2eSKui-Feng Lee */ 1612aef56f2eSKui-Feng Lee __u32 old_map_fd; 1613aef56f2eSKui-Feng Lee }; 16140c991ebcSAndrii Nakryiko } link_update; 16150c991ebcSAndrii Nakryiko 161673b11c2aSAndrii Nakryiko struct { 161773b11c2aSAndrii Nakryiko __u32 link_fd; 161873b11c2aSAndrii Nakryiko } link_detach; 161973b11c2aSAndrii Nakryiko 1620d46edd67SSong Liu struct { /* struct used by BPF_ENABLE_STATS command */ 1621d46edd67SSong Liu __u32 type; 1622d46edd67SSong Liu } enable_stats; 1623d46edd67SSong Liu 1624ac51d99bSYonghong Song struct { /* struct used by BPF_ITER_CREATE command */ 1625ac51d99bSYonghong Song __u32 link_fd; 1626ac51d99bSYonghong Song __u32 flags; 1627ac51d99bSYonghong Song } iter_create; 1628ac51d99bSYonghong Song 1629ef15314aSYiFei Zhu struct { /* struct used by BPF_PROG_BIND_MAP command */ 1630ef15314aSYiFei Zhu __u32 prog_fd; 1631ef15314aSYiFei Zhu __u32 map_fd; 1632ef15314aSYiFei Zhu __u32 flags; /* extra flags */ 1633ef15314aSYiFei Zhu } prog_bind_map; 1634ef15314aSYiFei Zhu 163599c55f7dSAlexei Starovoitov } __attribute__((aligned(8))); 163699c55f7dSAlexei Starovoitov 163756a092c8SQuentin Monnet /* The description below is an attempt at providing documentation to eBPF 163856a092c8SQuentin Monnet * developers about the multiple available eBPF helper functions. It can be 163956a092c8SQuentin Monnet * parsed and used to produce a manual page. The workflow is the following, 164056a092c8SQuentin Monnet * and requires the rst2man utility: 1641ebb676daSThomas Graf * 1642923a932cSJoe Stringer * $ ./scripts/bpf_doc.py \ 164356a092c8SQuentin Monnet * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 164456a092c8SQuentin Monnet * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 164556a092c8SQuentin Monnet * $ man /tmp/bpf-helpers.7 1646ebb676daSThomas Graf * 164756a092c8SQuentin Monnet * Note that in order to produce this external documentation, some RST 164856a092c8SQuentin Monnet * formatting is used in the descriptions to get "bold" and "italics" in 164956a092c8SQuentin Monnet * manual pages. Also note that the few trailing white spaces are 165056a092c8SQuentin Monnet * intentional, removing them would break paragraphs for rst2man. 1651ebb676daSThomas Graf * 165256a092c8SQuentin Monnet * Start of BPF helper function descriptions: 1653ad4a5223SQuentin Monnet * 1654ad4a5223SQuentin Monnet * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1655ad4a5223SQuentin Monnet * Description 1656ad4a5223SQuentin Monnet * Perform a lookup in *map* for an entry associated to *key*. 1657ad4a5223SQuentin Monnet * Return 1658ad4a5223SQuentin Monnet * Map value associated to *key*, or **NULL** if no entry was 1659ad4a5223SQuentin Monnet * found. 1660ad4a5223SQuentin Monnet * 1661bdb7b79bSAndrii Nakryiko * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1662ad4a5223SQuentin Monnet * Description 1663ad4a5223SQuentin Monnet * Add or update the value of the entry associated to *key* in 1664ad4a5223SQuentin Monnet * *map* with *value*. *flags* is one of: 1665ad4a5223SQuentin Monnet * 1666ad4a5223SQuentin Monnet * **BPF_NOEXIST** 1667ad4a5223SQuentin Monnet * The entry for *key* must not exist in the map. 1668ad4a5223SQuentin Monnet * **BPF_EXIST** 1669ad4a5223SQuentin Monnet * The entry for *key* must already exist in the map. 1670ad4a5223SQuentin Monnet * **BPF_ANY** 1671ad4a5223SQuentin Monnet * No condition on the existence of the entry for *key*. 1672ad4a5223SQuentin Monnet * 1673ad4a5223SQuentin Monnet * Flag value **BPF_NOEXIST** cannot be used for maps of types 1674ad4a5223SQuentin Monnet * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1675ad4a5223SQuentin Monnet * elements always exist), the helper would return an error. 1676ad4a5223SQuentin Monnet * Return 1677ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1678ad4a5223SQuentin Monnet * 1679bdb7b79bSAndrii Nakryiko * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1680ad4a5223SQuentin Monnet * Description 1681ad4a5223SQuentin Monnet * Delete entry with *key* from *map*. 1682ad4a5223SQuentin Monnet * Return 1683ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1684ad4a5223SQuentin Monnet * 1685bdb7b79bSAndrii Nakryiko * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1686ad4a5223SQuentin Monnet * Description 1687ad4a5223SQuentin Monnet * For tracing programs, safely attempt to read *size* bytes from 16886ae08ae3SDaniel Borkmann * kernel space address *unsafe_ptr* and store the data in *dst*. 16896ae08ae3SDaniel Borkmann * 1690ab8d7809SQuentin Monnet * Generally, use **bpf_probe_read_user**\ () or 1691ab8d7809SQuentin Monnet * **bpf_probe_read_kernel**\ () instead. 1692ad4a5223SQuentin Monnet * Return 1693ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1694ad4a5223SQuentin Monnet * 1695ad4a5223SQuentin Monnet * u64 bpf_ktime_get_ns(void) 1696ad4a5223SQuentin Monnet * Description 1697ad4a5223SQuentin Monnet * Return the time elapsed since system boot, in nanoseconds. 169871d19214SMaciej Żenczykowski * Does not include time the system was suspended. 1699ab8d7809SQuentin Monnet * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1700ad4a5223SQuentin Monnet * Return 1701ad4a5223SQuentin Monnet * Current *ktime*. 1702ad4a5223SQuentin Monnet * 1703bdb7b79bSAndrii Nakryiko * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1704ad4a5223SQuentin Monnet * Description 1705ad4a5223SQuentin Monnet * This helper is a "printk()-like" facility for debugging. It 1706ad4a5223SQuentin Monnet * prints a message defined by format *fmt* (of size *fmt_size*) 170727d7fdf0SRoss Zwisler * to file *\/sys/kernel/tracing/trace* from TraceFS, if 1708ad4a5223SQuentin Monnet * available. It can take up to three additional **u64** 1709ad4a5223SQuentin Monnet * arguments (as an eBPF helpers, the total number of arguments is 1710ad4a5223SQuentin Monnet * limited to five). 1711ad4a5223SQuentin Monnet * 1712ad4a5223SQuentin Monnet * Each time the helper is called, it appends a line to the trace. 171327d7fdf0SRoss Zwisler * Lines are discarded while *\/sys/kernel/tracing/trace* is 171427d7fdf0SRoss Zwisler * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this. 1715ad4a5223SQuentin Monnet * The format of the trace is customizable, and the exact output 1716ad4a5223SQuentin Monnet * one will get depends on the options set in 171727d7fdf0SRoss Zwisler * *\/sys/kernel/tracing/trace_options* (see also the 1718ad4a5223SQuentin Monnet * *README* file under the same directory). However, it usually 1719ad4a5223SQuentin Monnet * defaults to something like: 1720ad4a5223SQuentin Monnet * 1721ad4a5223SQuentin Monnet * :: 1722ad4a5223SQuentin Monnet * 1723ad4a5223SQuentin Monnet * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1724ad4a5223SQuentin Monnet * 1725ad4a5223SQuentin Monnet * In the above: 1726ad4a5223SQuentin Monnet * 1727ad4a5223SQuentin Monnet * * ``telnet`` is the name of the current task. 1728ad4a5223SQuentin Monnet * * ``470`` is the PID of the current task. 1729ad4a5223SQuentin Monnet * * ``001`` is the CPU number on which the task is 1730ad4a5223SQuentin Monnet * running. 1731ad4a5223SQuentin Monnet * * In ``.N..``, each character refers to a set of 1732ad4a5223SQuentin Monnet * options (whether irqs are enabled, scheduling 1733ad4a5223SQuentin Monnet * options, whether hard/softirqs are running, level of 1734ad4a5223SQuentin Monnet * preempt_disabled respectively). **N** means that 1735ad4a5223SQuentin Monnet * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1736ad4a5223SQuentin Monnet * are set. 1737ad4a5223SQuentin Monnet * * ``419421.045894`` is a timestamp. 1738ad4a5223SQuentin Monnet * * ``0x00000001`` is a fake value used by BPF for the 1739ad4a5223SQuentin Monnet * instruction pointer register. 1740ad4a5223SQuentin Monnet * * ``<formatted msg>`` is the message formatted with 1741ad4a5223SQuentin Monnet * *fmt*. 1742ad4a5223SQuentin Monnet * 1743ad4a5223SQuentin Monnet * The conversion specifiers supported by *fmt* are similar, but 1744ad4a5223SQuentin Monnet * more limited than for printk(). They are **%d**, **%i**, 1745ad4a5223SQuentin Monnet * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1746ad4a5223SQuentin Monnet * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1747ad4a5223SQuentin Monnet * of field, padding with zeroes, etc.) is available, and the 1748ad4a5223SQuentin Monnet * helper will return **-EINVAL** (but print nothing) if it 1749ad4a5223SQuentin Monnet * encounters an unknown specifier. 1750ad4a5223SQuentin Monnet * 1751ad4a5223SQuentin Monnet * Also, note that **bpf_trace_printk**\ () is slow, and should 1752ad4a5223SQuentin Monnet * only be used for debugging purposes. For this reason, a notice 1753b16fc097STobias Klauser * block (spanning several lines) is printed to kernel logs and 1754ad4a5223SQuentin Monnet * states that the helper should not be used "for production use" 1755ad4a5223SQuentin Monnet * the first time this helper is used (or more precisely, when 1756ad4a5223SQuentin Monnet * **trace_printk**\ () buffers are allocated). For passing values 1757ad4a5223SQuentin Monnet * to user space, perf events should be preferred. 1758ad4a5223SQuentin Monnet * Return 1759ad4a5223SQuentin Monnet * The number of bytes written to the buffer, or a negative error 1760ad4a5223SQuentin Monnet * in case of failure. 1761ad4a5223SQuentin Monnet * 17621fdd08beSQuentin Monnet * u32 bpf_get_prandom_u32(void) 17631fdd08beSQuentin Monnet * Description 17641fdd08beSQuentin Monnet * Get a pseudo-random number. 17651fdd08beSQuentin Monnet * 17661fdd08beSQuentin Monnet * From a security point of view, this helper uses its own 17671fdd08beSQuentin Monnet * pseudo-random internal state, and cannot be used to infer the 17681fdd08beSQuentin Monnet * seed of other random functions in the kernel. However, it is 17691fdd08beSQuentin Monnet * essential to note that the generator used by the helper is not 17701fdd08beSQuentin Monnet * cryptographically secure. 17711fdd08beSQuentin Monnet * Return 17721fdd08beSQuentin Monnet * A random 32-bit unsigned value. 17731fdd08beSQuentin Monnet * 17741fdd08beSQuentin Monnet * u32 bpf_get_smp_processor_id(void) 17751fdd08beSQuentin Monnet * Description 17761fdd08beSQuentin Monnet * Get the SMP (symmetric multiprocessing) processor id. Note that 177733656275SMatteo Croce * all programs run with migration disabled, which means that the 17781fdd08beSQuentin Monnet * SMP processor id is stable during all the execution of the 17791fdd08beSQuentin Monnet * program. 17801fdd08beSQuentin Monnet * Return 17811fdd08beSQuentin Monnet * The SMP id of the processor running the program. 17821fdd08beSQuentin Monnet * 1783bdb7b79bSAndrii Nakryiko * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1784ad4a5223SQuentin Monnet * Description 1785ad4a5223SQuentin Monnet * Store *len* bytes from address *from* into the packet 1786ad4a5223SQuentin Monnet * associated to *skb*, at *offset*. *flags* are a combination of 1787ad4a5223SQuentin Monnet * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1788ad4a5223SQuentin Monnet * checksum for the packet after storing the bytes) and 1789ad4a5223SQuentin Monnet * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1790ad4a5223SQuentin Monnet * **->swhash** and *skb*\ **->l4hash** to 0). 1791ad4a5223SQuentin Monnet * 179232e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1793ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1794ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1795ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1796ad4a5223SQuentin Monnet * direct packet access. 1797ad4a5223SQuentin Monnet * Return 1798ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1799ad4a5223SQuentin Monnet * 1800bdb7b79bSAndrii Nakryiko * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1801ad4a5223SQuentin Monnet * Description 1802ad4a5223SQuentin Monnet * Recompute the layer 3 (e.g. IP) checksum for the packet 1803ad4a5223SQuentin Monnet * associated to *skb*. Computation is incremental, so the helper 1804ad4a5223SQuentin Monnet * must know the former value of the header field that was 1805ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 1806ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored in *size*. 1807ad4a5223SQuentin Monnet * Alternatively, it is possible to store the difference between 1808ad4a5223SQuentin Monnet * the previous and the new values of the header field in *to*, by 1809ad4a5223SQuentin Monnet * setting *from* and *size* to 0. For both methods, *offset* 1810ad4a5223SQuentin Monnet * indicates the location of the IP checksum within the packet. 1811ad4a5223SQuentin Monnet * 1812ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 1813ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 1814ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 1815ad4a5223SQuentin Monnet * checksum to update. 1816ad4a5223SQuentin Monnet * 181732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1818ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1819ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1820ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1821ad4a5223SQuentin Monnet * direct packet access. 1822ad4a5223SQuentin Monnet * Return 1823ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1824ad4a5223SQuentin Monnet * 1825bdb7b79bSAndrii Nakryiko * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1826ad4a5223SQuentin Monnet * Description 1827ad4a5223SQuentin Monnet * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1828ad4a5223SQuentin Monnet * packet associated to *skb*. Computation is incremental, so the 1829ad4a5223SQuentin Monnet * helper must know the former value of the header field that was 1830ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 1831ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored on the lowest 1832ad4a5223SQuentin Monnet * four bits of *flags*. Alternatively, it is possible to store 1833ad4a5223SQuentin Monnet * the difference between the previous and the new values of the 1834ad4a5223SQuentin Monnet * header field in *to*, by setting *from* and the four lowest 1835ad4a5223SQuentin Monnet * bits of *flags* to 0. For both methods, *offset* indicates the 1836ad4a5223SQuentin Monnet * location of the IP checksum within the packet. In addition to 1837ad4a5223SQuentin Monnet * the size of the field, *flags* can be added (bitwise OR) actual 1838ad4a5223SQuentin Monnet * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1839ad4a5223SQuentin Monnet * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1840ad4a5223SQuentin Monnet * for updates resulting in a null checksum the value is set to 1841ad4a5223SQuentin Monnet * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1842ad4a5223SQuentin Monnet * the checksum is to be computed against a pseudo-header. 1843ad4a5223SQuentin Monnet * 1844ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 1845ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 1846ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 1847ad4a5223SQuentin Monnet * checksum to update. 1848ad4a5223SQuentin Monnet * 184932e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1850ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1851ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1852ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1853ad4a5223SQuentin Monnet * direct packet access. 1854ad4a5223SQuentin Monnet * Return 1855ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1856ad4a5223SQuentin Monnet * 1857bdb7b79bSAndrii Nakryiko * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 1858ad4a5223SQuentin Monnet * Description 1859ad4a5223SQuentin Monnet * This special helper is used to trigger a "tail call", or in 1860ad4a5223SQuentin Monnet * other words, to jump into another eBPF program. The same stack 1861ad4a5223SQuentin Monnet * frame is used (but values on stack and in registers for the 1862ad4a5223SQuentin Monnet * caller are not accessible to the callee). This mechanism allows 1863ad4a5223SQuentin Monnet * for program chaining, either for raising the maximum number of 1864ad4a5223SQuentin Monnet * available eBPF instructions, or to execute given programs in 1865ad4a5223SQuentin Monnet * conditional blocks. For security reasons, there is an upper 1866ad4a5223SQuentin Monnet * limit to the number of successive tail calls that can be 1867ad4a5223SQuentin Monnet * performed. 1868ad4a5223SQuentin Monnet * 1869ad4a5223SQuentin Monnet * Upon call of this helper, the program attempts to jump into a 1870ad4a5223SQuentin Monnet * program referenced at index *index* in *prog_array_map*, a 1871ad4a5223SQuentin Monnet * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 1872ad4a5223SQuentin Monnet * *ctx*, a pointer to the context. 1873ad4a5223SQuentin Monnet * 1874ad4a5223SQuentin Monnet * If the call succeeds, the kernel immediately runs the first 1875ad4a5223SQuentin Monnet * instruction of the new program. This is not a function call, 1876ad4a5223SQuentin Monnet * and it never returns to the previous program. If the call 1877ad4a5223SQuentin Monnet * fails, then the helper has no effect, and the caller continues 1878ad4a5223SQuentin Monnet * to run its subsequent instructions. A call can fail if the 1879ad4a5223SQuentin Monnet * destination program for the jump does not exist (i.e. *index* 1880ad4a5223SQuentin Monnet * is superior to the number of entries in *prog_array_map*), or 1881ad4a5223SQuentin Monnet * if the maximum number of tail calls has been reached for this 1882ad4a5223SQuentin Monnet * chain of programs. This limit is defined in the kernel by the 1883ad4a5223SQuentin Monnet * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 1884ebf7f6f0STiezhu Yang * which is currently set to 33. 1885ad4a5223SQuentin Monnet * Return 1886ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1887ad4a5223SQuentin Monnet * 1888bdb7b79bSAndrii Nakryiko * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 1889ad4a5223SQuentin Monnet * Description 1890ad4a5223SQuentin Monnet * Clone and redirect the packet associated to *skb* to another 1891ad4a5223SQuentin Monnet * net device of index *ifindex*. Both ingress and egress 1892ad4a5223SQuentin Monnet * interfaces can be used for redirection. The **BPF_F_INGRESS** 1893ad4a5223SQuentin Monnet * value in *flags* is used to make the distinction (ingress path 1894ad4a5223SQuentin Monnet * is selected if the flag is present, egress path otherwise). 1895ad4a5223SQuentin Monnet * This is the only flag supported for now. 1896ad4a5223SQuentin Monnet * 1897ad4a5223SQuentin Monnet * In comparison with **bpf_redirect**\ () helper, 1898ad4a5223SQuentin Monnet * **bpf_clone_redirect**\ () has the associated cost of 1899ad4a5223SQuentin Monnet * duplicating the packet buffer, but this can be executed out of 1900ad4a5223SQuentin Monnet * the eBPF program. Conversely, **bpf_redirect**\ () is more 1901ad4a5223SQuentin Monnet * efficient, but it is handled through an action code where the 1902ad4a5223SQuentin Monnet * redirection happens only after the eBPF program has returned. 1903ad4a5223SQuentin Monnet * 190432e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1905ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1906ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1907ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1908ad4a5223SQuentin Monnet * direct packet access. 1909ad4a5223SQuentin Monnet * Return 1910ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1911c456dec4SQuentin Monnet * 1912c456dec4SQuentin Monnet * u64 bpf_get_current_pid_tgid(void) 1913e40fbbf0SUsama Arif * Description 1914e40fbbf0SUsama Arif * Get the current pid and tgid. 1915c456dec4SQuentin Monnet * Return 1916c456dec4SQuentin Monnet * A 64-bit integer containing the current tgid and pid, and 1917c456dec4SQuentin Monnet * created as such: 1918c456dec4SQuentin Monnet * *current_task*\ **->tgid << 32 \|** 1919c456dec4SQuentin Monnet * *current_task*\ **->pid**. 1920c456dec4SQuentin Monnet * 1921c456dec4SQuentin Monnet * u64 bpf_get_current_uid_gid(void) 1922e40fbbf0SUsama Arif * Description 1923e40fbbf0SUsama Arif * Get the current uid and gid. 1924c456dec4SQuentin Monnet * Return 1925c456dec4SQuentin Monnet * A 64-bit integer containing the current GID and UID, and 1926c456dec4SQuentin Monnet * created as such: *current_gid* **<< 32 \|** *current_uid*. 1927c456dec4SQuentin Monnet * 1928bdb7b79bSAndrii Nakryiko * long bpf_get_current_comm(void *buf, u32 size_of_buf) 1929c456dec4SQuentin Monnet * Description 1930c456dec4SQuentin Monnet * Copy the **comm** attribute of the current task into *buf* of 1931c456dec4SQuentin Monnet * *size_of_buf*. The **comm** attribute contains the name of 1932c456dec4SQuentin Monnet * the executable (excluding the path) for the current task. The 1933c456dec4SQuentin Monnet * *size_of_buf* must be strictly positive. On success, the 1934c456dec4SQuentin Monnet * helper makes sure that the *buf* is NUL-terminated. On failure, 1935c456dec4SQuentin Monnet * it is filled with zeroes. 1936c456dec4SQuentin Monnet * Return 1937c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1938c456dec4SQuentin Monnet * 19391fdd08beSQuentin Monnet * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 19401fdd08beSQuentin Monnet * Description 19411fdd08beSQuentin Monnet * Retrieve the classid for the current task, i.e. for the net_cls 19421fdd08beSQuentin Monnet * cgroup to which *skb* belongs. 19431fdd08beSQuentin Monnet * 19441fdd08beSQuentin Monnet * This helper can be used on TC egress path, but not on ingress. 19451fdd08beSQuentin Monnet * 19461fdd08beSQuentin Monnet * The net_cls cgroup provides an interface to tag network packets 19471fdd08beSQuentin Monnet * based on a user-provided identifier for all traffic coming from 19481fdd08beSQuentin Monnet * the tasks belonging to the related cgroup. See also the related 19491fdd08beSQuentin Monnet * kernel documentation, available from the Linux sources in file 1950da82c92fSMauro Carvalho Chehab * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 19511fdd08beSQuentin Monnet * 19521fdd08beSQuentin Monnet * The Linux kernel has two versions for cgroups: there are 19531fdd08beSQuentin Monnet * cgroups v1 and cgroups v2. Both are available to users, who can 19541fdd08beSQuentin Monnet * use a mixture of them, but note that the net_cls cgroup is for 19551fdd08beSQuentin Monnet * cgroup v1 only. This makes it incompatible with BPF programs 19561fdd08beSQuentin Monnet * run on cgroups, which is a cgroup-v2-only feature (a socket can 19571fdd08beSQuentin Monnet * only hold data for one version of cgroups at a time). 19581fdd08beSQuentin Monnet * 19591fdd08beSQuentin Monnet * This helper is only available is the kernel was compiled with 19601fdd08beSQuentin Monnet * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 19611fdd08beSQuentin Monnet * "**y**" or to "**m**". 19621fdd08beSQuentin Monnet * Return 19631fdd08beSQuentin Monnet * The classid, or 0 for the default unconfigured classid. 19641fdd08beSQuentin Monnet * 1965bdb7b79bSAndrii Nakryiko * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1966c456dec4SQuentin Monnet * Description 1967c456dec4SQuentin Monnet * Push a *vlan_tci* (VLAN tag control information) of protocol 1968c456dec4SQuentin Monnet * *vlan_proto* to the packet associated to *skb*, then update 1969c456dec4SQuentin Monnet * the checksum. Note that if *vlan_proto* is different from 1970c456dec4SQuentin Monnet * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1971c456dec4SQuentin Monnet * be **ETH_P_8021Q**. 1972c456dec4SQuentin Monnet * 197332e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1974c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1975c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 1976c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 1977c456dec4SQuentin Monnet * direct packet access. 1978c456dec4SQuentin Monnet * Return 1979c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1980c456dec4SQuentin Monnet * 1981bdb7b79bSAndrii Nakryiko * long bpf_skb_vlan_pop(struct sk_buff *skb) 1982c456dec4SQuentin Monnet * Description 1983c456dec4SQuentin Monnet * Pop a VLAN header from the packet associated to *skb*. 1984c456dec4SQuentin Monnet * 198532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1986c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1987c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 1988c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 1989c456dec4SQuentin Monnet * direct packet access. 1990c456dec4SQuentin Monnet * Return 1991c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1992c456dec4SQuentin Monnet * 1993bdb7b79bSAndrii Nakryiko * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1994c456dec4SQuentin Monnet * Description 1995c456dec4SQuentin Monnet * Get tunnel metadata. This helper takes a pointer *key* to an 1996c456dec4SQuentin Monnet * empty **struct bpf_tunnel_key** of **size**, that will be 1997c456dec4SQuentin Monnet * filled with tunnel metadata for the packet associated to *skb*. 1998c456dec4SQuentin Monnet * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1999c456dec4SQuentin Monnet * indicates that the tunnel is based on IPv6 protocol instead of 2000c456dec4SQuentin Monnet * IPv4. 2001c456dec4SQuentin Monnet * 2002c456dec4SQuentin Monnet * The **struct bpf_tunnel_key** is an object that generalizes the 2003c456dec4SQuentin Monnet * principal parameters used by various tunneling protocols into a 2004c456dec4SQuentin Monnet * single struct. This way, it can be used to easily make a 2005c456dec4SQuentin Monnet * decision based on the contents of the encapsulation header, 2006c456dec4SQuentin Monnet * "summarized" in this struct. In particular, it holds the IP 2007c456dec4SQuentin Monnet * address of the remote end (IPv4 or IPv6, depending on the case) 2008c456dec4SQuentin Monnet * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 2009c456dec4SQuentin Monnet * this struct exposes the *key*\ **->tunnel_id**, which is 2010c456dec4SQuentin Monnet * generally mapped to a VNI (Virtual Network Identifier), making 2011c456dec4SQuentin Monnet * it programmable together with the **bpf_skb_set_tunnel_key**\ 2012c456dec4SQuentin Monnet * () helper. 2013c456dec4SQuentin Monnet * 2014c456dec4SQuentin Monnet * Let's imagine that the following code is part of a program 2015c456dec4SQuentin Monnet * attached to the TC ingress interface, on one end of a GRE 2016c456dec4SQuentin Monnet * tunnel, and is supposed to filter out all messages coming from 2017c456dec4SQuentin Monnet * remote ends with IPv4 address other than 10.0.0.1: 2018c456dec4SQuentin Monnet * 2019c456dec4SQuentin Monnet * :: 2020c456dec4SQuentin Monnet * 2021c456dec4SQuentin Monnet * int ret; 2022c456dec4SQuentin Monnet * struct bpf_tunnel_key key = {}; 2023c456dec4SQuentin Monnet * 2024c456dec4SQuentin Monnet * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 2025c456dec4SQuentin Monnet * if (ret < 0) 2026c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 2027c456dec4SQuentin Monnet * 2028c456dec4SQuentin Monnet * if (key.remote_ipv4 != 0x0a000001) 2029c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 2030c456dec4SQuentin Monnet * 2031c456dec4SQuentin Monnet * return TC_ACT_OK; // accept packet 2032c456dec4SQuentin Monnet * 2033c456dec4SQuentin Monnet * This interface can also be used with all encapsulation devices 2034c456dec4SQuentin Monnet * that can operate in "collect metadata" mode: instead of having 2035c456dec4SQuentin Monnet * one network device per specific configuration, the "collect 2036c456dec4SQuentin Monnet * metadata" mode only requires a single device where the 2037c456dec4SQuentin Monnet * configuration can be extracted from this helper. 2038c456dec4SQuentin Monnet * 2039c456dec4SQuentin Monnet * This can be used together with various tunnels such as VXLan, 2040c456dec4SQuentin Monnet * Geneve, GRE or IP in IP (IPIP). 2041c456dec4SQuentin Monnet * Return 2042c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 2043c456dec4SQuentin Monnet * 2044bdb7b79bSAndrii Nakryiko * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2045c456dec4SQuentin Monnet * Description 2046c456dec4SQuentin Monnet * Populate tunnel metadata for packet associated to *skb.* The 2047c456dec4SQuentin Monnet * tunnel metadata is set to the contents of *key*, of *size*. The 2048c456dec4SQuentin Monnet * *flags* can be set to a combination of the following values: 2049c456dec4SQuentin Monnet * 2050c456dec4SQuentin Monnet * **BPF_F_TUNINFO_IPV6** 2051c456dec4SQuentin Monnet * Indicate that the tunnel is based on IPv6 protocol 2052c456dec4SQuentin Monnet * instead of IPv4. 2053c456dec4SQuentin Monnet * **BPF_F_ZERO_CSUM_TX** 2054c456dec4SQuentin Monnet * For IPv4 packets, add a flag to tunnel metadata 2055c456dec4SQuentin Monnet * indicating that checksum computation should be skipped 2056c456dec4SQuentin Monnet * and checksum set to zeroes. 2057c456dec4SQuentin Monnet * **BPF_F_DONT_FRAGMENT** 2058c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that the 2059c456dec4SQuentin Monnet * packet should not be fragmented. 2060c456dec4SQuentin Monnet * **BPF_F_SEQ_NUMBER** 2061c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that a 2062c456dec4SQuentin Monnet * sequence number should be added to tunnel header before 2063c456dec4SQuentin Monnet * sending the packet. This flag was added for GRE 2064c456dec4SQuentin Monnet * encapsulation, but might be used with other protocols 2065c456dec4SQuentin Monnet * as well in the future. 2066e26aa600SChristian Ehrig * **BPF_F_NO_TUNNEL_KEY** 2067e26aa600SChristian Ehrig * Add a flag to tunnel metadata indicating that no tunnel 2068e26aa600SChristian Ehrig * key should be set in the resulting tunnel header. 2069c456dec4SQuentin Monnet * 2070c456dec4SQuentin Monnet * Here is a typical usage on the transmit path: 2071c456dec4SQuentin Monnet * 2072c456dec4SQuentin Monnet * :: 2073c456dec4SQuentin Monnet * 2074c456dec4SQuentin Monnet * struct bpf_tunnel_key key; 2075c456dec4SQuentin Monnet * populate key ... 2076c456dec4SQuentin Monnet * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 2077c456dec4SQuentin Monnet * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 2078c456dec4SQuentin Monnet * 2079c456dec4SQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_key**\ () 2080c456dec4SQuentin Monnet * helper for additional information. 2081c456dec4SQuentin Monnet * Return 2082c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 2083c456dec4SQuentin Monnet * 2084c6b5fb86SQuentin Monnet * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 2085c6b5fb86SQuentin Monnet * Description 2086c6b5fb86SQuentin Monnet * Read the value of a perf event counter. This helper relies on a 2087c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 2088c6b5fb86SQuentin Monnet * the perf event counter is selected when *map* is updated with 2089c6b5fb86SQuentin Monnet * perf event file descriptors. The *map* is an array whose size 2090c6b5fb86SQuentin Monnet * is the number of available CPUs, and each cell contains a value 2091c6b5fb86SQuentin Monnet * relative to one CPU. The value to retrieve is indicated by 2092c6b5fb86SQuentin Monnet * *flags*, that contains the index of the CPU to look up, masked 2093c6b5fb86SQuentin Monnet * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2094c6b5fb86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 2095c6b5fb86SQuentin Monnet * current CPU should be retrieved. 2096c6b5fb86SQuentin Monnet * 2097c6b5fb86SQuentin Monnet * Note that before Linux 4.13, only hardware perf event can be 2098c6b5fb86SQuentin Monnet * retrieved. 2099c6b5fb86SQuentin Monnet * 2100c6b5fb86SQuentin Monnet * Also, be aware that the newer helper 2101c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () is recommended over 21023bd5a09bSQuentin Monnet * **bpf_perf_event_read**\ () in general. The latter has some ABI 2103c6b5fb86SQuentin Monnet * quirks where error and counter value are used as a return code 2104c6b5fb86SQuentin Monnet * (which is wrong to do since ranges may overlap). This issue is 21053bd5a09bSQuentin Monnet * fixed with **bpf_perf_event_read_value**\ (), which at the same 21063bd5a09bSQuentin Monnet * time provides more features over the **bpf_perf_event_read**\ 21073bd5a09bSQuentin Monnet * () interface. Please refer to the description of 2108c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () for details. 2109c6b5fb86SQuentin Monnet * Return 2110c6b5fb86SQuentin Monnet * The value of the perf event counter read from the map, or a 2111c6b5fb86SQuentin Monnet * negative error code in case of failure. 2112c6b5fb86SQuentin Monnet * 2113bdb7b79bSAndrii Nakryiko * long bpf_redirect(u32 ifindex, u64 flags) 2114c456dec4SQuentin Monnet * Description 2115c456dec4SQuentin Monnet * Redirect the packet to another net device of index *ifindex*. 2116c456dec4SQuentin Monnet * This helper is somewhat similar to **bpf_clone_redirect**\ 2117c456dec4SQuentin Monnet * (), except that the packet is not cloned, which provides 2118c456dec4SQuentin Monnet * increased performance. 2119c456dec4SQuentin Monnet * 2120c456dec4SQuentin Monnet * Except for XDP, both ingress and egress interfaces can be used 2121c456dec4SQuentin Monnet * for redirection. The **BPF_F_INGRESS** value in *flags* is used 2122c456dec4SQuentin Monnet * to make the distinction (ingress path is selected if the flag 2123c456dec4SQuentin Monnet * is present, egress path otherwise). Currently, XDP only 2124c456dec4SQuentin Monnet * supports redirection to the egress interface, and accepts no 2125c456dec4SQuentin Monnet * flag at all. 2126c456dec4SQuentin Monnet * 2127f25975f4SToke Høiland-Jørgensen * The same effect can also be attained with the more generic 2128f25975f4SToke Høiland-Jørgensen * **bpf_redirect_map**\ (), which uses a BPF map to store the 2129f25975f4SToke Høiland-Jørgensen * redirect target instead of providing it directly to the helper. 2130c456dec4SQuentin Monnet * Return 2131c456dec4SQuentin Monnet * For XDP, the helper returns **XDP_REDIRECT** on success or 2132c456dec4SQuentin Monnet * **XDP_ABORTED** on error. For other program types, the values 2133c456dec4SQuentin Monnet * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2134c456dec4SQuentin Monnet * error. 2135c456dec4SQuentin Monnet * 21361fdd08beSQuentin Monnet * u32 bpf_get_route_realm(struct sk_buff *skb) 21371fdd08beSQuentin Monnet * Description 21381fdd08beSQuentin Monnet * Retrieve the realm or the route, that is to say the 21391fdd08beSQuentin Monnet * **tclassid** field of the destination for the *skb*. The 2140b16fc097STobias Klauser * identifier retrieved is a user-provided tag, similar to the 21411fdd08beSQuentin Monnet * one used with the net_cls cgroup (see description for 21421fdd08beSQuentin Monnet * **bpf_get_cgroup_classid**\ () helper), but here this tag is 21431fdd08beSQuentin Monnet * held by a route (a destination entry), not by a task. 21441fdd08beSQuentin Monnet * 21451fdd08beSQuentin Monnet * Retrieving this identifier works with the clsact TC egress hook 21461fdd08beSQuentin Monnet * (see also **tc-bpf(8)**), or alternatively on conventional 21471fdd08beSQuentin Monnet * classful egress qdiscs, but not on TC ingress path. In case of 21481fdd08beSQuentin Monnet * clsact TC egress hook, this has the advantage that, internally, 21491fdd08beSQuentin Monnet * the destination entry has not been dropped yet in the transmit 21501fdd08beSQuentin Monnet * path. Therefore, the destination entry does not need to be 21511fdd08beSQuentin Monnet * artificially held via **netif_keep_dst**\ () for a classful 21521fdd08beSQuentin Monnet * qdisc until the *skb* is freed. 21531fdd08beSQuentin Monnet * 21541fdd08beSQuentin Monnet * This helper is available only if the kernel was compiled with 21551fdd08beSQuentin Monnet * **CONFIG_IP_ROUTE_CLASSID** configuration option. 21561fdd08beSQuentin Monnet * Return 21571fdd08beSQuentin Monnet * The realm of the route for the packet associated to *skb*, or 0 21581fdd08beSQuentin Monnet * if none was found. 21591fdd08beSQuentin Monnet * 2160bdb7b79bSAndrii Nakryiko * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2161c456dec4SQuentin Monnet * Description 2162c456dec4SQuentin Monnet * Write raw *data* blob into a special BPF perf event held by 2163c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2164c456dec4SQuentin Monnet * event must have the following attributes: **PERF_SAMPLE_RAW** 2165c456dec4SQuentin Monnet * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2166c456dec4SQuentin Monnet * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2167c456dec4SQuentin Monnet * 2168c456dec4SQuentin Monnet * The *flags* are used to indicate the index in *map* for which 2169c456dec4SQuentin Monnet * the value must be put, masked with **BPF_F_INDEX_MASK**. 2170c456dec4SQuentin Monnet * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2171c456dec4SQuentin Monnet * to indicate that the index of the current CPU core should be 2172c456dec4SQuentin Monnet * used. 2173c456dec4SQuentin Monnet * 2174c456dec4SQuentin Monnet * The value to write, of *size*, is passed through eBPF stack and 2175c456dec4SQuentin Monnet * pointed by *data*. 2176c456dec4SQuentin Monnet * 2177c456dec4SQuentin Monnet * The context of the program *ctx* needs also be passed to the 2178c456dec4SQuentin Monnet * helper. 2179c456dec4SQuentin Monnet * 2180c456dec4SQuentin Monnet * On user space, a program willing to read the values needs to 2181c456dec4SQuentin Monnet * call **perf_event_open**\ () on the perf event (either for 2182c456dec4SQuentin Monnet * one or for all CPUs) and to store the file descriptor into the 2183c456dec4SQuentin Monnet * *map*. This must be done before the eBPF program can send data 2184c456dec4SQuentin Monnet * into it. An example is available in file 2185c456dec4SQuentin Monnet * *samples/bpf/trace_output_user.c* in the Linux kernel source 2186c456dec4SQuentin Monnet * tree (the eBPF program counterpart is in 2187c456dec4SQuentin Monnet * *samples/bpf/trace_output_kern.c*). 2188c456dec4SQuentin Monnet * 2189c456dec4SQuentin Monnet * **bpf_perf_event_output**\ () achieves better performance 2190c456dec4SQuentin Monnet * than **bpf_trace_printk**\ () for sharing data with user 2191c456dec4SQuentin Monnet * space, and is much better suitable for streaming data from eBPF 2192c456dec4SQuentin Monnet * programs. 2193c456dec4SQuentin Monnet * 2194c456dec4SQuentin Monnet * Note that this helper is not restricted to tracing use cases 2195c456dec4SQuentin Monnet * and can be used with programs attached to TC or XDP as well, 2196c456dec4SQuentin Monnet * where it allows for passing data to user space listeners. Data 2197c456dec4SQuentin Monnet * can be: 2198c456dec4SQuentin Monnet * 2199c456dec4SQuentin Monnet * * Only custom structs, 2200c456dec4SQuentin Monnet * * Only the packet payload, or 2201c456dec4SQuentin Monnet * * A combination of both. 2202c456dec4SQuentin Monnet * Return 2203c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 2204c456dec4SQuentin Monnet * 2205bdb7b79bSAndrii Nakryiko * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 22061fdd08beSQuentin Monnet * Description 22071fdd08beSQuentin Monnet * This helper was provided as an easy way to load data from a 22081fdd08beSQuentin Monnet * packet. It can be used to load *len* bytes from *offset* from 22091fdd08beSQuentin Monnet * the packet associated to *skb*, into the buffer pointed by 22101fdd08beSQuentin Monnet * *to*. 22111fdd08beSQuentin Monnet * 22121fdd08beSQuentin Monnet * Since Linux 4.7, usage of this helper has mostly been replaced 22131fdd08beSQuentin Monnet * by "direct packet access", enabling packet data to be 22141fdd08beSQuentin Monnet * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 22151fdd08beSQuentin Monnet * pointing respectively to the first byte of packet data and to 22161fdd08beSQuentin Monnet * the byte after the last byte of packet data. However, it 22171fdd08beSQuentin Monnet * remains useful if one wishes to read large quantities of data 22181fdd08beSQuentin Monnet * at once from a packet into the eBPF stack. 22191fdd08beSQuentin Monnet * Return 22201fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 22211fdd08beSQuentin Monnet * 2222bdb7b79bSAndrii Nakryiko * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2223c456dec4SQuentin Monnet * Description 2224c456dec4SQuentin Monnet * Walk a user or a kernel stack and return its id. To achieve 2225c456dec4SQuentin Monnet * this, the helper needs *ctx*, which is a pointer to the context 2226c456dec4SQuentin Monnet * on which the tracing program is executed, and a pointer to a 2227c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2228c456dec4SQuentin Monnet * 2229c456dec4SQuentin Monnet * The last argument, *flags*, holds the number of stack frames to 2230c456dec4SQuentin Monnet * skip (from 0 to 255), masked with 2231c456dec4SQuentin Monnet * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2232c456dec4SQuentin Monnet * a combination of the following flags: 2233c456dec4SQuentin Monnet * 2234c456dec4SQuentin Monnet * **BPF_F_USER_STACK** 2235c456dec4SQuentin Monnet * Collect a user space stack instead of a kernel stack. 2236c456dec4SQuentin Monnet * **BPF_F_FAST_STACK_CMP** 2237c456dec4SQuentin Monnet * Compare stacks by hash only. 2238c456dec4SQuentin Monnet * **BPF_F_REUSE_STACKID** 2239c456dec4SQuentin Monnet * If two different stacks hash into the same *stackid*, 2240c456dec4SQuentin Monnet * discard the old one. 2241c456dec4SQuentin Monnet * 2242c456dec4SQuentin Monnet * The stack id retrieved is a 32 bit long integer handle which 2243c456dec4SQuentin Monnet * can be further combined with other data (including other stack 2244c456dec4SQuentin Monnet * ids) and used as a key into maps. This can be useful for 2245c456dec4SQuentin Monnet * generating a variety of graphs (such as flame graphs or off-cpu 2246c456dec4SQuentin Monnet * graphs). 2247c456dec4SQuentin Monnet * 2248c456dec4SQuentin Monnet * For walking a stack, this helper is an improvement over 2249c456dec4SQuentin Monnet * **bpf_probe_read**\ (), which can be used with unrolled loops 2250c456dec4SQuentin Monnet * but is not efficient and consumes a lot of eBPF instructions. 2251c456dec4SQuentin Monnet * Instead, **bpf_get_stackid**\ () can collect up to 2252c456dec4SQuentin Monnet * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2253c456dec4SQuentin Monnet * this limit can be controlled with the **sysctl** program, and 2254c456dec4SQuentin Monnet * that it should be manually increased in order to profile long 2255c456dec4SQuentin Monnet * user stacks (such as stacks for Java programs). To do so, use: 2256c456dec4SQuentin Monnet * 2257c456dec4SQuentin Monnet * :: 2258c456dec4SQuentin Monnet * 2259c456dec4SQuentin Monnet * # sysctl kernel.perf_event_max_stack=<new value> 2260c456dec4SQuentin Monnet * Return 2261c456dec4SQuentin Monnet * The positive or null stack id on success, or a negative error 2262c456dec4SQuentin Monnet * in case of failure. 2263c456dec4SQuentin Monnet * 22641fdd08beSQuentin Monnet * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 22651fdd08beSQuentin Monnet * Description 22661fdd08beSQuentin Monnet * Compute a checksum difference, from the raw buffer pointed by 22671fdd08beSQuentin Monnet * *from*, of length *from_size* (that must be a multiple of 4), 22681fdd08beSQuentin Monnet * towards the raw buffer pointed by *to*, of size *to_size* 22691fdd08beSQuentin Monnet * (same remark). An optional *seed* can be added to the value 22701fdd08beSQuentin Monnet * (this can be cascaded, the seed may come from a previous call 22711fdd08beSQuentin Monnet * to the helper). 22721fdd08beSQuentin Monnet * 22731fdd08beSQuentin Monnet * This is flexible enough to be used in several ways: 22741fdd08beSQuentin Monnet * 22751fdd08beSQuentin Monnet * * With *from_size* == 0, *to_size* > 0 and *seed* set to 22761fdd08beSQuentin Monnet * checksum, it can be used when pushing new data. 22771fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* == 0 and *seed* set to 22781fdd08beSQuentin Monnet * checksum, it can be used when removing data from a packet. 22791fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 22801fdd08beSQuentin Monnet * can be used to compute a diff. Note that *from_size* and 22811fdd08beSQuentin Monnet * *to_size* do not need to be equal. 22821fdd08beSQuentin Monnet * 22831fdd08beSQuentin Monnet * This helper can be used in combination with 22841fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 22851fdd08beSQuentin Monnet * which one can feed in the difference computed with 22861fdd08beSQuentin Monnet * **bpf_csum_diff**\ (). 22871fdd08beSQuentin Monnet * Return 22881fdd08beSQuentin Monnet * The checksum result, or a negative error code in case of 22891fdd08beSQuentin Monnet * failure. 22901fdd08beSQuentin Monnet * 2291bdb7b79bSAndrii Nakryiko * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 22921fdd08beSQuentin Monnet * Description 22931fdd08beSQuentin Monnet * Retrieve tunnel options metadata for the packet associated to 22941fdd08beSQuentin Monnet * *skb*, and store the raw tunnel option data to the buffer *opt* 22951fdd08beSQuentin Monnet * of *size*. 22961fdd08beSQuentin Monnet * 22971fdd08beSQuentin Monnet * This helper can be used with encapsulation devices that can 22981fdd08beSQuentin Monnet * operate in "collect metadata" mode (please refer to the related 22991fdd08beSQuentin Monnet * note in the description of **bpf_skb_get_tunnel_key**\ () for 23001fdd08beSQuentin Monnet * more details). A particular example where this can be used is 23011fdd08beSQuentin Monnet * in combination with the Geneve encapsulation protocol, where it 23021fdd08beSQuentin Monnet * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 23031fdd08beSQuentin Monnet * and retrieving arbitrary TLVs (Type-Length-Value headers) from 23041fdd08beSQuentin Monnet * the eBPF program. This allows for full customization of these 23051fdd08beSQuentin Monnet * headers. 23061fdd08beSQuentin Monnet * Return 23071fdd08beSQuentin Monnet * The size of the option data retrieved. 23081fdd08beSQuentin Monnet * 2309bdb7b79bSAndrii Nakryiko * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 23101fdd08beSQuentin Monnet * Description 23111fdd08beSQuentin Monnet * Set tunnel options metadata for the packet associated to *skb* 23121fdd08beSQuentin Monnet * to the option data contained in the raw buffer *opt* of *size*. 23131fdd08beSQuentin Monnet * 23141fdd08beSQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_opt**\ () 23151fdd08beSQuentin Monnet * helper for additional information. 23161fdd08beSQuentin Monnet * Return 23171fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 23181fdd08beSQuentin Monnet * 2319bdb7b79bSAndrii Nakryiko * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 23201fdd08beSQuentin Monnet * Description 23211fdd08beSQuentin Monnet * Change the protocol of the *skb* to *proto*. Currently 23221fdd08beSQuentin Monnet * supported are transition from IPv4 to IPv6, and from IPv6 to 23231fdd08beSQuentin Monnet * IPv4. The helper takes care of the groundwork for the 23241fdd08beSQuentin Monnet * transition, including resizing the socket buffer. The eBPF 23251fdd08beSQuentin Monnet * program is expected to fill the new headers, if any, via 23261fdd08beSQuentin Monnet * **skb_store_bytes**\ () and to recompute the checksums with 23271fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 23281fdd08beSQuentin Monnet * (). The main case for this helper is to perform NAT64 23291fdd08beSQuentin Monnet * operations out of an eBPF program. 23301fdd08beSQuentin Monnet * 23311fdd08beSQuentin Monnet * Internally, the GSO type is marked as dodgy so that headers are 23321fdd08beSQuentin Monnet * checked and segments are recalculated by the GSO/GRO engine. 23331fdd08beSQuentin Monnet * The size for GSO target is adapted as well. 23341fdd08beSQuentin Monnet * 23351fdd08beSQuentin Monnet * All values for *flags* are reserved for future usage, and must 23361fdd08beSQuentin Monnet * be left at zero. 23371fdd08beSQuentin Monnet * 233832e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 23391fdd08beSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 23401fdd08beSQuentin Monnet * previously done by the verifier are invalidated and must be 23411fdd08beSQuentin Monnet * performed again, if the helper is used in combination with 23421fdd08beSQuentin Monnet * direct packet access. 23431fdd08beSQuentin Monnet * Return 23441fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 23451fdd08beSQuentin Monnet * 2346bdb7b79bSAndrii Nakryiko * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 23471fdd08beSQuentin Monnet * Description 23481fdd08beSQuentin Monnet * Change the packet type for the packet associated to *skb*. This 23491fdd08beSQuentin Monnet * comes down to setting *skb*\ **->pkt_type** to *type*, except 23501fdd08beSQuentin Monnet * the eBPF program does not have a write access to *skb*\ 23511fdd08beSQuentin Monnet * **->pkt_type** beside this helper. Using a helper here allows 23521fdd08beSQuentin Monnet * for graceful handling of errors. 23531fdd08beSQuentin Monnet * 23541fdd08beSQuentin Monnet * The major use case is to change incoming *skb*s to 23551fdd08beSQuentin Monnet * **PACKET_HOST** in a programmatic way instead of having to 23561fdd08beSQuentin Monnet * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 23571fdd08beSQuentin Monnet * example. 23581fdd08beSQuentin Monnet * 23591fdd08beSQuentin Monnet * Note that *type* only allows certain values. At this time, they 23601fdd08beSQuentin Monnet * are: 23611fdd08beSQuentin Monnet * 23621fdd08beSQuentin Monnet * **PACKET_HOST** 23631fdd08beSQuentin Monnet * Packet is for us. 23641fdd08beSQuentin Monnet * **PACKET_BROADCAST** 23651fdd08beSQuentin Monnet * Send packet to all. 23661fdd08beSQuentin Monnet * **PACKET_MULTICAST** 23671fdd08beSQuentin Monnet * Send packet to group. 23681fdd08beSQuentin Monnet * **PACKET_OTHERHOST** 23691fdd08beSQuentin Monnet * Send packet to someone else. 23701fdd08beSQuentin Monnet * Return 23711fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 23721fdd08beSQuentin Monnet * 2373bdb7b79bSAndrii Nakryiko * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2374c6b5fb86SQuentin Monnet * Description 2375c6b5fb86SQuentin Monnet * Check whether *skb* is a descendant of the cgroup2 held by 2376c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2377c6b5fb86SQuentin Monnet * Return 2378c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 2379c6b5fb86SQuentin Monnet * 2380c6b5fb86SQuentin Monnet * * 0, if the *skb* failed the cgroup2 descendant test. 2381c6b5fb86SQuentin Monnet * * 1, if the *skb* succeeded the cgroup2 descendant test. 2382c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 2383c6b5fb86SQuentin Monnet * 2384fa15601aSQuentin Monnet * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2385fa15601aSQuentin Monnet * Description 2386fa15601aSQuentin Monnet * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2387fa15601aSQuentin Monnet * not set, in particular if the hash was cleared due to mangling, 2388fa15601aSQuentin Monnet * recompute this hash. Later accesses to the hash can be done 2389fa15601aSQuentin Monnet * directly with *skb*\ **->hash**. 2390fa15601aSQuentin Monnet * 2391fa15601aSQuentin Monnet * Calling **bpf_set_hash_invalid**\ (), changing a packet 2392fa15601aSQuentin Monnet * prototype with **bpf_skb_change_proto**\ (), or calling 2393fa15601aSQuentin Monnet * **bpf_skb_store_bytes**\ () with the 2394fa15601aSQuentin Monnet * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2395fa15601aSQuentin Monnet * the hash and to trigger a new computation for the next call to 2396fa15601aSQuentin Monnet * **bpf_get_hash_recalc**\ (). 2397fa15601aSQuentin Monnet * Return 2398fa15601aSQuentin Monnet * The 32-bit hash. 2399fa15601aSQuentin Monnet * 2400c456dec4SQuentin Monnet * u64 bpf_get_current_task(void) 2401e40fbbf0SUsama Arif * Description 2402e40fbbf0SUsama Arif * Get the current task. 2403c456dec4SQuentin Monnet * Return 2404c456dec4SQuentin Monnet * A pointer to the current task struct. 2405fa15601aSQuentin Monnet * 2406bdb7b79bSAndrii Nakryiko * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2407c6b5fb86SQuentin Monnet * Description 2408c6b5fb86SQuentin Monnet * Attempt in a safe way to write *len* bytes from the buffer 2409c6b5fb86SQuentin Monnet * *src* to *dst* in memory. It only works for threads that are in 2410c6b5fb86SQuentin Monnet * user context, and *dst* must be a valid user space address. 2411c6b5fb86SQuentin Monnet * 2412c6b5fb86SQuentin Monnet * This helper should not be used to implement any kind of 2413c6b5fb86SQuentin Monnet * security mechanism because of TOC-TOU attacks, but rather to 2414c6b5fb86SQuentin Monnet * debug, divert, and manipulate execution of semi-cooperative 2415c6b5fb86SQuentin Monnet * processes. 2416c6b5fb86SQuentin Monnet * 2417c6b5fb86SQuentin Monnet * Keep in mind that this feature is meant for experiments, and it 2418c6b5fb86SQuentin Monnet * has a risk of crashing the system and running programs. 2419c6b5fb86SQuentin Monnet * Therefore, when an eBPF program using this helper is attached, 2420c6b5fb86SQuentin Monnet * a warning including PID and process name is printed to kernel 2421c6b5fb86SQuentin Monnet * logs. 2422c6b5fb86SQuentin Monnet * Return 2423c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2424c6b5fb86SQuentin Monnet * 2425bdb7b79bSAndrii Nakryiko * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2426c6b5fb86SQuentin Monnet * Description 2427c6b5fb86SQuentin Monnet * Check whether the probe is being run is the context of a given 2428c6b5fb86SQuentin Monnet * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2429c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2430c6b5fb86SQuentin Monnet * Return 2431c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 2432c6b5fb86SQuentin Monnet * 243358617014SHengqi Chen * * 1, if current task belongs to the cgroup2. 243458617014SHengqi Chen * * 0, if current task does not belong to the cgroup2. 2435c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 2436c6b5fb86SQuentin Monnet * 2437bdb7b79bSAndrii Nakryiko * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2438fa15601aSQuentin Monnet * Description 2439fa15601aSQuentin Monnet * Resize (trim or grow) the packet associated to *skb* to the 2440fa15601aSQuentin Monnet * new *len*. The *flags* are reserved for future usage, and must 2441fa15601aSQuentin Monnet * be left at zero. 2442fa15601aSQuentin Monnet * 2443fa15601aSQuentin Monnet * The basic idea is that the helper performs the needed work to 2444fa15601aSQuentin Monnet * change the size of the packet, then the eBPF program rewrites 2445fa15601aSQuentin Monnet * the rest via helpers like **bpf_skb_store_bytes**\ (), 2446fa15601aSQuentin Monnet * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2447fa15601aSQuentin Monnet * and others. This helper is a slow path utility intended for 2448fa15601aSQuentin Monnet * replies with control messages. And because it is targeted for 2449fa15601aSQuentin Monnet * slow path, the helper itself can afford to be slow: it 2450fa15601aSQuentin Monnet * implicitly linearizes, unclones and drops offloads from the 2451fa15601aSQuentin Monnet * *skb*. 2452fa15601aSQuentin Monnet * 245332e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2454fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2455fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2456fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2457fa15601aSQuentin Monnet * direct packet access. 2458fa15601aSQuentin Monnet * Return 2459fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2460fa15601aSQuentin Monnet * 2461bdb7b79bSAndrii Nakryiko * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2462fa15601aSQuentin Monnet * Description 2463fa15601aSQuentin Monnet * Pull in non-linear data in case the *skb* is non-linear and not 2464fa15601aSQuentin Monnet * all of *len* are part of the linear section. Make *len* bytes 2465fa15601aSQuentin Monnet * from *skb* readable and writable. If a zero value is passed for 2466bdb2bc75SJoanne Koong * *len*, then all bytes in the linear part of *skb* will be made 2467bdb2bc75SJoanne Koong * readable and writable. 2468fa15601aSQuentin Monnet * 2469fa15601aSQuentin Monnet * This helper is only needed for reading and writing with direct 2470fa15601aSQuentin Monnet * packet access. 2471fa15601aSQuentin Monnet * 2472fa15601aSQuentin Monnet * For direct packet access, testing that offsets to access 2473fa15601aSQuentin Monnet * are within packet boundaries (test on *skb*\ **->data_end**) is 2474fa15601aSQuentin Monnet * susceptible to fail if offsets are invalid, or if the requested 2475fa15601aSQuentin Monnet * data is in non-linear parts of the *skb*. On failure the 2476fa15601aSQuentin Monnet * program can just bail out, or in the case of a non-linear 2477fa15601aSQuentin Monnet * buffer, use a helper to make the data available. The 2478fa15601aSQuentin Monnet * **bpf_skb_load_bytes**\ () helper is a first solution to access 2479fa15601aSQuentin Monnet * the data. Another one consists in using **bpf_skb_pull_data** 2480fa15601aSQuentin Monnet * to pull in once the non-linear parts, then retesting and 2481fa15601aSQuentin Monnet * eventually access the data. 2482fa15601aSQuentin Monnet * 2483fa15601aSQuentin Monnet * At the same time, this also makes sure the *skb* is uncloned, 2484fa15601aSQuentin Monnet * which is a necessary condition for direct write. As this needs 2485fa15601aSQuentin Monnet * to be an invariant for the write part only, the verifier 2486fa15601aSQuentin Monnet * detects writes and adds a prologue that is calling 2487fa15601aSQuentin Monnet * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2488fa15601aSQuentin Monnet * the very beginning in case it is indeed cloned. 2489fa15601aSQuentin Monnet * 249032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2491fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2492fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2493fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2494fa15601aSQuentin Monnet * direct packet access. 2495fa15601aSQuentin Monnet * Return 2496fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2497fa15601aSQuentin Monnet * 2498fa15601aSQuentin Monnet * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2499fa15601aSQuentin Monnet * Description 2500fa15601aSQuentin Monnet * Add the checksum *csum* into *skb*\ **->csum** in case the 2501fa15601aSQuentin Monnet * driver has supplied a checksum for the entire packet into that 2502fa15601aSQuentin Monnet * field. Return an error otherwise. This helper is intended to be 2503fa15601aSQuentin Monnet * used in combination with **bpf_csum_diff**\ (), in particular 2504fa15601aSQuentin Monnet * when the checksum needs to be updated after data has been 2505fa15601aSQuentin Monnet * written into the packet through direct packet access. 2506fa15601aSQuentin Monnet * Return 2507fa15601aSQuentin Monnet * The checksum on success, or a negative error code in case of 2508fa15601aSQuentin Monnet * failure. 2509fa15601aSQuentin Monnet * 2510fa15601aSQuentin Monnet * void bpf_set_hash_invalid(struct sk_buff *skb) 2511fa15601aSQuentin Monnet * Description 2512fa15601aSQuentin Monnet * Invalidate the current *skb*\ **->hash**. It can be used after 2513fa15601aSQuentin Monnet * mangling on headers through direct packet access, in order to 2514fa15601aSQuentin Monnet * indicate that the hash is outdated and to trigger a 2515fa15601aSQuentin Monnet * recalculation the next time the kernel tries to access this 2516fa15601aSQuentin Monnet * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2517e40fbbf0SUsama Arif * Return 2518e40fbbf0SUsama Arif * void. 2519fa15601aSQuentin Monnet * 2520bdb7b79bSAndrii Nakryiko * long bpf_get_numa_node_id(void) 2521fa15601aSQuentin Monnet * Description 2522fa15601aSQuentin Monnet * Return the id of the current NUMA node. The primary use case 2523fa15601aSQuentin Monnet * for this helper is the selection of sockets for the local NUMA 2524fa15601aSQuentin Monnet * node, when the program is attached to sockets using the 2525fa15601aSQuentin Monnet * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2526fa15601aSQuentin Monnet * but the helper is also available to other eBPF program types, 2527fa15601aSQuentin Monnet * similarly to **bpf_get_smp_processor_id**\ (). 2528fa15601aSQuentin Monnet * Return 2529fa15601aSQuentin Monnet * The id of current NUMA node. 2530fa15601aSQuentin Monnet * 2531bdb7b79bSAndrii Nakryiko * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2532c6b5fb86SQuentin Monnet * Description 2533c6b5fb86SQuentin Monnet * Grows headroom of packet associated to *skb* and adjusts the 2534c6b5fb86SQuentin Monnet * offset of the MAC header accordingly, adding *len* bytes of 2535c6b5fb86SQuentin Monnet * space. It automatically extends and reallocates memory as 2536c6b5fb86SQuentin Monnet * required. 2537c6b5fb86SQuentin Monnet * 2538c6b5fb86SQuentin Monnet * This helper can be used on a layer 3 *skb* to push a MAC header 2539c6b5fb86SQuentin Monnet * for redirection into a layer 2 device. 2540c6b5fb86SQuentin Monnet * 2541c6b5fb86SQuentin Monnet * All values for *flags* are reserved for future usage, and must 2542c6b5fb86SQuentin Monnet * be left at zero. 2543c6b5fb86SQuentin Monnet * 254432e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2545c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2546c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 2547c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 2548c6b5fb86SQuentin Monnet * direct packet access. 2549c6b5fb86SQuentin Monnet * Return 2550c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2551c6b5fb86SQuentin Monnet * 2552bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2553c6b5fb86SQuentin Monnet * Description 2554c6b5fb86SQuentin Monnet * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2555c6b5fb86SQuentin Monnet * it is possible to use a negative value for *delta*. This helper 2556c6b5fb86SQuentin Monnet * can be used to prepare the packet for pushing or popping 2557c6b5fb86SQuentin Monnet * headers. 2558c6b5fb86SQuentin Monnet * 255932e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2560c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2561c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 2562c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 2563c6b5fb86SQuentin Monnet * direct packet access. 2564c6b5fb86SQuentin Monnet * Return 2565c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2566c6b5fb86SQuentin Monnet * 2567bdb7b79bSAndrii Nakryiko * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2568c6b5fb86SQuentin Monnet * Description 25696ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe kernel address 2570ab8d7809SQuentin Monnet * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 25716ae08ae3SDaniel Borkmann * more details. 2572c6b5fb86SQuentin Monnet * 2573ab8d7809SQuentin Monnet * Generally, use **bpf_probe_read_user_str**\ () or 2574ab8d7809SQuentin Monnet * **bpf_probe_read_kernel_str**\ () instead. 2575c6b5fb86SQuentin Monnet * Return 2576c6b5fb86SQuentin Monnet * On success, the strictly positive length of the string, 2577c6b5fb86SQuentin Monnet * including the trailing NUL character. On error, a negative 2578c6b5fb86SQuentin Monnet * value. 2579c6b5fb86SQuentin Monnet * 2580c6b5fb86SQuentin Monnet * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2581c6b5fb86SQuentin Monnet * Description 2582c6b5fb86SQuentin Monnet * If the **struct sk_buff** pointed by *skb* has a known socket, 2583c6b5fb86SQuentin Monnet * retrieve the cookie (generated by the kernel) of this socket. 2584c6b5fb86SQuentin Monnet * If no cookie has been set yet, generate a new cookie. Once 2585c6b5fb86SQuentin Monnet * generated, the socket cookie remains stable for the life of the 2586c6b5fb86SQuentin Monnet * socket. This helper can be useful for monitoring per socket 2587cd48bddaSDaniel Borkmann * networking traffic statistics as it provides a global socket 2588cd48bddaSDaniel Borkmann * identifier that can be assumed unique. 2589c6b5fb86SQuentin Monnet * Return 259007881ccbSFlorent Revest * A 8-byte long unique number on success, or 0 if the socket 259107881ccbSFlorent Revest * field is missing inside *skb*. 2592c6b5fb86SQuentin Monnet * 2593d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2594d692f113SAndrey Ignatov * Description 2595d692f113SAndrey Ignatov * Equivalent to bpf_get_socket_cookie() helper that accepts 259662369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_addr** context. 2597d692f113SAndrey Ignatov * Return 259807881ccbSFlorent Revest * A 8-byte long unique number. 2599d692f113SAndrey Ignatov * 2600d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2601d692f113SAndrey Ignatov * Description 2602ab8d7809SQuentin Monnet * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 260362369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_ops** context. 2604d692f113SAndrey Ignatov * Return 260507881ccbSFlorent Revest * A 8-byte long unique number. 2606d692f113SAndrey Ignatov * 2607c5dbb89fSFlorent Revest * u64 bpf_get_socket_cookie(struct sock *sk) 2608c5dbb89fSFlorent Revest * Description 2609c5dbb89fSFlorent Revest * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2610c5dbb89fSFlorent Revest * *sk*, but gets socket from a BTF **struct sock**. This helper 2611c5dbb89fSFlorent Revest * also works for sleepable programs. 2612c5dbb89fSFlorent Revest * Return 2613c5dbb89fSFlorent Revest * A 8-byte long unique number or 0 if *sk* is NULL. 2614c5dbb89fSFlorent Revest * 2615c6b5fb86SQuentin Monnet * u32 bpf_get_socket_uid(struct sk_buff *skb) 2616e40fbbf0SUsama Arif * Description 2617e40fbbf0SUsama Arif * Get the owner UID of the socked associated to *skb*. 2618c6b5fb86SQuentin Monnet * Return 2619c6b5fb86SQuentin Monnet * The owner UID of the socket associated to *skb*. If the socket 2620c6b5fb86SQuentin Monnet * is **NULL**, or if it is not a full socket (i.e. if it is a 2621c6b5fb86SQuentin Monnet * time-wait or a request socket instead), **overflowuid** value 2622c6b5fb86SQuentin Monnet * is returned (note that **overflowuid** might also be the actual 2623c6b5fb86SQuentin Monnet * UID value for the socket). 2624c6b5fb86SQuentin Monnet * 2625bdb7b79bSAndrii Nakryiko * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2626fa15601aSQuentin Monnet * Description 2627fa15601aSQuentin Monnet * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2628fa15601aSQuentin Monnet * to value *hash*. 2629fa15601aSQuentin Monnet * Return 2630fa15601aSQuentin Monnet * 0 2631fa15601aSQuentin Monnet * 2632bdb7b79bSAndrii Nakryiko * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 26337aa79a86SQuentin Monnet * Description 26347aa79a86SQuentin Monnet * Emulate a call to **setsockopt()** on the socket associated to 26357aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 26367aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 26377aa79a86SQuentin Monnet * must be specified, see **setsockopt(2)** for more information. 26387aa79a86SQuentin Monnet * The option value of length *optlen* is pointed by *optval*. 26397aa79a86SQuentin Monnet * 2640beecf11bSStanislav Fomichev * *bpf_socket* should be one of the following: 2641ab8d7809SQuentin Monnet * 2642beecf11bSStanislav Fomichev * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2643beecf11bSStanislav Fomichev * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2644beecf11bSStanislav Fomichev * and **BPF_CGROUP_INET6_CONNECT**. 2645beecf11bSStanislav Fomichev * 26467aa79a86SQuentin Monnet * This helper actually implements a subset of **setsockopt()**. 26477aa79a86SQuentin Monnet * It supports the following *level*\ s: 26487aa79a86SQuentin Monnet * 26497aa79a86SQuentin Monnet * * **SOL_SOCKET**, which supports the following *optname*\ s: 26507aa79a86SQuentin Monnet * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2651f9bcf968SDmitry Yakunin * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 265272b43bdeSJi Rongfeng * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**, 265372b43bdeSJi Rongfeng * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**. 26547aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports the following *optname*\ s: 26557aa79a86SQuentin Monnet * **TCP_CONGESTION**, **TCP_BPF_IW**, 2656f9bcf968SDmitry Yakunin * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2657f9bcf968SDmitry Yakunin * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 265872b43bdeSJi Rongfeng * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**, 265972b43bdeSJi Rongfeng * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**, 266072b43bdeSJi Rongfeng * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**, 266172b43bdeSJi Rongfeng * **TCP_BPF_RTO_MIN**. 26627aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 266372b43bdeSJi Rongfeng * * **IPPROTO_IPV6**, which supports the following *optname*\ s: 266472b43bdeSJi Rongfeng * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**. 26657aa79a86SQuentin Monnet * Return 26667aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 26677aa79a86SQuentin Monnet * 2668bdb7b79bSAndrii Nakryiko * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2669fa15601aSQuentin Monnet * Description 2670fa15601aSQuentin Monnet * Grow or shrink the room for data in the packet associated to 2671fa15601aSQuentin Monnet * *skb* by *len_diff*, and according to the selected *mode*. 2672fa15601aSQuentin Monnet * 2673836e66c2SDaniel Borkmann * By default, the helper will reset any offloaded checksum 2674836e66c2SDaniel Borkmann * indicator of the skb to CHECKSUM_NONE. This can be avoided 2675836e66c2SDaniel Borkmann * by the following flag: 2676836e66c2SDaniel Borkmann * 2677836e66c2SDaniel Borkmann * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2678836e66c2SDaniel Borkmann * checksum data of the skb to CHECKSUM_NONE. 2679836e66c2SDaniel Borkmann * 268014aa3192SWillem de Bruijn * There are two supported modes at this time: 268114aa3192SWillem de Bruijn * 268214aa3192SWillem de Bruijn * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 26834961d077SQuentin Monnet * (room space is added or removed between the layer 2 and 26844961d077SQuentin Monnet * layer 3 headers). 2685fa15601aSQuentin Monnet * 2686fa15601aSQuentin Monnet * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 26874961d077SQuentin Monnet * (room space is added or removed between the layer 3 and 26884961d077SQuentin Monnet * layer 4 headers). 2689fa15601aSQuentin Monnet * 2690868d5235SWillem de Bruijn * The following flags are supported at this time: 26912278f6ccSWillem de Bruijn * 26922278f6ccSWillem de Bruijn * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 26932278f6ccSWillem de Bruijn * Adjusting mss in this way is not allowed for datagrams. 2694fa15601aSQuentin Monnet * 269580867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 269680867c5eSQuentin Monnet * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2697868d5235SWillem de Bruijn * Any new space is reserved to hold a tunnel header. 2698868d5235SWillem de Bruijn * Configure skb offsets and other fields accordingly. 2699868d5235SWillem de Bruijn * 270080867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 270180867c5eSQuentin Monnet * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2702868d5235SWillem de Bruijn * Use with ENCAP_L3 flags to further specify the tunnel type. 2703868d5235SWillem de Bruijn * 270480867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 270558dfc900SAlan Maguire * Use with ENCAP_L3/L4 flags to further specify the tunnel 270680867c5eSQuentin Monnet * type; *len* is the length of the inner MAC header. 270758dfc900SAlan Maguire * 2708d01b59c9SXuesen Huang * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2709d01b59c9SXuesen Huang * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2710d01b59c9SXuesen Huang * L2 type as Ethernet. 2711d01b59c9SXuesen Huang * 2712d219df60SZiyang Xuan * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**, 2713d219df60SZiyang Xuan * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**: 2714d219df60SZiyang Xuan * Indicate the new IP header version after decapsulating the outer 2715d219df60SZiyang Xuan * IP header. Used when the inner and outer IP versions are different. 2716d219df60SZiyang Xuan * 271732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2718fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2719fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2720fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2721fa15601aSQuentin Monnet * direct packet access. 2722fa15601aSQuentin Monnet * Return 2723fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2724fa15601aSQuentin Monnet * 272532637e33SToke Høiland-Jørgensen * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags) 2726ab127040SQuentin Monnet * Description 2727ab127040SQuentin Monnet * Redirect the packet to the endpoint referenced by *map* at 2728ab127040SQuentin Monnet * index *key*. Depending on its type, this *map* can contain 2729ab127040SQuentin Monnet * references to net devices (for forwarding packets through other 2730ab127040SQuentin Monnet * ports), or to CPUs (for redirecting XDP frames to another CPU; 2731ab127040SQuentin Monnet * but this is only implemented for native XDP (with driver 2732ab127040SQuentin Monnet * support) as of this writing). 2733ab127040SQuentin Monnet * 273443e74c02SToke Høiland-Jørgensen * The lower two bits of *flags* are used as the return code if 273543e74c02SToke Høiland-Jørgensen * the map lookup fails. This is so that the return value can be 2736ab8d7809SQuentin Monnet * one of the XDP program return codes up to **XDP_TX**, as chosen 2737e624d4edSHangbin Liu * by the caller. The higher bits of *flags* can be set to 2738e624d4edSHangbin Liu * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2739e624d4edSHangbin Liu * 2740e624d4edSHangbin Liu * With BPF_F_BROADCAST the packet will be broadcasted to all the 2741e624d4edSHangbin Liu * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2742e624d4edSHangbin Liu * interface will be excluded when do broadcasting. 2743ab127040SQuentin Monnet * 2744ab8d7809SQuentin Monnet * See also **bpf_redirect**\ (), which only supports redirecting 2745ab8d7809SQuentin Monnet * to an ifindex, but doesn't require a map to do so. 2746ab127040SQuentin Monnet * Return 2747f25975f4SToke Høiland-Jørgensen * **XDP_REDIRECT** on success, or the value of the two lower bits 2748a33d3147SJakub Wilk * of the *flags* argument on error. 2749ab127040SQuentin Monnet * 2750bdb7b79bSAndrii Nakryiko * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2751ab127040SQuentin Monnet * Description 2752ab127040SQuentin Monnet * Redirect the packet to the socket referenced by *map* (of type 2753ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2754ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 2755ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 2756ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 2757ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 2758ab127040SQuentin Monnet * Return 2759ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 2760ab127040SQuentin Monnet * 2761bdb7b79bSAndrii Nakryiko * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2762ab127040SQuentin Monnet * Description 2763ab127040SQuentin Monnet * Add an entry to, or update a *map* referencing sockets. The 2764ab127040SQuentin Monnet * *skops* is used as a new value for the entry associated to 2765ab127040SQuentin Monnet * *key*. *flags* is one of: 2766ab127040SQuentin Monnet * 2767ab127040SQuentin Monnet * **BPF_NOEXIST** 2768ab127040SQuentin Monnet * The entry for *key* must not exist in the map. 2769ab127040SQuentin Monnet * **BPF_EXIST** 2770ab127040SQuentin Monnet * The entry for *key* must already exist in the map. 2771ab127040SQuentin Monnet * **BPF_ANY** 2772ab127040SQuentin Monnet * No condition on the existence of the entry for *key*. 2773ab127040SQuentin Monnet * 2774ab127040SQuentin Monnet * If the *map* has eBPF programs (parser and verdict), those will 2775ab127040SQuentin Monnet * be inherited by the socket being added. If the socket is 2776ab127040SQuentin Monnet * already attached to eBPF programs, this results in an error. 2777ab127040SQuentin Monnet * Return 2778ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 2779ab127040SQuentin Monnet * 2780bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2781fa15601aSQuentin Monnet * Description 2782fa15601aSQuentin Monnet * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2783fa15601aSQuentin Monnet * *delta* (which can be positive or negative). Note that this 2784fa15601aSQuentin Monnet * operation modifies the address stored in *xdp_md*\ **->data**, 2785fa15601aSQuentin Monnet * so the latter must be loaded only after the helper has been 2786fa15601aSQuentin Monnet * called. 2787fa15601aSQuentin Monnet * 2788fa15601aSQuentin Monnet * The use of *xdp_md*\ **->data_meta** is optional and programs 2789fa15601aSQuentin Monnet * are not required to use it. The rationale is that when the 2790fa15601aSQuentin Monnet * packet is processed with XDP (e.g. as DoS filter), it is 2791fa15601aSQuentin Monnet * possible to push further meta data along with it before passing 2792fa15601aSQuentin Monnet * to the stack, and to give the guarantee that an ingress eBPF 2793fa15601aSQuentin Monnet * program attached as a TC classifier on the same device can pick 2794fa15601aSQuentin Monnet * this up for further post-processing. Since TC works with socket 2795fa15601aSQuentin Monnet * buffers, it remains possible to set from XDP the **mark** or 2796fa15601aSQuentin Monnet * **priority** pointers, or other pointers for the socket buffer. 2797fa15601aSQuentin Monnet * Having this scratch space generic and programmable allows for 2798fa15601aSQuentin Monnet * more flexibility as the user is free to store whatever meta 2799fa15601aSQuentin Monnet * data they need. 2800fa15601aSQuentin Monnet * 280132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2802fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2803fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2804fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2805fa15601aSQuentin Monnet * direct packet access. 2806fa15601aSQuentin Monnet * Return 2807fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 28087aa79a86SQuentin Monnet * 2809bdb7b79bSAndrii Nakryiko * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 28107aa79a86SQuentin Monnet * Description 28117aa79a86SQuentin Monnet * Read the value of a perf event counter, and store it into *buf* 28127aa79a86SQuentin Monnet * of size *buf_size*. This helper relies on a *map* of type 28137aa79a86SQuentin Monnet * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 28147aa79a86SQuentin Monnet * counter is selected when *map* is updated with perf event file 28157aa79a86SQuentin Monnet * descriptors. The *map* is an array whose size is the number of 28167aa79a86SQuentin Monnet * available CPUs, and each cell contains a value relative to one 28177aa79a86SQuentin Monnet * CPU. The value to retrieve is indicated by *flags*, that 28187aa79a86SQuentin Monnet * contains the index of the CPU to look up, masked with 28197aa79a86SQuentin Monnet * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 28207aa79a86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 28217aa79a86SQuentin Monnet * current CPU should be retrieved. 28227aa79a86SQuentin Monnet * 28237aa79a86SQuentin Monnet * This helper behaves in a way close to 28247aa79a86SQuentin Monnet * **bpf_perf_event_read**\ () helper, save that instead of 28257aa79a86SQuentin Monnet * just returning the value observed, it fills the *buf* 28267aa79a86SQuentin Monnet * structure. This allows for additional data to be retrieved: in 28277aa79a86SQuentin Monnet * particular, the enabled and running times (in *buf*\ 28287aa79a86SQuentin Monnet * **->enabled** and *buf*\ **->running**, respectively) are 28297aa79a86SQuentin Monnet * copied. In general, **bpf_perf_event_read_value**\ () is 28307aa79a86SQuentin Monnet * recommended over **bpf_perf_event_read**\ (), which has some 28317aa79a86SQuentin Monnet * ABI issues and provides fewer functionalities. 28327aa79a86SQuentin Monnet * 28337aa79a86SQuentin Monnet * These values are interesting, because hardware PMU (Performance 28347aa79a86SQuentin Monnet * Monitoring Unit) counters are limited resources. When there are 28357aa79a86SQuentin Monnet * more PMU based perf events opened than available counters, 28367aa79a86SQuentin Monnet * kernel will multiplex these events so each event gets certain 28377aa79a86SQuentin Monnet * percentage (but not all) of the PMU time. In case that 28387aa79a86SQuentin Monnet * multiplexing happens, the number of samples or counter value 28397aa79a86SQuentin Monnet * will not reflect the case compared to when no multiplexing 28407aa79a86SQuentin Monnet * occurs. This makes comparison between different runs difficult. 28417aa79a86SQuentin Monnet * Typically, the counter value should be normalized before 28427aa79a86SQuentin Monnet * comparing to other experiments. The usual normalization is done 28437aa79a86SQuentin Monnet * as follows. 28447aa79a86SQuentin Monnet * 28457aa79a86SQuentin Monnet * :: 28467aa79a86SQuentin Monnet * 28477aa79a86SQuentin Monnet * normalized_counter = counter * t_enabled / t_running 28487aa79a86SQuentin Monnet * 28497aa79a86SQuentin Monnet * Where t_enabled is the time enabled for event and t_running is 28507aa79a86SQuentin Monnet * the time running for event since last normalization. The 28517aa79a86SQuentin Monnet * enabled and running times are accumulated since the perf event 28527aa79a86SQuentin Monnet * open. To achieve scaling factor between two invocations of an 2853ab8d7809SQuentin Monnet * eBPF program, users can use CPU id as the key (which is 28547aa79a86SQuentin Monnet * typical for perf array usage model) to remember the previous 28557aa79a86SQuentin Monnet * value and do the calculation inside the eBPF program. 28567aa79a86SQuentin Monnet * Return 28577aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 28587aa79a86SQuentin Monnet * 2859bdb7b79bSAndrii Nakryiko * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 28607aa79a86SQuentin Monnet * Description 286117c9b4e1SFlorian Lehner * For an eBPF program attached to a perf event, retrieve the 28627aa79a86SQuentin Monnet * value of the event counter associated to *ctx* and store it in 28637aa79a86SQuentin Monnet * the structure pointed by *buf* and of size *buf_size*. Enabled 28647aa79a86SQuentin Monnet * and running times are also stored in the structure (see 28657aa79a86SQuentin Monnet * description of helper **bpf_perf_event_read_value**\ () for 28667aa79a86SQuentin Monnet * more details). 28677aa79a86SQuentin Monnet * Return 28687aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 28697aa79a86SQuentin Monnet * 2870bdb7b79bSAndrii Nakryiko * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 28717aa79a86SQuentin Monnet * Description 28727aa79a86SQuentin Monnet * Emulate a call to **getsockopt()** on the socket associated to 28737aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 28747aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 28757aa79a86SQuentin Monnet * must be specified, see **getsockopt(2)** for more information. 28767aa79a86SQuentin Monnet * The retrieved value is stored in the structure pointed by 28777aa79a86SQuentin Monnet * *opval* and of length *optlen*. 28787aa79a86SQuentin Monnet * 2879beecf11bSStanislav Fomichev * *bpf_socket* should be one of the following: 2880ab8d7809SQuentin Monnet * 2881beecf11bSStanislav Fomichev * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2882beecf11bSStanislav Fomichev * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2883beecf11bSStanislav Fomichev * and **BPF_CGROUP_INET6_CONNECT**. 2884beecf11bSStanislav Fomichev * 28857aa79a86SQuentin Monnet * This helper actually implements a subset of **getsockopt()**. 288672b43bdeSJi Rongfeng * It supports the same set of *optname*\ s that is supported by 288772b43bdeSJi Rongfeng * the **bpf_setsockopt**\ () helper. The exceptions are 288872b43bdeSJi Rongfeng * **TCP_BPF_*** is **bpf_setsockopt**\ () only and 288972b43bdeSJi Rongfeng * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only. 28907aa79a86SQuentin Monnet * Return 28917aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 28927aa79a86SQuentin Monnet * 2893bdb7b79bSAndrii Nakryiko * long bpf_override_return(struct pt_regs *regs, u64 rc) 28947aa79a86SQuentin Monnet * Description 28957aa79a86SQuentin Monnet * Used for error injection, this helper uses kprobes to override 28967aa79a86SQuentin Monnet * the return value of the probed function, and to set it to *rc*. 28977aa79a86SQuentin Monnet * The first argument is the context *regs* on which the kprobe 28987aa79a86SQuentin Monnet * works. 28997aa79a86SQuentin Monnet * 2900ab8d7809SQuentin Monnet * This helper works by setting the PC (program counter) 29017aa79a86SQuentin Monnet * to an override function which is run in place of the original 29027aa79a86SQuentin Monnet * probed function. This means the probed function is not run at 29037aa79a86SQuentin Monnet * all. The replacement function just returns with the required 29047aa79a86SQuentin Monnet * value. 29057aa79a86SQuentin Monnet * 29067aa79a86SQuentin Monnet * This helper has security implications, and thus is subject to 29077aa79a86SQuentin Monnet * restrictions. It is only available if the kernel was compiled 29087aa79a86SQuentin Monnet * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 29097aa79a86SQuentin Monnet * option, and in this case it only works on functions tagged with 29107aa79a86SQuentin Monnet * **ALLOW_ERROR_INJECTION** in the kernel code. 29117aa79a86SQuentin Monnet * 29127aa79a86SQuentin Monnet * Also, the helper is only available for the architectures having 29137aa79a86SQuentin Monnet * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 29147aa79a86SQuentin Monnet * x86 architecture is the only one to support this feature. 29157aa79a86SQuentin Monnet * Return 29167aa79a86SQuentin Monnet * 0 29177aa79a86SQuentin Monnet * 2918bdb7b79bSAndrii Nakryiko * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 29197aa79a86SQuentin Monnet * Description 29207aa79a86SQuentin Monnet * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 29217aa79a86SQuentin Monnet * for the full TCP socket associated to *bpf_sock_ops* to 29227aa79a86SQuentin Monnet * *argval*. 29237aa79a86SQuentin Monnet * 29247aa79a86SQuentin Monnet * The primary use of this field is to determine if there should 29257aa79a86SQuentin Monnet * be calls to eBPF programs of type 29267aa79a86SQuentin Monnet * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 29277aa79a86SQuentin Monnet * code. A program of the same type can change its value, per 29287aa79a86SQuentin Monnet * connection and as necessary, when the connection is 29297aa79a86SQuentin Monnet * established. This field is directly accessible for reading, but 29307aa79a86SQuentin Monnet * this helper must be used for updates in order to return an 29317aa79a86SQuentin Monnet * error if an eBPF program tries to set a callback that is not 29327aa79a86SQuentin Monnet * supported in the current kernel. 29337aa79a86SQuentin Monnet * 2934725721a6SViet Hoang Tran * *argval* is a flag array which can combine these flags: 29357aa79a86SQuentin Monnet * 29367aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 29377aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 29387aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 293923729ff2SStanislav Fomichev * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 29407aa79a86SQuentin Monnet * 2941725721a6SViet Hoang Tran * Therefore, this function can be used to clear a callback flag by 2942725721a6SViet Hoang Tran * setting the appropriate bit to zero. e.g. to disable the RTO 2943725721a6SViet Hoang Tran * callback: 2944725721a6SViet Hoang Tran * 2945725721a6SViet Hoang Tran * **bpf_sock_ops_cb_flags_set(bpf_sock,** 2946725721a6SViet Hoang Tran * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 2947725721a6SViet Hoang Tran * 29487aa79a86SQuentin Monnet * Here are some examples of where one could call such eBPF 29497aa79a86SQuentin Monnet * program: 29507aa79a86SQuentin Monnet * 29517aa79a86SQuentin Monnet * * When RTO fires. 29527aa79a86SQuentin Monnet * * When a packet is retransmitted. 29537aa79a86SQuentin Monnet * * When the connection terminates. 29547aa79a86SQuentin Monnet * * When a packet is sent. 29557aa79a86SQuentin Monnet * * When a packet is received. 29567aa79a86SQuentin Monnet * Return 29577aa79a86SQuentin Monnet * Code **-EINVAL** if the socket is not a full TCP socket; 29587aa79a86SQuentin Monnet * otherwise, a positive number containing the bits that could not 29597aa79a86SQuentin Monnet * be set is returned (which comes down to 0 if all bits were set 29607aa79a86SQuentin Monnet * as required). 29617aa79a86SQuentin Monnet * 2962bdb7b79bSAndrii Nakryiko * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 2963ab127040SQuentin Monnet * Description 2964ab127040SQuentin Monnet * This helper is used in programs implementing policies at the 2965ab127040SQuentin Monnet * socket level. If the message *msg* is allowed to pass (i.e. if 2966ab127040SQuentin Monnet * the verdict eBPF program returns **SK_PASS**), redirect it to 2967ab127040SQuentin Monnet * the socket referenced by *map* (of type 2968ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2969ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 2970ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 2971ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 2972ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 2973ab127040SQuentin Monnet * Return 2974ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 2975ab127040SQuentin Monnet * 2976bdb7b79bSAndrii Nakryiko * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 2977ab127040SQuentin Monnet * Description 2978ab127040SQuentin Monnet * For socket policies, apply the verdict of the eBPF program to 2979ab127040SQuentin Monnet * the next *bytes* (number of bytes) of message *msg*. 2980ab127040SQuentin Monnet * 2981ab127040SQuentin Monnet * For example, this helper can be used in the following cases: 2982ab127040SQuentin Monnet * 2983ab127040SQuentin Monnet * * A single **sendmsg**\ () or **sendfile**\ () system call 2984ab127040SQuentin Monnet * contains multiple logical messages that the eBPF program is 2985ab127040SQuentin Monnet * supposed to read and for which it should apply a verdict. 2986ab127040SQuentin Monnet * * An eBPF program only cares to read the first *bytes* of a 2987ab127040SQuentin Monnet * *msg*. If the message has a large payload, then setting up 2988ab127040SQuentin Monnet * and calling the eBPF program repeatedly for all bytes, even 2989ab127040SQuentin Monnet * though the verdict is already known, would create unnecessary 2990ab127040SQuentin Monnet * overhead. 2991ab127040SQuentin Monnet * 2992ab127040SQuentin Monnet * When called from within an eBPF program, the helper sets a 2993ab127040SQuentin Monnet * counter internal to the BPF infrastructure, that is used to 2994ab127040SQuentin Monnet * apply the last verdict to the next *bytes*. If *bytes* is 2995ab127040SQuentin Monnet * smaller than the current data being processed from a 2996ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () system call, the first 2997ab127040SQuentin Monnet * *bytes* will be sent and the eBPF program will be re-run with 2998ab127040SQuentin Monnet * the pointer for start of data pointing to byte number *bytes* 2999ab127040SQuentin Monnet * **+ 1**. If *bytes* is larger than the current data being 3000ab127040SQuentin Monnet * processed, then the eBPF verdict will be applied to multiple 3001ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 3002ab127040SQuentin Monnet * consumed. 3003ab127040SQuentin Monnet * 3004ab127040SQuentin Monnet * Note that if a socket closes with the internal counter holding 3005ab127040SQuentin Monnet * a non-zero value, this is not a problem because data is not 3006ab127040SQuentin Monnet * being buffered for *bytes* and is sent as it is received. 3007ab127040SQuentin Monnet * Return 3008ab127040SQuentin Monnet * 0 3009ab127040SQuentin Monnet * 3010bdb7b79bSAndrii Nakryiko * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 3011ab127040SQuentin Monnet * Description 3012ab127040SQuentin Monnet * For socket policies, prevent the execution of the verdict eBPF 3013ab127040SQuentin Monnet * program for message *msg* until *bytes* (byte number) have been 3014ab127040SQuentin Monnet * accumulated. 3015ab127040SQuentin Monnet * 3016ab127040SQuentin Monnet * This can be used when one needs a specific number of bytes 3017ab127040SQuentin Monnet * before a verdict can be assigned, even if the data spans 3018ab127040SQuentin Monnet * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 3019ab127040SQuentin Monnet * case would be a user calling **sendmsg**\ () repeatedly with 3020ab127040SQuentin Monnet * 1-byte long message segments. Obviously, this is bad for 3021ab127040SQuentin Monnet * performance, but it is still valid. If the eBPF program needs 3022ab127040SQuentin Monnet * *bytes* bytes to validate a header, this helper can be used to 3023ab127040SQuentin Monnet * prevent the eBPF program to be called again until *bytes* have 3024ab127040SQuentin Monnet * been accumulated. 3025ab127040SQuentin Monnet * Return 3026ab127040SQuentin Monnet * 0 3027ab127040SQuentin Monnet * 3028bdb7b79bSAndrii Nakryiko * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 3029ab127040SQuentin Monnet * Description 3030ab127040SQuentin Monnet * For socket policies, pull in non-linear data from user space 3031ab127040SQuentin Monnet * for *msg* and set pointers *msg*\ **->data** and *msg*\ 3032ab127040SQuentin Monnet * **->data_end** to *start* and *end* bytes offsets into *msg*, 3033ab127040SQuentin Monnet * respectively. 3034ab127040SQuentin Monnet * 3035ab127040SQuentin Monnet * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3036ab127040SQuentin Monnet * *msg* it can only parse data that the (**data**, **data_end**) 3037ab127040SQuentin Monnet * pointers have already consumed. For **sendmsg**\ () hooks this 3038ab127040SQuentin Monnet * is likely the first scatterlist element. But for calls relying 3039ab127040SQuentin Monnet * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 3040ab127040SQuentin Monnet * be the range (**0**, **0**) because the data is shared with 3041ab127040SQuentin Monnet * user space and by default the objective is to avoid allowing 3042ab127040SQuentin Monnet * user space to modify data while (or after) eBPF verdict is 3043ab127040SQuentin Monnet * being decided. This helper can be used to pull in data and to 3044ab127040SQuentin Monnet * set the start and end pointer to given values. Data will be 3045ab127040SQuentin Monnet * copied if necessary (i.e. if data was not linear and if start 3046ab127040SQuentin Monnet * and end pointers do not point to the same chunk). 3047ab127040SQuentin Monnet * 304832e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3049ab127040SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 3050ab127040SQuentin Monnet * previously done by the verifier are invalidated and must be 3051ab127040SQuentin Monnet * performed again, if the helper is used in combination with 3052ab127040SQuentin Monnet * direct packet access. 3053ab127040SQuentin Monnet * 3054ab127040SQuentin Monnet * All values for *flags* are reserved for future usage, and must 3055ab127040SQuentin Monnet * be left at zero. 3056ab127040SQuentin Monnet * Return 3057ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 3058ab127040SQuentin Monnet * 3059bdb7b79bSAndrii Nakryiko * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 30607aa79a86SQuentin Monnet * Description 30617aa79a86SQuentin Monnet * Bind the socket associated to *ctx* to the address pointed by 30627aa79a86SQuentin Monnet * *addr*, of length *addr_len*. This allows for making outgoing 30637aa79a86SQuentin Monnet * connection from the desired IP address, which can be useful for 30647aa79a86SQuentin Monnet * example when all processes inside a cgroup should use one 30657aa79a86SQuentin Monnet * single IP address on a host that has multiple IP configured. 30667aa79a86SQuentin Monnet * 30677aa79a86SQuentin Monnet * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 30687aa79a86SQuentin Monnet * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 30698086fbafSStanislav Fomichev * **AF_INET6**). It's advised to pass zero port (**sin_port** 30708086fbafSStanislav Fomichev * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 30718086fbafSStanislav Fomichev * behavior and lets the kernel efficiently pick up an unused 30728086fbafSStanislav Fomichev * port as long as 4-tuple is unique. Passing non-zero port might 30738086fbafSStanislav Fomichev * lead to degraded performance. 30747aa79a86SQuentin Monnet * Return 30757aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 30762d020dd7SQuentin Monnet * 3077bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 30782d020dd7SQuentin Monnet * Description 30792d020dd7SQuentin Monnet * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 3080c8741e2bSJesper Dangaard Brouer * possible to both shrink and grow the packet tail. 3081c8741e2bSJesper Dangaard Brouer * Shrink done via *delta* being a negative integer. 30822d020dd7SQuentin Monnet * 308332e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 30842d020dd7SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 30852d020dd7SQuentin Monnet * previously done by the verifier are invalidated and must be 30862d020dd7SQuentin Monnet * performed again, if the helper is used in combination with 30872d020dd7SQuentin Monnet * direct packet access. 30882d020dd7SQuentin Monnet * Return 30892d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 30902d020dd7SQuentin Monnet * 3091bdb7b79bSAndrii Nakryiko * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 30922d020dd7SQuentin Monnet * Description 30932d020dd7SQuentin Monnet * Retrieve the XFRM state (IP transform framework, see also 30942d020dd7SQuentin Monnet * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 30952d020dd7SQuentin Monnet * 30962d020dd7SQuentin Monnet * The retrieved value is stored in the **struct bpf_xfrm_state** 30972d020dd7SQuentin Monnet * pointed by *xfrm_state* and of length *size*. 30982d020dd7SQuentin Monnet * 30992d020dd7SQuentin Monnet * All values for *flags* are reserved for future usage, and must 31002d020dd7SQuentin Monnet * be left at zero. 31012d020dd7SQuentin Monnet * 31022d020dd7SQuentin Monnet * This helper is available only if the kernel was compiled with 31032d020dd7SQuentin Monnet * **CONFIG_XFRM** configuration option. 31042d020dd7SQuentin Monnet * Return 31052d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 3106c195651eSYonghong Song * 3107bdb7b79bSAndrii Nakryiko * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 3108c195651eSYonghong Song * Description 3109c195651eSYonghong Song * Return a user or a kernel stack in bpf program provided buffer. 3110c195651eSYonghong Song * To achieve this, the helper needs *ctx*, which is a pointer 3111c195651eSYonghong Song * to the context on which the tracing program is executed. 3112c195651eSYonghong Song * To store the stacktrace, the bpf program provides *buf* with 3113c195651eSYonghong Song * a nonnegative *size*. 3114c195651eSYonghong Song * 3115c195651eSYonghong Song * The last argument, *flags*, holds the number of stack frames to 3116c195651eSYonghong Song * skip (from 0 to 255), masked with 3117c195651eSYonghong Song * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3118c195651eSYonghong Song * the following flags: 3119c195651eSYonghong Song * 3120c195651eSYonghong Song * **BPF_F_USER_STACK** 3121c195651eSYonghong Song * Collect a user space stack instead of a kernel stack. 3122c195651eSYonghong Song * **BPF_F_USER_BUILD_ID** 3123ca34ce29SDave Marchevsky * Collect (build_id, file_offset) instead of ips for user 3124ca34ce29SDave Marchevsky * stack, only valid if **BPF_F_USER_STACK** is also 3125ca34ce29SDave Marchevsky * specified. 3126ca34ce29SDave Marchevsky * 3127ca34ce29SDave Marchevsky * *file_offset* is an offset relative to the beginning 3128ca34ce29SDave Marchevsky * of the executable or shared object file backing the vma 3129ca34ce29SDave Marchevsky * which the *ip* falls in. It is *not* an offset relative 3130ca34ce29SDave Marchevsky * to that object's base address. Accordingly, it must be 3131ca34ce29SDave Marchevsky * adjusted by adding (sh_addr - sh_offset), where 3132ca34ce29SDave Marchevsky * sh_{addr,offset} correspond to the executable section 3133ca34ce29SDave Marchevsky * containing *file_offset* in the object, for comparisons 3134ca34ce29SDave Marchevsky * to symbols' st_value to be valid. 3135c195651eSYonghong Song * 3136c195651eSYonghong Song * **bpf_get_stack**\ () can collect up to 3137c195651eSYonghong Song * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3138c195651eSYonghong Song * to sufficient large buffer size. Note that 3139c195651eSYonghong Song * this limit can be controlled with the **sysctl** program, and 3140c195651eSYonghong Song * that it should be manually increased in order to profile long 3141c195651eSYonghong Song * user stacks (such as stacks for Java programs). To do so, use: 3142c195651eSYonghong Song * 3143c195651eSYonghong Song * :: 3144c195651eSYonghong Song * 3145c195651eSYonghong Song * # sysctl kernel.perf_event_max_stack=<new value> 3146c195651eSYonghong Song * Return 3147ee2a0988SNamhyung Kim * The non-negative copied *buf* length equal to or less than 3148ee2a0988SNamhyung Kim * *size* on success, or a negative error in case of failure. 31494e1ec56cSDaniel Borkmann * 3150bdb7b79bSAndrii Nakryiko * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 31514e1ec56cSDaniel Borkmann * Description 31524e1ec56cSDaniel Borkmann * This helper is similar to **bpf_skb_load_bytes**\ () in that 31534e1ec56cSDaniel Borkmann * it provides an easy way to load *len* bytes from *offset* 31544e1ec56cSDaniel Borkmann * from the packet associated to *skb*, into the buffer pointed 31554e1ec56cSDaniel Borkmann * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 31564e1ec56cSDaniel Borkmann * a fifth argument *start_header* exists in order to select a 31574e1ec56cSDaniel Borkmann * base offset to start from. *start_header* can be one of: 31584e1ec56cSDaniel Borkmann * 31594e1ec56cSDaniel Borkmann * **BPF_HDR_START_MAC** 31604e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s mac header. 31614e1ec56cSDaniel Borkmann * **BPF_HDR_START_NET** 31624e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s network header. 31634e1ec56cSDaniel Borkmann * 31644e1ec56cSDaniel Borkmann * In general, "direct packet access" is the preferred method to 31654e1ec56cSDaniel Borkmann * access packet data, however, this helper is in particular useful 31664e1ec56cSDaniel Borkmann * in socket filters where *skb*\ **->data** does not always point 31674e1ec56cSDaniel Borkmann * to the start of the mac header and where "direct packet access" 31684e1ec56cSDaniel Borkmann * is not available. 31694e1ec56cSDaniel Borkmann * Return 31704e1ec56cSDaniel Borkmann * 0 on success, or a negative error in case of failure. 31714e1ec56cSDaniel Borkmann * 3172bdb7b79bSAndrii Nakryiko * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 317387f5fc7eSDavid Ahern * Description 317487f5fc7eSDavid Ahern * Do FIB lookup in kernel tables using parameters in *params*. 317587f5fc7eSDavid Ahern * If lookup is successful and result shows packet is to be 317687f5fc7eSDavid Ahern * forwarded, the neighbor tables are searched for the nexthop. 317787f5fc7eSDavid Ahern * If successful (ie., FIB lookup shows forwarding and nexthop 3178fa898d76SDavid Ahern * is resolved), the nexthop address is returned in ipv4_dst 3179fa898d76SDavid Ahern * or ipv6_dst based on family, smac is set to mac address of 3180fa898d76SDavid Ahern * egress device, dmac is set to nexthop mac address, rt_metric 31814c79579bSDavid Ahern * is set to metric from route (IPv4/IPv6 only), and ifindex 31824c79579bSDavid Ahern * is set to the device index of the nexthop from the FIB lookup. 318387f5fc7eSDavid Ahern * 318487f5fc7eSDavid Ahern * *plen* argument is the size of the passed in struct. 31857a279e93SQuentin Monnet * *flags* argument can be a combination of one or more of the 31867a279e93SQuentin Monnet * following values: 318787f5fc7eSDavid Ahern * 31887a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_DIRECT** 31897a279e93SQuentin Monnet * Do a direct table lookup vs full lookup using FIB 31907a279e93SQuentin Monnet * rules. 31918ad77e72SLouis DeLosSantos * **BPF_FIB_LOOKUP_TBID** 31928ad77e72SLouis DeLosSantos * Used with BPF_FIB_LOOKUP_DIRECT. 31938ad77e72SLouis DeLosSantos * Use the routing table ID present in *params*->tbid 31948ad77e72SLouis DeLosSantos * for the fib lookup. 31957a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_OUTPUT** 31967a279e93SQuentin Monnet * Perform lookup from an egress perspective (default is 31977a279e93SQuentin Monnet * ingress). 319831de4105SMartin KaFai Lau * **BPF_FIB_LOOKUP_SKIP_NEIGH** 319931de4105SMartin KaFai Lau * Skip the neighbour table lookup. *params*->dmac 320031de4105SMartin KaFai Lau * and *params*->smac will not be set as output. A common 320131de4105SMartin KaFai Lau * use case is to call **bpf_redirect_neigh**\ () after 320231de4105SMartin KaFai Lau * doing **bpf_fib_lookup**\ (). 320387f5fc7eSDavid Ahern * 320487f5fc7eSDavid Ahern * *ctx* is either **struct xdp_md** for XDP programs or 320587f5fc7eSDavid Ahern * **struct sk_buff** tc cls_act programs. 320687f5fc7eSDavid Ahern * Return 32074c79579bSDavid Ahern * * < 0 if any input argument is invalid 32084c79579bSDavid Ahern * * 0 on success (packet is forwarded, nexthop neighbor exists) 32094c79579bSDavid Ahern * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 32102bae79d2SQuentin Monnet * packet is not forwarded or needs assist from full stack 321181110384SJohn Fastabend * 3212e1850ea9SJesper Dangaard Brouer * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3213e1850ea9SJesper Dangaard Brouer * was exceeded and output params->mtu_result contains the MTU. 3214e1850ea9SJesper Dangaard Brouer * 3215bdb7b79bSAndrii Nakryiko * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 321681110384SJohn Fastabend * Description 321781110384SJohn Fastabend * Add an entry to, or update a sockhash *map* referencing sockets. 321881110384SJohn Fastabend * The *skops* is used as a new value for the entry associated to 321981110384SJohn Fastabend * *key*. *flags* is one of: 322081110384SJohn Fastabend * 322181110384SJohn Fastabend * **BPF_NOEXIST** 322281110384SJohn Fastabend * The entry for *key* must not exist in the map. 322381110384SJohn Fastabend * **BPF_EXIST** 322481110384SJohn Fastabend * The entry for *key* must already exist in the map. 322581110384SJohn Fastabend * **BPF_ANY** 322681110384SJohn Fastabend * No condition on the existence of the entry for *key*. 322781110384SJohn Fastabend * 322881110384SJohn Fastabend * If the *map* has eBPF programs (parser and verdict), those will 322981110384SJohn Fastabend * be inherited by the socket being added. If the socket is 323081110384SJohn Fastabend * already attached to eBPF programs, this results in an error. 323181110384SJohn Fastabend * Return 323281110384SJohn Fastabend * 0 on success, or a negative error in case of failure. 323381110384SJohn Fastabend * 3234bdb7b79bSAndrii Nakryiko * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 323581110384SJohn Fastabend * Description 323681110384SJohn Fastabend * This helper is used in programs implementing policies at the 323781110384SJohn Fastabend * socket level. If the message *msg* is allowed to pass (i.e. if 323881110384SJohn Fastabend * the verdict eBPF program returns **SK_PASS**), redirect it to 323981110384SJohn Fastabend * the socket referenced by *map* (of type 324081110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 324181110384SJohn Fastabend * egress interfaces can be used for redirection. The 324281110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 324381110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 324481110384SJohn Fastabend * egress path otherwise). This is the only flag supported for now. 324581110384SJohn Fastabend * Return 324681110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 324781110384SJohn Fastabend * 3248bdb7b79bSAndrii Nakryiko * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 324981110384SJohn Fastabend * Description 325081110384SJohn Fastabend * This helper is used in programs implementing policies at the 325181110384SJohn Fastabend * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 325249f3d12bSJakub Wilk * if the verdict eBPF program returns **SK_PASS**), redirect it 325381110384SJohn Fastabend * to the socket referenced by *map* (of type 325481110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 325581110384SJohn Fastabend * egress interfaces can be used for redirection. The 325681110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 325781110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 325881110384SJohn Fastabend * egress otherwise). This is the only flag supported for now. 325981110384SJohn Fastabend * Return 326081110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 3261fe94cc29SMathieu Xhonneux * 3262bdb7b79bSAndrii Nakryiko * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3263fe94cc29SMathieu Xhonneux * Description 3264fe94cc29SMathieu Xhonneux * Encapsulate the packet associated to *skb* within a Layer 3 3265fe94cc29SMathieu Xhonneux * protocol header. This header is provided in the buffer at 3266fe94cc29SMathieu Xhonneux * address *hdr*, with *len* its size in bytes. *type* indicates 3267fe94cc29SMathieu Xhonneux * the protocol of the header and can be one of: 3268fe94cc29SMathieu Xhonneux * 3269fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6** 3270fe94cc29SMathieu Xhonneux * IPv6 encapsulation with Segment Routing Header 3271fe94cc29SMathieu Xhonneux * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3272fe94cc29SMathieu Xhonneux * the IPv6 header is computed by the kernel. 3273fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6_INLINE** 3274fe94cc29SMathieu Xhonneux * Only works if *skb* contains an IPv6 packet. Insert a 3275fe94cc29SMathieu Xhonneux * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3276fe94cc29SMathieu Xhonneux * the IPv6 header. 32773e0bd37cSPeter Oskolkov * **BPF_LWT_ENCAP_IP** 32783e0bd37cSPeter Oskolkov * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 32793e0bd37cSPeter Oskolkov * must be IPv4 or IPv6, followed by zero or more 328080867c5eSQuentin Monnet * additional headers, up to **LWT_BPF_MAX_HEADROOM** 328180867c5eSQuentin Monnet * total bytes in all prepended headers. Please note that 328280867c5eSQuentin Monnet * if **skb_is_gso**\ (*skb*) is true, no more than two 328380867c5eSQuentin Monnet * headers can be prepended, and the inner header, if 328480867c5eSQuentin Monnet * present, should be either GRE or UDP/GUE. 32853e0bd37cSPeter Oskolkov * 328680867c5eSQuentin Monnet * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 328780867c5eSQuentin Monnet * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 328880867c5eSQuentin Monnet * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 328980867c5eSQuentin Monnet * **BPF_PROG_TYPE_LWT_XMIT**. 3290fe94cc29SMathieu Xhonneux * 329132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3292fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3293fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3294fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3295fe94cc29SMathieu Xhonneux * direct packet access. 3296fe94cc29SMathieu Xhonneux * Return 3297fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3298fe94cc29SMathieu Xhonneux * 3299bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3300fe94cc29SMathieu Xhonneux * Description 3301fe94cc29SMathieu Xhonneux * Store *len* bytes from address *from* into the packet 3302fe94cc29SMathieu Xhonneux * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3303fe94cc29SMathieu Xhonneux * inside the outermost IPv6 Segment Routing Header can be 3304fe94cc29SMathieu Xhonneux * modified through this helper. 3305fe94cc29SMathieu Xhonneux * 330632e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3307fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3308fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3309fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3310fe94cc29SMathieu Xhonneux * direct packet access. 3311fe94cc29SMathieu Xhonneux * Return 3312fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3313fe94cc29SMathieu Xhonneux * 3314bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3315fe94cc29SMathieu Xhonneux * Description 3316fe94cc29SMathieu Xhonneux * Adjust the size allocated to TLVs in the outermost IPv6 3317fe94cc29SMathieu Xhonneux * Segment Routing Header contained in the packet associated to 3318fe94cc29SMathieu Xhonneux * *skb*, at position *offset* by *delta* bytes. Only offsets 3319fe94cc29SMathieu Xhonneux * after the segments are accepted. *delta* can be as well 3320fe94cc29SMathieu Xhonneux * positive (growing) as negative (shrinking). 3321fe94cc29SMathieu Xhonneux * 332232e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3323fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3324fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3325fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3326fe94cc29SMathieu Xhonneux * direct packet access. 3327fe94cc29SMathieu Xhonneux * Return 3328fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3329fe94cc29SMathieu Xhonneux * 3330bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3331fe94cc29SMathieu Xhonneux * Description 3332fe94cc29SMathieu Xhonneux * Apply an IPv6 Segment Routing action of type *action* to the 3333fe94cc29SMathieu Xhonneux * packet associated to *skb*. Each action takes a parameter 3334fe94cc29SMathieu Xhonneux * contained at address *param*, and of length *param_len* bytes. 3335fe94cc29SMathieu Xhonneux * *action* can be one of: 3336fe94cc29SMathieu Xhonneux * 3337fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_X** 3338fe94cc29SMathieu Xhonneux * End.X action: Endpoint with Layer-3 cross-connect. 3339fe94cc29SMathieu Xhonneux * Type of *param*: **struct in6_addr**. 3340fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_T** 3341fe94cc29SMathieu Xhonneux * End.T action: Endpoint with specific IPv6 table lookup. 3342fe94cc29SMathieu Xhonneux * Type of *param*: **int**. 3343fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6** 3344fe94cc29SMathieu Xhonneux * End.B6 action: Endpoint bound to an SRv6 policy. 334580867c5eSQuentin Monnet * Type of *param*: **struct ipv6_sr_hdr**. 3346fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3347fe94cc29SMathieu Xhonneux * End.B6.Encap action: Endpoint bound to an SRv6 3348fe94cc29SMathieu Xhonneux * encapsulation policy. 334980867c5eSQuentin Monnet * Type of *param*: **struct ipv6_sr_hdr**. 3350fe94cc29SMathieu Xhonneux * 335132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3352fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3353fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3354fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3355fe94cc29SMathieu Xhonneux * direct packet access. 3356fe94cc29SMathieu Xhonneux * Return 3357fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3358f4364dcfSSean Young * 3359bdb7b79bSAndrii Nakryiko * long bpf_rc_repeat(void *ctx) 336062369db2SQuentin Monnet * Description 336162369db2SQuentin Monnet * This helper is used in programs implementing IR decoding, to 336262369db2SQuentin Monnet * report a successfully decoded repeat key message. This delays 336362369db2SQuentin Monnet * the generation of a key up event for previously generated 336462369db2SQuentin Monnet * key down event. 336562369db2SQuentin Monnet * 336662369db2SQuentin Monnet * Some IR protocols like NEC have a special IR message for 336762369db2SQuentin Monnet * repeating last button, for when a button is held down. 336862369db2SQuentin Monnet * 336962369db2SQuentin Monnet * The *ctx* should point to the lirc sample as passed into 337062369db2SQuentin Monnet * the program. 337162369db2SQuentin Monnet * 337262369db2SQuentin Monnet * This helper is only available is the kernel was compiled with 337362369db2SQuentin Monnet * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 337462369db2SQuentin Monnet * "**y**". 337562369db2SQuentin Monnet * Return 337662369db2SQuentin Monnet * 0 337762369db2SQuentin Monnet * 3378bdb7b79bSAndrii Nakryiko * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3379f4364dcfSSean Young * Description 3380f4364dcfSSean Young * This helper is used in programs implementing IR decoding, to 3381f4364dcfSSean Young * report a successfully decoded key press with *scancode*, 3382f4364dcfSSean Young * *toggle* value in the given *protocol*. The scancode will be 3383f4364dcfSSean Young * translated to a keycode using the rc keymap, and reported as 3384f4364dcfSSean Young * an input key down event. After a period a key up event is 3385f4364dcfSSean Young * generated. This period can be extended by calling either 338690b1023fSQuentin Monnet * **bpf_rc_keydown**\ () again with the same values, or calling 338790b1023fSQuentin Monnet * **bpf_rc_repeat**\ (). 3388f4364dcfSSean Young * 3389f4364dcfSSean Young * Some protocols include a toggle bit, in case the button was 3390f4364dcfSSean Young * released and pressed again between consecutive scancodes. 3391f4364dcfSSean Young * 3392f4364dcfSSean Young * The *ctx* should point to the lirc sample as passed into 3393f4364dcfSSean Young * the program. 3394f4364dcfSSean Young * 3395f4364dcfSSean Young * The *protocol* is the decoded protocol number (see 3396f4364dcfSSean Young * **enum rc_proto** for some predefined values). 3397f4364dcfSSean Young * 3398f4364dcfSSean Young * This helper is only available is the kernel was compiled with 3399f4364dcfSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3400f4364dcfSSean Young * "**y**". 3401f4364dcfSSean Young * Return 3402f4364dcfSSean Young * 0 3403f4364dcfSSean Young * 340462369db2SQuentin Monnet * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3405cb20b08eSDaniel Borkmann * Description 3406cb20b08eSDaniel Borkmann * Return the cgroup v2 id of the socket associated with the *skb*. 3407cb20b08eSDaniel Borkmann * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3408cb20b08eSDaniel Borkmann * helper for cgroup v1 by providing a tag resp. identifier that 3409cb20b08eSDaniel Borkmann * can be matched on or used for map lookups e.g. to implement 3410cb20b08eSDaniel Borkmann * policy. The cgroup v2 id of a given path in the hierarchy is 3411cb20b08eSDaniel Borkmann * exposed in user space through the f_handle API in order to get 3412cb20b08eSDaniel Borkmann * to the same 64-bit id. 3413cb20b08eSDaniel Borkmann * 3414cb20b08eSDaniel Borkmann * This helper can be used on TC egress path, but not on ingress, 3415cb20b08eSDaniel Borkmann * and is available only if the kernel was compiled with the 3416cb20b08eSDaniel Borkmann * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3417cb20b08eSDaniel Borkmann * Return 3418cb20b08eSDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 3419bf6fa2c8SYonghong Song * 3420bf6fa2c8SYonghong Song * u64 bpf_get_current_cgroup_id(void) 3421e40fbbf0SUsama Arif * Description 3422e40fbbf0SUsama Arif * Get the current cgroup id based on the cgroup within which 3423e40fbbf0SUsama Arif * the current task is running. 3424bf6fa2c8SYonghong Song * Return 3425bf6fa2c8SYonghong Song * A 64-bit integer containing the current cgroup id based 3426bf6fa2c8SYonghong Song * on the cgroup within which the current task is running. 3427cd339431SRoman Gushchin * 342862369db2SQuentin Monnet * void *bpf_get_local_storage(void *map, u64 flags) 3429cd339431SRoman Gushchin * Description 3430cd339431SRoman Gushchin * Get the pointer to the local storage area. 3431cd339431SRoman Gushchin * The type and the size of the local storage is defined 3432cd339431SRoman Gushchin * by the *map* argument. 3433cd339431SRoman Gushchin * The *flags* meaning is specific for each map type, 3434cd339431SRoman Gushchin * and has to be 0 for cgroup local storage. 3435cd339431SRoman Gushchin * 343690b1023fSQuentin Monnet * Depending on the BPF program type, a local storage area 343790b1023fSQuentin Monnet * can be shared between multiple instances of the BPF program, 3438cd339431SRoman Gushchin * running simultaneously. 3439cd339431SRoman Gushchin * 3440cd339431SRoman Gushchin * A user should care about the synchronization by himself. 344191c960b0SBrendan Jackman * For example, by using the **BPF_ATOMIC** instructions to alter 3442cd339431SRoman Gushchin * the shared data. 3443cd339431SRoman Gushchin * Return 344490b1023fSQuentin Monnet * A pointer to the local storage area. 34452dbb9b9eSMartin KaFai Lau * 3446bdb7b79bSAndrii Nakryiko * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 34472dbb9b9eSMartin KaFai Lau * Description 344890b1023fSQuentin Monnet * Select a **SO_REUSEPORT** socket from a 3449f170acdaSKuniyuki Iwashima * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 345090b1023fSQuentin Monnet * It checks the selected socket is matching the incoming 345190b1023fSQuentin Monnet * request in the socket buffer. 34522dbb9b9eSMartin KaFai Lau * Return 34532dbb9b9eSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 34546acc9b43SJoe Stringer * 345562369db2SQuentin Monnet * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 345662369db2SQuentin Monnet * Description 345762369db2SQuentin Monnet * Return id of cgroup v2 that is ancestor of cgroup associated 345862369db2SQuentin Monnet * with the *skb* at the *ancestor_level*. The root cgroup is at 345962369db2SQuentin Monnet * *ancestor_level* zero and each step down the hierarchy 346062369db2SQuentin Monnet * increments the level. If *ancestor_level* == level of cgroup 346162369db2SQuentin Monnet * associated with *skb*, then return value will be same as that 346262369db2SQuentin Monnet * of **bpf_skb_cgroup_id**\ (). 346362369db2SQuentin Monnet * 346462369db2SQuentin Monnet * The helper is useful to implement policies based on cgroups 346562369db2SQuentin Monnet * that are upper in hierarchy than immediate cgroup associated 346662369db2SQuentin Monnet * with *skb*. 346762369db2SQuentin Monnet * 346862369db2SQuentin Monnet * The format of returned id and helper limitations are same as in 346962369db2SQuentin Monnet * **bpf_skb_cgroup_id**\ (). 347062369db2SQuentin Monnet * Return 347162369db2SQuentin Monnet * The id is returned or 0 in case the id could not be retrieved. 347262369db2SQuentin Monnet * 3473f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 34746acc9b43SJoe Stringer * Description 34756acc9b43SJoe Stringer * Look for TCP socket matching *tuple*, optionally in a child 34766acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 347790b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 34786acc9b43SJoe Stringer * 34796acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 34806acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 34816acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 34826acc9b43SJoe Stringer * 34836acc9b43SJoe Stringer * *tuple_size* must be one of: 34846acc9b43SJoe Stringer * 34856acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 34866acc9b43SJoe Stringer * Look for an IPv4 socket. 34876acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 34886acc9b43SJoe Stringer * Look for an IPv6 socket. 34896acc9b43SJoe Stringer * 3490f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 3491bfdfa517SRandy Dunlap * socket lookup table in the netns associated with the *ctx* 3492f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 3493f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 3494f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 3495f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 3496f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 3497f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 34986acc9b43SJoe Stringer * 34996acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 35006acc9b43SJoe Stringer * be left at zero. 35016acc9b43SJoe Stringer * 35026acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 35036acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 35046acc9b43SJoe Stringer * Return 35050bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 35060bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 350780867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 350880867c5eSQuentin Monnet * tuple. 35096acc9b43SJoe Stringer * 3510f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 35116acc9b43SJoe Stringer * Description 35126acc9b43SJoe Stringer * Look for UDP socket matching *tuple*, optionally in a child 35136acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 351490b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 35156acc9b43SJoe Stringer * 35166acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 35176acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 35186acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 35196acc9b43SJoe Stringer * 35206acc9b43SJoe Stringer * *tuple_size* must be one of: 35216acc9b43SJoe Stringer * 35226acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 35236acc9b43SJoe Stringer * Look for an IPv4 socket. 35246acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 35256acc9b43SJoe Stringer * Look for an IPv6 socket. 35266acc9b43SJoe Stringer * 3527f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 3528bfdfa517SRandy Dunlap * socket lookup table in the netns associated with the *ctx* 3529f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 3530f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 3531f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 3532f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 3533f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 3534f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 35356acc9b43SJoe Stringer * 35366acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 35376acc9b43SJoe Stringer * be left at zero. 35386acc9b43SJoe Stringer * 35396acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 35406acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 35416acc9b43SJoe Stringer * Return 35420bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 35430bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 354480867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 354580867c5eSQuentin Monnet * tuple. 35466acc9b43SJoe Stringer * 3547a5fa25adSMartin KaFai Lau * long bpf_sk_release(void *sock) 35486acc9b43SJoe Stringer * Description 354990b1023fSQuentin Monnet * Release the reference held by *sock*. *sock* must be a 355090b1023fSQuentin Monnet * non-**NULL** pointer that was returned from 355190b1023fSQuentin Monnet * **bpf_sk_lookup_xxx**\ (). 355290b1023fSQuentin Monnet * Return 355390b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 355490b1023fSQuentin Monnet * 3555bdb7b79bSAndrii Nakryiko * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 355662369db2SQuentin Monnet * Description 355762369db2SQuentin Monnet * Push an element *value* in *map*. *flags* is one of: 355862369db2SQuentin Monnet * 355962369db2SQuentin Monnet * **BPF_EXIST** 356062369db2SQuentin Monnet * If the queue/stack is full, the oldest element is 356162369db2SQuentin Monnet * removed to make room for this. 356262369db2SQuentin Monnet * Return 356362369db2SQuentin Monnet * 0 on success, or a negative error in case of failure. 356462369db2SQuentin Monnet * 3565bdb7b79bSAndrii Nakryiko * long bpf_map_pop_elem(struct bpf_map *map, void *value) 356690b1023fSQuentin Monnet * Description 356790b1023fSQuentin Monnet * Pop an element from *map*. 356890b1023fSQuentin Monnet * Return 356990b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 357090b1023fSQuentin Monnet * 3571bdb7b79bSAndrii Nakryiko * long bpf_map_peek_elem(struct bpf_map *map, void *value) 357290b1023fSQuentin Monnet * Description 357390b1023fSQuentin Monnet * Get an element from *map* without removing it. 35746acc9b43SJoe Stringer * Return 35756acc9b43SJoe Stringer * 0 on success, or a negative error in case of failure. 35766fff607eSJohn Fastabend * 3577bdb7b79bSAndrii Nakryiko * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 35786fff607eSJohn Fastabend * Description 357990b1023fSQuentin Monnet * For socket policies, insert *len* bytes into *msg* at offset 35806fff607eSJohn Fastabend * *start*. 35816fff607eSJohn Fastabend * 35826fff607eSJohn Fastabend * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 358390b1023fSQuentin Monnet * *msg* it may want to insert metadata or options into the *msg*. 35846fff607eSJohn Fastabend * This can later be read and used by any of the lower layer BPF 35856fff607eSJohn Fastabend * hooks. 35866fff607eSJohn Fastabend * 35876fff607eSJohn Fastabend * This helper may fail if under memory pressure (a malloc 35886fff607eSJohn Fastabend * fails) in these cases BPF programs will get an appropriate 35896fff607eSJohn Fastabend * error and BPF programs will need to handle them. 35906fff607eSJohn Fastabend * Return 35916fff607eSJohn Fastabend * 0 on success, or a negative error in case of failure. 35927246d8edSJohn Fastabend * 3593bdb7b79bSAndrii Nakryiko * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 35947246d8edSJohn Fastabend * Description 35955f0e5412SAndrii Nakryiko * Will remove *len* bytes from a *msg* starting at byte *start*. 35967246d8edSJohn Fastabend * This may result in **ENOMEM** errors under certain situations if 35977246d8edSJohn Fastabend * an allocation and copy are required due to a full ring buffer. 35987246d8edSJohn Fastabend * However, the helper will try to avoid doing the allocation 35997246d8edSJohn Fastabend * if possible. Other errors can occur if input parameters are 360090b1023fSQuentin Monnet * invalid either due to *start* byte not being valid part of *msg* 36017246d8edSJohn Fastabend * payload and/or *pop* value being to large. 36027246d8edSJohn Fastabend * Return 360390b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 360401d3240aSSean Young * 3605bdb7b79bSAndrii Nakryiko * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 360601d3240aSSean Young * Description 360701d3240aSSean Young * This helper is used in programs implementing IR decoding, to 360801d3240aSSean Young * report a successfully decoded pointer movement. 360901d3240aSSean Young * 361001d3240aSSean Young * The *ctx* should point to the lirc sample as passed into 361101d3240aSSean Young * the program. 361201d3240aSSean Young * 361301d3240aSSean Young * This helper is only available is the kernel was compiled with 361401d3240aSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 361501d3240aSSean Young * "**y**". 361601d3240aSSean Young * Return 361701d3240aSSean Young * 0 361846f8bc92SMartin KaFai Lau * 3619bdb7b79bSAndrii Nakryiko * long bpf_spin_lock(struct bpf_spin_lock *lock) 36200eb09785SQuentin Monnet * Description 36210eb09785SQuentin Monnet * Acquire a spinlock represented by the pointer *lock*, which is 36220eb09785SQuentin Monnet * stored as part of a value of a map. Taking the lock allows to 36230eb09785SQuentin Monnet * safely update the rest of the fields in that value. The 36240eb09785SQuentin Monnet * spinlock can (and must) later be released with a call to 36250eb09785SQuentin Monnet * **bpf_spin_unlock**\ (\ *lock*\ ). 36260eb09785SQuentin Monnet * 36270eb09785SQuentin Monnet * Spinlocks in BPF programs come with a number of restrictions 36280eb09785SQuentin Monnet * and constraints: 36290eb09785SQuentin Monnet * 36300eb09785SQuentin Monnet * * **bpf_spin_lock** objects are only allowed inside maps of 36310eb09785SQuentin Monnet * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 36320eb09785SQuentin Monnet * list could be extended in the future). 36330eb09785SQuentin Monnet * * BTF description of the map is mandatory. 36340eb09785SQuentin Monnet * * The BPF program can take ONE lock at a time, since taking two 36350eb09785SQuentin Monnet * or more could cause dead locks. 36360eb09785SQuentin Monnet * * Only one **struct bpf_spin_lock** is allowed per map element. 36370eb09785SQuentin Monnet * * When the lock is taken, calls (either BPF to BPF or helpers) 36380eb09785SQuentin Monnet * are not allowed. 36390eb09785SQuentin Monnet * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 36400eb09785SQuentin Monnet * allowed inside a spinlock-ed region. 36410eb09785SQuentin Monnet * * The BPF program MUST call **bpf_spin_unlock**\ () to release 36420eb09785SQuentin Monnet * the lock, on all execution paths, before it returns. 36430eb09785SQuentin Monnet * * The BPF program can access **struct bpf_spin_lock** only via 36440eb09785SQuentin Monnet * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 36450eb09785SQuentin Monnet * helpers. Loading or storing data into the **struct 36460eb09785SQuentin Monnet * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 36470eb09785SQuentin Monnet * * To use the **bpf_spin_lock**\ () helper, the BTF description 36480eb09785SQuentin Monnet * of the map value must be a struct and have **struct 36490eb09785SQuentin Monnet * bpf_spin_lock** *anyname*\ **;** field at the top level. 36500eb09785SQuentin Monnet * Nested lock inside another struct is not allowed. 36510eb09785SQuentin Monnet * * The **struct bpf_spin_lock** *lock* field in a map value must 36520eb09785SQuentin Monnet * be aligned on a multiple of 4 bytes in that value. 36530eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 36540eb09785SQuentin Monnet * the **bpf_spin_lock** field to user space. 36550eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 36560eb09785SQuentin Monnet * a BPF program, do not update the **bpf_spin_lock** field. 36570eb09785SQuentin Monnet * * **bpf_spin_lock** cannot be on the stack or inside a 36580eb09785SQuentin Monnet * networking packet (it can only be inside of a map values). 36590eb09785SQuentin Monnet * * **bpf_spin_lock** is available to root only. 36600eb09785SQuentin Monnet * * Tracing programs and socket filter programs cannot use 36610eb09785SQuentin Monnet * **bpf_spin_lock**\ () due to insufficient preemption checks 36620eb09785SQuentin Monnet * (but this may change in the future). 36630eb09785SQuentin Monnet * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 36640eb09785SQuentin Monnet * Return 36650eb09785SQuentin Monnet * 0 36660eb09785SQuentin Monnet * 3667bdb7b79bSAndrii Nakryiko * long bpf_spin_unlock(struct bpf_spin_lock *lock) 36680eb09785SQuentin Monnet * Description 36690eb09785SQuentin Monnet * Release the *lock* previously locked by a call to 36700eb09785SQuentin Monnet * **bpf_spin_lock**\ (\ *lock*\ ). 36710eb09785SQuentin Monnet * Return 36720eb09785SQuentin Monnet * 0 36730eb09785SQuentin Monnet * 367446f8bc92SMartin KaFai Lau * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 367546f8bc92SMartin KaFai Lau * Description 367646f8bc92SMartin KaFai Lau * This helper gets a **struct bpf_sock** pointer such 367762369db2SQuentin Monnet * that all the fields in this **bpf_sock** can be accessed. 367846f8bc92SMartin KaFai Lau * Return 367962369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 368046f8bc92SMartin KaFai Lau * case of failure. 3681655a51e5SMartin KaFai Lau * 3682655a51e5SMartin KaFai Lau * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3683655a51e5SMartin KaFai Lau * Description 3684655a51e5SMartin KaFai Lau * This helper gets a **struct bpf_tcp_sock** pointer from a 3685655a51e5SMartin KaFai Lau * **struct bpf_sock** pointer. 3686655a51e5SMartin KaFai Lau * Return 368762369db2SQuentin Monnet * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3688655a51e5SMartin KaFai Lau * case of failure. 3689f7c917baSbrakmo * 3690bdb7b79bSAndrii Nakryiko * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3691f7c917baSbrakmo * Description 369262369db2SQuentin Monnet * Set ECN (Explicit Congestion Notification) field of IP header 369362369db2SQuentin Monnet * to **CE** (Congestion Encountered) if current value is **ECT** 369462369db2SQuentin Monnet * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 369562369db2SQuentin Monnet * and IPv4. 3696f7c917baSbrakmo * Return 369762369db2SQuentin Monnet * 1 if the **CE** flag is set (either by the current helper call 369862369db2SQuentin Monnet * or because it was already present), 0 if it is not set. 3699dbafd7ddSMartin KaFai Lau * 3700dbafd7ddSMartin KaFai Lau * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3701dbafd7ddSMartin KaFai Lau * Description 370262369db2SQuentin Monnet * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 370362369db2SQuentin Monnet * **bpf_sk_release**\ () is unnecessary and not allowed. 3704dbafd7ddSMartin KaFai Lau * Return 370562369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 3706dbafd7ddSMartin KaFai Lau * case of failure. 3707edbf8c01SLorenz Bauer * 3708edbf8c01SLorenz Bauer * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3709edbf8c01SLorenz Bauer * Description 3710edbf8c01SLorenz Bauer * Look for TCP socket matching *tuple*, optionally in a child 3711edbf8c01SLorenz Bauer * network namespace *netns*. The return value must be checked, 3712edbf8c01SLorenz Bauer * and if non-**NULL**, released via **bpf_sk_release**\ (). 3713edbf8c01SLorenz Bauer * 371480867c5eSQuentin Monnet * This function is identical to **bpf_sk_lookup_tcp**\ (), except 371580867c5eSQuentin Monnet * that it also returns timewait or request sockets. Use 371680867c5eSQuentin Monnet * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 371780867c5eSQuentin Monnet * full structure. 3718edbf8c01SLorenz Bauer * 3719edbf8c01SLorenz Bauer * This helper is available only if the kernel was compiled with 3720edbf8c01SLorenz Bauer * **CONFIG_NET** configuration option. 3721edbf8c01SLorenz Bauer * Return 3722edbf8c01SLorenz Bauer * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3723edbf8c01SLorenz Bauer * For sockets with reuseport option, the **struct bpf_sock** 372480867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 372580867c5eSQuentin Monnet * tuple. 372639904084SLorenz Bauer * 3727c0df236eSMartin KaFai Lau * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 372839904084SLorenz Bauer * Description 372980867c5eSQuentin Monnet * Check whether *iph* and *th* contain a valid SYN cookie ACK for 373080867c5eSQuentin Monnet * the listening socket in *sk*. 373139904084SLorenz Bauer * 373280867c5eSQuentin Monnet * *iph* points to the start of the IPv4 or IPv6 header, while 373380867c5eSQuentin Monnet * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3734ac80287aSMaxim Mikityanskiy * **sizeof**\ (**struct ipv6hdr**). 373539904084SLorenz Bauer * 373680867c5eSQuentin Monnet * *th* points to the start of the TCP header, while *th_len* 3737ac80287aSMaxim Mikityanskiy * contains the length of the TCP header (at least 3738ac80287aSMaxim Mikityanskiy * **sizeof**\ (**struct tcphdr**)). 373939904084SLorenz Bauer * Return 374080867c5eSQuentin Monnet * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 374180867c5eSQuentin Monnet * error otherwise. 3742808649fbSAndrey Ignatov * 3743bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3744808649fbSAndrey Ignatov * Description 3745808649fbSAndrey Ignatov * Get name of sysctl in /proc/sys/ and copy it into provided by 3746808649fbSAndrey Ignatov * program buffer *buf* of size *buf_len*. 3747808649fbSAndrey Ignatov * 3748808649fbSAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 3749808649fbSAndrey Ignatov * 3750808649fbSAndrey Ignatov * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3751808649fbSAndrey Ignatov * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3752808649fbSAndrey Ignatov * only (e.g. "tcp_mem"). 3753808649fbSAndrey Ignatov * Return 3754808649fbSAndrey Ignatov * Number of character copied (not including the trailing NUL). 3755808649fbSAndrey Ignatov * 3756808649fbSAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3757808649fbSAndrey Ignatov * truncated name in this case). 37581d11b301SAndrey Ignatov * 3759bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 37601d11b301SAndrey Ignatov * Description 37611d11b301SAndrey Ignatov * Get current value of sysctl as it is presented in /proc/sys 37621d11b301SAndrey Ignatov * (incl. newline, etc), and copy it as a string into provided 37631d11b301SAndrey Ignatov * by program buffer *buf* of size *buf_len*. 37641d11b301SAndrey Ignatov * 37651d11b301SAndrey Ignatov * The whole value is copied, no matter what file position user 37661d11b301SAndrey Ignatov * space issued e.g. sys_read at. 37671d11b301SAndrey Ignatov * 37681d11b301SAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 37691d11b301SAndrey Ignatov * Return 37701d11b301SAndrey Ignatov * Number of character copied (not including the trailing NUL). 37711d11b301SAndrey Ignatov * 37721d11b301SAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 37731d11b301SAndrey Ignatov * truncated name in this case). 37741d11b301SAndrey Ignatov * 37751d11b301SAndrey Ignatov * **-EINVAL** if current value was unavailable, e.g. because 37761d11b301SAndrey Ignatov * sysctl is uninitialized and read returns -EIO for it. 37774e63acdfSAndrey Ignatov * 3778bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 37794e63acdfSAndrey Ignatov * Description 37804e63acdfSAndrey Ignatov * Get new value being written by user space to sysctl (before 37814e63acdfSAndrey Ignatov * the actual write happens) and copy it as a string into 37824e63acdfSAndrey Ignatov * provided by program buffer *buf* of size *buf_len*. 37834e63acdfSAndrey Ignatov * 37844e63acdfSAndrey Ignatov * User space may write new value at file position > 0. 37854e63acdfSAndrey Ignatov * 37864e63acdfSAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 37874e63acdfSAndrey Ignatov * Return 37884e63acdfSAndrey Ignatov * Number of character copied (not including the trailing NUL). 37894e63acdfSAndrey Ignatov * 37904e63acdfSAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 37914e63acdfSAndrey Ignatov * truncated name in this case). 37924e63acdfSAndrey Ignatov * 37934e63acdfSAndrey Ignatov * **-EINVAL** if sysctl is being read. 37944e63acdfSAndrey Ignatov * 3795bdb7b79bSAndrii Nakryiko * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 37964e63acdfSAndrey Ignatov * Description 37974e63acdfSAndrey Ignatov * Override new value being written by user space to sysctl with 37984e63acdfSAndrey Ignatov * value provided by program in buffer *buf* of size *buf_len*. 37994e63acdfSAndrey Ignatov * 38004e63acdfSAndrey Ignatov * *buf* should contain a string in same form as provided by user 38014e63acdfSAndrey Ignatov * space on sysctl write. 38024e63acdfSAndrey Ignatov * 38034e63acdfSAndrey Ignatov * User space may write new value at file position > 0. To override 38044e63acdfSAndrey Ignatov * the whole sysctl value file position should be set to zero. 38054e63acdfSAndrey Ignatov * Return 38064e63acdfSAndrey Ignatov * 0 on success. 38074e63acdfSAndrey Ignatov * 38084e63acdfSAndrey Ignatov * **-E2BIG** if the *buf_len* is too big. 38094e63acdfSAndrey Ignatov * 38104e63acdfSAndrey Ignatov * **-EINVAL** if sysctl is being read. 3811d7a4cb9bSAndrey Ignatov * 3812bdb7b79bSAndrii Nakryiko * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3813d7a4cb9bSAndrey Ignatov * Description 3814d7a4cb9bSAndrey Ignatov * Convert the initial part of the string from buffer *buf* of 3815d7a4cb9bSAndrey Ignatov * size *buf_len* to a long integer according to the given base 3816d7a4cb9bSAndrey Ignatov * and save the result in *res*. 3817d7a4cb9bSAndrey Ignatov * 3818d7a4cb9bSAndrey Ignatov * The string may begin with an arbitrary amount of white space 381980867c5eSQuentin Monnet * (as determined by **isspace**\ (3)) followed by a single 382080867c5eSQuentin Monnet * optional '**-**' sign. 3821d7a4cb9bSAndrey Ignatov * 3822d7a4cb9bSAndrey Ignatov * Five least significant bits of *flags* encode base, other bits 3823d7a4cb9bSAndrey Ignatov * are currently unused. 3824d7a4cb9bSAndrey Ignatov * 3825d7a4cb9bSAndrey Ignatov * Base must be either 8, 10, 16 or 0 to detect it automatically 382680867c5eSQuentin Monnet * similar to user space **strtol**\ (3). 3827d7a4cb9bSAndrey Ignatov * Return 3828d7a4cb9bSAndrey Ignatov * Number of characters consumed on success. Must be positive but 382980867c5eSQuentin Monnet * no more than *buf_len*. 3830d7a4cb9bSAndrey Ignatov * 3831d7a4cb9bSAndrey Ignatov * **-EINVAL** if no valid digits were found or unsupported base 3832d7a4cb9bSAndrey Ignatov * was provided. 3833d7a4cb9bSAndrey Ignatov * 3834d7a4cb9bSAndrey Ignatov * **-ERANGE** if resulting value was out of range. 3835d7a4cb9bSAndrey Ignatov * 3836bdb7b79bSAndrii Nakryiko * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3837d7a4cb9bSAndrey Ignatov * Description 3838d7a4cb9bSAndrey Ignatov * Convert the initial part of the string from buffer *buf* of 3839d7a4cb9bSAndrey Ignatov * size *buf_len* to an unsigned long integer according to the 3840d7a4cb9bSAndrey Ignatov * given base and save the result in *res*. 3841d7a4cb9bSAndrey Ignatov * 3842d7a4cb9bSAndrey Ignatov * The string may begin with an arbitrary amount of white space 384380867c5eSQuentin Monnet * (as determined by **isspace**\ (3)). 3844d7a4cb9bSAndrey Ignatov * 3845d7a4cb9bSAndrey Ignatov * Five least significant bits of *flags* encode base, other bits 3846d7a4cb9bSAndrey Ignatov * are currently unused. 3847d7a4cb9bSAndrey Ignatov * 3848d7a4cb9bSAndrey Ignatov * Base must be either 8, 10, 16 or 0 to detect it automatically 384980867c5eSQuentin Monnet * similar to user space **strtoul**\ (3). 3850d7a4cb9bSAndrey Ignatov * Return 3851d7a4cb9bSAndrey Ignatov * Number of characters consumed on success. Must be positive but 385280867c5eSQuentin Monnet * no more than *buf_len*. 3853d7a4cb9bSAndrey Ignatov * 3854d7a4cb9bSAndrey Ignatov * **-EINVAL** if no valid digits were found or unsupported base 3855d7a4cb9bSAndrey Ignatov * was provided. 3856d7a4cb9bSAndrey Ignatov * 3857d7a4cb9bSAndrey Ignatov * **-ERANGE** if resulting value was out of range. 38586ac99e8fSMartin KaFai Lau * 385930897832SKP Singh * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 38606ac99e8fSMartin KaFai Lau * Description 386180867c5eSQuentin Monnet * Get a bpf-local-storage from a *sk*. 38626ac99e8fSMartin KaFai Lau * 38636ac99e8fSMartin KaFai Lau * Logically, it could be thought of getting the value from 38646ac99e8fSMartin KaFai Lau * a *map* with *sk* as the **key**. From this 38656ac99e8fSMartin KaFai Lau * perspective, the usage is not much different from 386680867c5eSQuentin Monnet * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 386780867c5eSQuentin Monnet * helper enforces the key must be a full socket and the map must 386880867c5eSQuentin Monnet * be a **BPF_MAP_TYPE_SK_STORAGE** also. 38696ac99e8fSMartin KaFai Lau * 38706ac99e8fSMartin KaFai Lau * Underneath, the value is stored locally at *sk* instead of 387180867c5eSQuentin Monnet * the *map*. The *map* is used as the bpf-local-storage 387280867c5eSQuentin Monnet * "type". The bpf-local-storage "type" (i.e. the *map*) is 387380867c5eSQuentin Monnet * searched against all bpf-local-storages residing at *sk*. 38746ac99e8fSMartin KaFai Lau * 387530897832SKP Singh * *sk* is a kernel **struct sock** pointer for LSM program. 387630897832SKP Singh * *sk* is a **struct bpf_sock** pointer for other program types. 387730897832SKP Singh * 387880867c5eSQuentin Monnet * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 38796ac99e8fSMartin KaFai Lau * used such that a new bpf-local-storage will be 38806ac99e8fSMartin KaFai Lau * created if one does not exist. *value* can be used 388180867c5eSQuentin Monnet * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 38826ac99e8fSMartin KaFai Lau * the initial value of a bpf-local-storage. If *value* is 388380867c5eSQuentin Monnet * **NULL**, the new bpf-local-storage will be zero initialized. 38846ac99e8fSMartin KaFai Lau * Return 38856ac99e8fSMartin KaFai Lau * A bpf-local-storage pointer is returned on success. 38866ac99e8fSMartin KaFai Lau * 38876ac99e8fSMartin KaFai Lau * **NULL** if not found or there was an error in adding 38886ac99e8fSMartin KaFai Lau * a new bpf-local-storage. 38896ac99e8fSMartin KaFai Lau * 389030897832SKP Singh * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 38916ac99e8fSMartin KaFai Lau * Description 389280867c5eSQuentin Monnet * Delete a bpf-local-storage from a *sk*. 38936ac99e8fSMartin KaFai Lau * Return 38946ac99e8fSMartin KaFai Lau * 0 on success. 38956ac99e8fSMartin KaFai Lau * 38966ac99e8fSMartin KaFai Lau * **-ENOENT** if the bpf-local-storage cannot be found. 3897592a3498SMartin KaFai Lau * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 38988b401f9eSYonghong Song * 3899bdb7b79bSAndrii Nakryiko * long bpf_send_signal(u32 sig) 39008b401f9eSYonghong Song * Description 39018482941fSYonghong Song * Send signal *sig* to the process of the current task. 39028482941fSYonghong Song * The signal may be delivered to any of this process's threads. 39038b401f9eSYonghong Song * Return 39048b401f9eSYonghong Song * 0 on success or successfully queued. 39058b401f9eSYonghong Song * 39068b401f9eSYonghong Song * **-EBUSY** if work queue under nmi is full. 39078b401f9eSYonghong Song * 39088b401f9eSYonghong Song * **-EINVAL** if *sig* is invalid. 39098b401f9eSYonghong Song * 39108b401f9eSYonghong Song * **-EPERM** if no permission to send the *sig*. 39118b401f9eSYonghong Song * 39128b401f9eSYonghong Song * **-EAGAIN** if bpf program can try again. 391370d66244SPetar Penkov * 3914c0df236eSMartin KaFai Lau * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 391570d66244SPetar Penkov * Description 391670d66244SPetar Penkov * Try to issue a SYN cookie for the packet with corresponding 391770d66244SPetar Penkov * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 391870d66244SPetar Penkov * 391970d66244SPetar Penkov * *iph* points to the start of the IPv4 or IPv6 header, while 392070d66244SPetar Penkov * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3921ac80287aSMaxim Mikityanskiy * **sizeof**\ (**struct ipv6hdr**). 392270d66244SPetar Penkov * 392370d66244SPetar Penkov * *th* points to the start of the TCP header, while *th_len* 3924ac80287aSMaxim Mikityanskiy * contains the length of the TCP header with options (at least 3925ac80287aSMaxim Mikityanskiy * **sizeof**\ (**struct tcphdr**)). 392670d66244SPetar Penkov * Return 392770d66244SPetar Penkov * On success, lower 32 bits hold the generated SYN cookie in 392870d66244SPetar Penkov * followed by 16 bits which hold the MSS value for that cookie, 392970d66244SPetar Penkov * and the top 16 bits are unused. 393070d66244SPetar Penkov * 393170d66244SPetar Penkov * On failure, the returned value is one of the following: 393270d66244SPetar Penkov * 393370d66244SPetar Penkov * **-EINVAL** SYN cookie cannot be issued due to error 393470d66244SPetar Penkov * 393570d66244SPetar Penkov * **-ENOENT** SYN cookie should not be issued (no SYN flood) 393670d66244SPetar Penkov * 393770d66244SPetar Penkov * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 393870d66244SPetar Penkov * 393970d66244SPetar Penkov * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 3940a7658e1aSAlexei Starovoitov * 3941bdb7b79bSAndrii Nakryiko * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3942a7658e1aSAlexei Starovoitov * Description 3943a7658e1aSAlexei Starovoitov * Write raw *data* blob into a special BPF perf event held by 3944a7658e1aSAlexei Starovoitov * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3945a7658e1aSAlexei Starovoitov * event must have the following attributes: **PERF_SAMPLE_RAW** 3946a7658e1aSAlexei Starovoitov * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3947a7658e1aSAlexei Starovoitov * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3948a7658e1aSAlexei Starovoitov * 3949a7658e1aSAlexei Starovoitov * The *flags* are used to indicate the index in *map* for which 3950a7658e1aSAlexei Starovoitov * the value must be put, masked with **BPF_F_INDEX_MASK**. 3951a7658e1aSAlexei Starovoitov * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3952a7658e1aSAlexei Starovoitov * to indicate that the index of the current CPU core should be 3953a7658e1aSAlexei Starovoitov * used. 3954a7658e1aSAlexei Starovoitov * 3955a7658e1aSAlexei Starovoitov * The value to write, of *size*, is passed through eBPF stack and 3956a7658e1aSAlexei Starovoitov * pointed by *data*. 3957a7658e1aSAlexei Starovoitov * 3958a7658e1aSAlexei Starovoitov * *ctx* is a pointer to in-kernel struct sk_buff. 3959a7658e1aSAlexei Starovoitov * 3960a7658e1aSAlexei Starovoitov * This helper is similar to **bpf_perf_event_output**\ () but 3961a7658e1aSAlexei Starovoitov * restricted to raw_tracepoint bpf programs. 3962a7658e1aSAlexei Starovoitov * Return 3963a7658e1aSAlexei Starovoitov * 0 on success, or a negative error in case of failure. 39646ae08ae3SDaniel Borkmann * 3965bdb7b79bSAndrii Nakryiko * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 39666ae08ae3SDaniel Borkmann * Description 39676ae08ae3SDaniel Borkmann * Safely attempt to read *size* bytes from user space address 39686ae08ae3SDaniel Borkmann * *unsafe_ptr* and store the data in *dst*. 39696ae08ae3SDaniel Borkmann * Return 39706ae08ae3SDaniel Borkmann * 0 on success, or a negative error in case of failure. 39716ae08ae3SDaniel Borkmann * 3972bdb7b79bSAndrii Nakryiko * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 39736ae08ae3SDaniel Borkmann * Description 39746ae08ae3SDaniel Borkmann * Safely attempt to read *size* bytes from kernel space address 39756ae08ae3SDaniel Borkmann * *unsafe_ptr* and store the data in *dst*. 39766ae08ae3SDaniel Borkmann * Return 39776ae08ae3SDaniel Borkmann * 0 on success, or a negative error in case of failure. 39786ae08ae3SDaniel Borkmann * 3979bdb7b79bSAndrii Nakryiko * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 39806ae08ae3SDaniel Borkmann * Description 39816ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe user address 39826ae08ae3SDaniel Borkmann * *unsafe_ptr* to *dst*. The *size* should include the 39836ae08ae3SDaniel Borkmann * terminating NUL byte. In case the string length is smaller than 39846ae08ae3SDaniel Borkmann * *size*, the target is not padded with further NUL bytes. If the 39856ae08ae3SDaniel Borkmann * string length is larger than *size*, just *size*-1 bytes are 39866ae08ae3SDaniel Borkmann * copied and the last byte is set to NUL. 39876ae08ae3SDaniel Borkmann * 3988c6458e72SBrendan Jackman * On success, returns the number of bytes that were written, 3989c6458e72SBrendan Jackman * including the terminal NUL. This makes this helper useful in 3990c6458e72SBrendan Jackman * tracing programs for reading strings, and more importantly to 3991c6458e72SBrendan Jackman * get its length at runtime. See the following snippet: 39926ae08ae3SDaniel Borkmann * 39936ae08ae3SDaniel Borkmann * :: 39946ae08ae3SDaniel Borkmann * 39956ae08ae3SDaniel Borkmann * SEC("kprobe/sys_open") 39966ae08ae3SDaniel Borkmann * void bpf_sys_open(struct pt_regs *ctx) 39976ae08ae3SDaniel Borkmann * { 39986ae08ae3SDaniel Borkmann * char buf[PATHLEN]; // PATHLEN is defined to 256 39996ae08ae3SDaniel Borkmann * int res = bpf_probe_read_user_str(buf, sizeof(buf), 40006ae08ae3SDaniel Borkmann * ctx->di); 40016ae08ae3SDaniel Borkmann * 40026ae08ae3SDaniel Borkmann * // Consume buf, for example push it to 40036ae08ae3SDaniel Borkmann * // userspace via bpf_perf_event_output(); we 40046ae08ae3SDaniel Borkmann * // can use res (the string length) as event 40056ae08ae3SDaniel Borkmann * // size, after checking its boundaries. 40066ae08ae3SDaniel Borkmann * } 40076ae08ae3SDaniel Borkmann * 4008ab8d7809SQuentin Monnet * In comparison, using **bpf_probe_read_user**\ () helper here 40096ae08ae3SDaniel Borkmann * instead to read the string would require to estimate the length 40106ae08ae3SDaniel Borkmann * at compile time, and would often result in copying more memory 40116ae08ae3SDaniel Borkmann * than necessary. 40126ae08ae3SDaniel Borkmann * 40136ae08ae3SDaniel Borkmann * Another useful use case is when parsing individual process 40146ae08ae3SDaniel Borkmann * arguments or individual environment variables navigating 40156ae08ae3SDaniel Borkmann * *current*\ **->mm->arg_start** and *current*\ 40166ae08ae3SDaniel Borkmann * **->mm->env_start**: using this helper and the return value, 40176ae08ae3SDaniel Borkmann * one can quickly iterate at the right offset of the memory area. 40186ae08ae3SDaniel Borkmann * Return 4019c6458e72SBrendan Jackman * On success, the strictly positive length of the output string, 40206ae08ae3SDaniel Borkmann * including the trailing NUL character. On error, a negative 40216ae08ae3SDaniel Borkmann * value. 40226ae08ae3SDaniel Borkmann * 4023bdb7b79bSAndrii Nakryiko * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 40246ae08ae3SDaniel Borkmann * Description 40256ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 4026ab8d7809SQuentin Monnet * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 40276ae08ae3SDaniel Borkmann * Return 40286ae08ae3SDaniel Borkmann * On success, the strictly positive length of the string, including 40296ae08ae3SDaniel Borkmann * the trailing NUL character. On error, a negative value. 4030206057feSMartin KaFai Lau * 4031bdb7b79bSAndrii Nakryiko * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 4032206057feSMartin KaFai Lau * Description 4033ab8d7809SQuentin Monnet * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 4034206057feSMartin KaFai Lau * *rcv_nxt* is the ack_seq to be sent out. 4035206057feSMartin KaFai Lau * Return 4036206057feSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 4037206057feSMartin KaFai Lau * 4038bdb7b79bSAndrii Nakryiko * long bpf_send_signal_thread(u32 sig) 40398482941fSYonghong Song * Description 40408482941fSYonghong Song * Send signal *sig* to the thread corresponding to the current task. 40418482941fSYonghong Song * Return 40428482941fSYonghong Song * 0 on success or successfully queued. 40438482941fSYonghong Song * 40448482941fSYonghong Song * **-EBUSY** if work queue under nmi is full. 40458482941fSYonghong Song * 40468482941fSYonghong Song * **-EINVAL** if *sig* is invalid. 40478482941fSYonghong Song * 40488482941fSYonghong Song * **-EPERM** if no permission to send the *sig*. 40498482941fSYonghong Song * 40508482941fSYonghong Song * **-EAGAIN** if bpf program can try again. 40515576b991SMartin KaFai Lau * 40525576b991SMartin KaFai Lau * u64 bpf_jiffies64(void) 40535576b991SMartin KaFai Lau * Description 40545576b991SMartin KaFai Lau * Obtain the 64bit jiffies 40555576b991SMartin KaFai Lau * Return 40565576b991SMartin KaFai Lau * The 64 bit jiffies 4057fff7b643SDaniel Xu * 4058bdb7b79bSAndrii Nakryiko * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 4059fff7b643SDaniel Xu * Description 4060fff7b643SDaniel Xu * For an eBPF program attached to a perf event, retrieve the 4061ab8d7809SQuentin Monnet * branch records (**struct perf_branch_entry**) associated to *ctx* 4062fff7b643SDaniel Xu * and store it in the buffer pointed by *buf* up to size 4063fff7b643SDaniel Xu * *size* bytes. 4064fff7b643SDaniel Xu * Return 4065fff7b643SDaniel Xu * On success, number of bytes written to *buf*. On error, a 4066fff7b643SDaniel Xu * negative value. 4067fff7b643SDaniel Xu * 4068fff7b643SDaniel Xu * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 4069fff7b643SDaniel Xu * instead return the number of bytes required to store all the 4070fff7b643SDaniel Xu * branch entries. If this flag is set, *buf* may be NULL. 4071fff7b643SDaniel Xu * 4072fff7b643SDaniel Xu * **-EINVAL** if arguments invalid or **size** not a multiple 4073ab8d7809SQuentin Monnet * of **sizeof**\ (**struct perf_branch_entry**\ ). 4074fff7b643SDaniel Xu * 4075fff7b643SDaniel Xu * **-ENOENT** if architecture does not support branch records. 4076b4490c5cSCarlos Neira * 4077bdb7b79bSAndrii Nakryiko * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 4078b4490c5cSCarlos Neira * Description 4079b4490c5cSCarlos Neira * Returns 0 on success, values for *pid* and *tgid* as seen from the current 4080b4490c5cSCarlos Neira * *namespace* will be returned in *nsdata*. 4081ab8d7809SQuentin Monnet * Return 4082ab8d7809SQuentin Monnet * 0 on success, or one of the following in case of failure: 4083b4490c5cSCarlos Neira * 4084b4490c5cSCarlos Neira * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 4085b4490c5cSCarlos Neira * with nsfs of current task, or if dev conversion to dev_t lost high bits. 4086b4490c5cSCarlos Neira * 4087b4490c5cSCarlos Neira * **-ENOENT** if pidns does not exists for the current task. 4088b4490c5cSCarlos Neira * 4089bdb7b79bSAndrii Nakryiko * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4090d831ee84SEelco Chaudron * Description 4091d831ee84SEelco Chaudron * Write raw *data* blob into a special BPF perf event held by 4092d831ee84SEelco Chaudron * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4093d831ee84SEelco Chaudron * event must have the following attributes: **PERF_SAMPLE_RAW** 4094d831ee84SEelco Chaudron * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4095d831ee84SEelco Chaudron * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4096d831ee84SEelco Chaudron * 4097d831ee84SEelco Chaudron * The *flags* are used to indicate the index in *map* for which 4098d831ee84SEelco Chaudron * the value must be put, masked with **BPF_F_INDEX_MASK**. 4099d831ee84SEelco Chaudron * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4100d831ee84SEelco Chaudron * to indicate that the index of the current CPU core should be 4101d831ee84SEelco Chaudron * used. 4102d831ee84SEelco Chaudron * 4103d831ee84SEelco Chaudron * The value to write, of *size*, is passed through eBPF stack and 4104d831ee84SEelco Chaudron * pointed by *data*. 4105d831ee84SEelco Chaudron * 4106d831ee84SEelco Chaudron * *ctx* is a pointer to in-kernel struct xdp_buff. 4107d831ee84SEelco Chaudron * 4108d831ee84SEelco Chaudron * This helper is similar to **bpf_perf_eventoutput**\ () but 4109d831ee84SEelco Chaudron * restricted to raw_tracepoint bpf programs. 4110d831ee84SEelco Chaudron * Return 4111d831ee84SEelco Chaudron * 0 on success, or a negative error in case of failure. 4112f318903cSDaniel Borkmann * 4113f318903cSDaniel Borkmann * u64 bpf_get_netns_cookie(void *ctx) 4114f318903cSDaniel Borkmann * Description 4115f318903cSDaniel Borkmann * Retrieve the cookie (generated by the kernel) of the network 4116f318903cSDaniel Borkmann * namespace the input *ctx* is associated with. The network 4117f318903cSDaniel Borkmann * namespace cookie remains stable for its lifetime and provides 4118f318903cSDaniel Borkmann * a global identifier that can be assumed unique. If *ctx* is 4119f318903cSDaniel Borkmann * NULL, then the helper returns the cookie for the initial 4120f318903cSDaniel Borkmann * network namespace. The cookie itself is very similar to that 4121ab8d7809SQuentin Monnet * of **bpf_get_socket_cookie**\ () helper, but for network 4122ab8d7809SQuentin Monnet * namespaces instead of sockets. 4123f318903cSDaniel Borkmann * Return 4124f318903cSDaniel Borkmann * A 8-byte long opaque number. 41250f09abd1SDaniel Borkmann * 41260f09abd1SDaniel Borkmann * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 41270f09abd1SDaniel Borkmann * Description 41280f09abd1SDaniel Borkmann * Return id of cgroup v2 that is ancestor of the cgroup associated 41290f09abd1SDaniel Borkmann * with the current task at the *ancestor_level*. The root cgroup 41300f09abd1SDaniel Borkmann * is at *ancestor_level* zero and each step down the hierarchy 41310f09abd1SDaniel Borkmann * increments the level. If *ancestor_level* == level of cgroup 41320f09abd1SDaniel Borkmann * associated with the current task, then return value will be the 41330f09abd1SDaniel Borkmann * same as that of **bpf_get_current_cgroup_id**\ (). 41340f09abd1SDaniel Borkmann * 41350f09abd1SDaniel Borkmann * The helper is useful to implement policies based on cgroups 41360f09abd1SDaniel Borkmann * that are upper in hierarchy than immediate cgroup associated 41370f09abd1SDaniel Borkmann * with the current task. 41380f09abd1SDaniel Borkmann * 41390f09abd1SDaniel Borkmann * The format of returned id and helper limitations are same as in 41400f09abd1SDaniel Borkmann * **bpf_get_current_cgroup_id**\ (). 41410f09abd1SDaniel Borkmann * Return 41420f09abd1SDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 4143cf7fbe66SJoe Stringer * 414427e5203bSMartin KaFai Lau * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 4145cf7fbe66SJoe Stringer * Description 4146e9ddbb77SJakub Sitnicki * Helper is overloaded depending on BPF program type. This 4147e9ddbb77SJakub Sitnicki * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 4148e9ddbb77SJakub Sitnicki * **BPF_PROG_TYPE_SCHED_ACT** programs. 4149e9ddbb77SJakub Sitnicki * 4150cf7fbe66SJoe Stringer * Assign the *sk* to the *skb*. When combined with appropriate 4151cf7fbe66SJoe Stringer * routing configuration to receive the packet towards the socket, 4152cf7fbe66SJoe Stringer * will cause *skb* to be delivered to the specified socket. 4153cf7fbe66SJoe Stringer * Subsequent redirection of *skb* via **bpf_redirect**\ (), 4154cf7fbe66SJoe Stringer * **bpf_clone_redirect**\ () or other methods outside of BPF may 4155cf7fbe66SJoe Stringer * interfere with successful delivery to the socket. 4156cf7fbe66SJoe Stringer * 4157cf7fbe66SJoe Stringer * This operation is only valid from TC ingress path. 4158cf7fbe66SJoe Stringer * 4159cf7fbe66SJoe Stringer * The *flags* argument must be zero. 4160cf7fbe66SJoe Stringer * Return 4161ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 4162cf7fbe66SJoe Stringer * 4163ab8d7809SQuentin Monnet * **-EINVAL** if specified *flags* are not supported. 4164ab8d7809SQuentin Monnet * 4165ab8d7809SQuentin Monnet * **-ENOENT** if the socket is unavailable for assignment. 4166ab8d7809SQuentin Monnet * 4167ab8d7809SQuentin Monnet * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4168ab8d7809SQuentin Monnet * 4169ab8d7809SQuentin Monnet * **-EOPNOTSUPP** if the operation is not supported, for example 4170ab8d7809SQuentin Monnet * a call from outside of TC ingress. 4171ab8d7809SQuentin Monnet * 4172ab8d7809SQuentin Monnet * **-ESOCKTNOSUPPORT** if the socket type is not supported 4173ab8d7809SQuentin Monnet * (reuseport). 417471d19214SMaciej Żenczykowski * 4175e9ddbb77SJakub Sitnicki * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4176e9ddbb77SJakub Sitnicki * Description 4177e9ddbb77SJakub Sitnicki * Helper is overloaded depending on BPF program type. This 4178e9ddbb77SJakub Sitnicki * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4179e9ddbb77SJakub Sitnicki * 4180e9ddbb77SJakub Sitnicki * Select the *sk* as a result of a socket lookup. 4181e9ddbb77SJakub Sitnicki * 4182e9ddbb77SJakub Sitnicki * For the operation to succeed passed socket must be compatible 4183e9ddbb77SJakub Sitnicki * with the packet description provided by the *ctx* object. 4184e9ddbb77SJakub Sitnicki * 4185e9ddbb77SJakub Sitnicki * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4186e9ddbb77SJakub Sitnicki * be an exact match. While IP family (**AF_INET** or 4187e9ddbb77SJakub Sitnicki * **AF_INET6**) must be compatible, that is IPv6 sockets 4188e9ddbb77SJakub Sitnicki * that are not v6-only can be selected for IPv4 packets. 4189e9ddbb77SJakub Sitnicki * 4190e9ddbb77SJakub Sitnicki * Only TCP listeners and UDP unconnected sockets can be 4191e9ddbb77SJakub Sitnicki * selected. *sk* can also be NULL to reset any previous 4192e9ddbb77SJakub Sitnicki * selection. 4193e9ddbb77SJakub Sitnicki * 4194e9ddbb77SJakub Sitnicki * *flags* argument can combination of following values: 4195e9ddbb77SJakub Sitnicki * 4196e9ddbb77SJakub Sitnicki * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4197e9ddbb77SJakub Sitnicki * socket selection, potentially done by a BPF program 4198e9ddbb77SJakub Sitnicki * that ran before us. 4199e9ddbb77SJakub Sitnicki * 4200e9ddbb77SJakub Sitnicki * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4201e9ddbb77SJakub Sitnicki * load-balancing within reuseport group for the socket 4202e9ddbb77SJakub Sitnicki * being selected. 4203e9ddbb77SJakub Sitnicki * 4204e9ddbb77SJakub Sitnicki * On success *ctx->sk* will point to the selected socket. 4205e9ddbb77SJakub Sitnicki * 4206e9ddbb77SJakub Sitnicki * Return 4207e9ddbb77SJakub Sitnicki * 0 on success, or a negative errno in case of failure. 4208e9ddbb77SJakub Sitnicki * 4209e9ddbb77SJakub Sitnicki * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4210e9ddbb77SJakub Sitnicki * not compatible with packet family (*ctx->family*). 4211e9ddbb77SJakub Sitnicki * 4212e9ddbb77SJakub Sitnicki * * **-EEXIST** if socket has been already selected, 4213e9ddbb77SJakub Sitnicki * potentially by another program, and 4214e9ddbb77SJakub Sitnicki * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4215e9ddbb77SJakub Sitnicki * 4216e9ddbb77SJakub Sitnicki * * **-EINVAL** if unsupported flags were specified. 4217e9ddbb77SJakub Sitnicki * 4218e9ddbb77SJakub Sitnicki * * **-EPROTOTYPE** if socket L4 protocol 4219e9ddbb77SJakub Sitnicki * (*sk->protocol*) doesn't match packet protocol 4220e9ddbb77SJakub Sitnicki * (*ctx->protocol*). 4221e9ddbb77SJakub Sitnicki * 4222e9ddbb77SJakub Sitnicki * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4223e9ddbb77SJakub Sitnicki * state (TCP listening or UDP unconnected). 4224e9ddbb77SJakub Sitnicki * 422571d19214SMaciej Żenczykowski * u64 bpf_ktime_get_boot_ns(void) 422671d19214SMaciej Żenczykowski * Description 422771d19214SMaciej Żenczykowski * Return the time elapsed since system boot, in nanoseconds. 422871d19214SMaciej Żenczykowski * Does include the time the system was suspended. 4229ab8d7809SQuentin Monnet * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 423071d19214SMaciej Żenczykowski * Return 423171d19214SMaciej Żenczykowski * Current *ktime*. 4232492e639fSYonghong Song * 4233bdb7b79bSAndrii Nakryiko * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4234492e639fSYonghong Song * Description 4235ab8d7809SQuentin Monnet * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4236ab8d7809SQuentin Monnet * out the format string. 4237492e639fSYonghong Song * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4238492e639fSYonghong Song * the format string itself. The *data* and *data_len* are format string 4239ab8d7809SQuentin Monnet * arguments. The *data* are a **u64** array and corresponding format string 4240492e639fSYonghong Song * values are stored in the array. For strings and pointers where pointees 4241492e639fSYonghong Song * are accessed, only the pointer values are stored in the *data* array. 4242a42effb0SDave Marchevsky * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4243492e639fSYonghong Song * 4244492e639fSYonghong Song * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4245492e639fSYonghong Song * Reading kernel memory may fail due to either invalid address or 4246492e639fSYonghong Song * valid address but requiring a major memory fault. If reading kernel memory 4247492e639fSYonghong Song * fails, the string for **%s** will be an empty string, and the ip 4248492e639fSYonghong Song * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4249ab8d7809SQuentin Monnet * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4250492e639fSYonghong Song * Return 4251ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 4252492e639fSYonghong Song * 4253ab8d7809SQuentin Monnet * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4254492e639fSYonghong Song * by returning 1 from bpf program. 4255ab8d7809SQuentin Monnet * 4256ab8d7809SQuentin Monnet * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4257ab8d7809SQuentin Monnet * 4258ab8d7809SQuentin Monnet * **-E2BIG** if *fmt* contains too many format specifiers. 4259ab8d7809SQuentin Monnet * 4260ab8d7809SQuentin Monnet * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4261492e639fSYonghong Song * 4262bdb7b79bSAndrii Nakryiko * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4263492e639fSYonghong Song * Description 4264ab8d7809SQuentin Monnet * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4265492e639fSYonghong Song * The *m* represents the seq_file. The *data* and *len* represent the 4266492e639fSYonghong Song * data to write in bytes. 4267492e639fSYonghong Song * Return 4268ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 4269492e639fSYonghong Song * 4270ab8d7809SQuentin Monnet * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4271f307fa2cSAndrey Ignatov * 4272a5fa25adSMartin KaFai Lau * u64 bpf_sk_cgroup_id(void *sk) 4273f307fa2cSAndrey Ignatov * Description 4274f307fa2cSAndrey Ignatov * Return the cgroup v2 id of the socket *sk*. 4275f307fa2cSAndrey Ignatov * 4276a5fa25adSMartin KaFai Lau * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4277f307fa2cSAndrey Ignatov * returned from **bpf_sk_lookup_xxx**\ (), 4278f307fa2cSAndrey Ignatov * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4279f307fa2cSAndrey Ignatov * same as in **bpf_skb_cgroup_id**\ (). 4280f307fa2cSAndrey Ignatov * 4281f307fa2cSAndrey Ignatov * This helper is available only if the kernel was compiled with 4282f307fa2cSAndrey Ignatov * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4283f307fa2cSAndrey Ignatov * Return 4284f307fa2cSAndrey Ignatov * The id is returned or 0 in case the id could not be retrieved. 4285f307fa2cSAndrey Ignatov * 4286a5fa25adSMartin KaFai Lau * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4287f307fa2cSAndrey Ignatov * Description 4288f307fa2cSAndrey Ignatov * Return id of cgroup v2 that is ancestor of cgroup associated 4289f307fa2cSAndrey Ignatov * with the *sk* at the *ancestor_level*. The root cgroup is at 4290f307fa2cSAndrey Ignatov * *ancestor_level* zero and each step down the hierarchy 4291f307fa2cSAndrey Ignatov * increments the level. If *ancestor_level* == level of cgroup 4292f307fa2cSAndrey Ignatov * associated with *sk*, then return value will be same as that 4293f307fa2cSAndrey Ignatov * of **bpf_sk_cgroup_id**\ (). 4294f307fa2cSAndrey Ignatov * 4295f307fa2cSAndrey Ignatov * The helper is useful to implement policies based on cgroups 4296f307fa2cSAndrey Ignatov * that are upper in hierarchy than immediate cgroup associated 4297f307fa2cSAndrey Ignatov * with *sk*. 4298f307fa2cSAndrey Ignatov * 4299f307fa2cSAndrey Ignatov * The format of returned id and helper limitations are same as in 4300f307fa2cSAndrey Ignatov * **bpf_sk_cgroup_id**\ (). 4301f307fa2cSAndrey Ignatov * Return 4302f307fa2cSAndrey Ignatov * The id is returned or 0 in case the id could not be retrieved. 4303457f4436SAndrii Nakryiko * 4304e1613b57SAndrii Nakryiko * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4305457f4436SAndrii Nakryiko * Description 4306457f4436SAndrii Nakryiko * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4307bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4308bcc7f554SQuentin Monnet * of new data availability is sent. 4309bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4310bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 43115c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 43125c507329SPedro Tammela * of new data availability is sent. 43135c507329SPedro Tammela * 43145c507329SPedro Tammela * An adaptive notification is a notification sent whenever the user-space 43155c507329SPedro Tammela * process has caught up and consumed all available payloads. In case the user-space 43165c507329SPedro Tammela * process is still processing a previous payload, then no notification is needed 43175c507329SPedro Tammela * as it will process the newly added payload automatically. 4318457f4436SAndrii Nakryiko * Return 4319bcc7f554SQuentin Monnet * 0 on success, or a negative error in case of failure. 4320457f4436SAndrii Nakryiko * 4321457f4436SAndrii Nakryiko * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4322457f4436SAndrii Nakryiko * Description 4323457f4436SAndrii Nakryiko * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 43245c507329SPedro Tammela * *flags* must be 0. 4325457f4436SAndrii Nakryiko * Return 4326457f4436SAndrii Nakryiko * Valid pointer with *size* bytes of memory available; NULL, 4327457f4436SAndrii Nakryiko * otherwise. 4328457f4436SAndrii Nakryiko * 4329457f4436SAndrii Nakryiko * void bpf_ringbuf_submit(void *data, u64 flags) 4330457f4436SAndrii Nakryiko * Description 4331457f4436SAndrii Nakryiko * Submit reserved ring buffer sample, pointed to by *data*. 4332bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4333bcc7f554SQuentin Monnet * of new data availability is sent. 4334bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4335bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 43365c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 43375c507329SPedro Tammela * of new data availability is sent. 43385c507329SPedro Tammela * 43395c507329SPedro Tammela * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4340457f4436SAndrii Nakryiko * Return 4341457f4436SAndrii Nakryiko * Nothing. Always succeeds. 4342457f4436SAndrii Nakryiko * 4343457f4436SAndrii Nakryiko * void bpf_ringbuf_discard(void *data, u64 flags) 4344457f4436SAndrii Nakryiko * Description 4345457f4436SAndrii Nakryiko * Discard reserved ring buffer sample, pointed to by *data*. 4346bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4347bcc7f554SQuentin Monnet * of new data availability is sent. 4348bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4349bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 43505c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 43515c507329SPedro Tammela * of new data availability is sent. 43525c507329SPedro Tammela * 43535c507329SPedro Tammela * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4354457f4436SAndrii Nakryiko * Return 4355457f4436SAndrii Nakryiko * Nothing. Always succeeds. 4356457f4436SAndrii Nakryiko * 4357457f4436SAndrii Nakryiko * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4358457f4436SAndrii Nakryiko * Description 4359457f4436SAndrii Nakryiko * Query various characteristics of provided ring buffer. What 4360457f4436SAndrii Nakryiko * exactly is queries is determined by *flags*: 4361bcc7f554SQuentin Monnet * 4362bcc7f554SQuentin Monnet * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4363bcc7f554SQuentin Monnet * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4364bcc7f554SQuentin Monnet * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4365bcc7f554SQuentin Monnet * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4366bcc7f554SQuentin Monnet * 4367bcc7f554SQuentin Monnet * Data returned is just a momentary snapshot of actual values 4368457f4436SAndrii Nakryiko * and could be inaccurate, so this facility should be used to 4369457f4436SAndrii Nakryiko * power heuristics and for reporting, not to make 100% correct 4370457f4436SAndrii Nakryiko * calculation. 4371457f4436SAndrii Nakryiko * Return 4372bcc7f554SQuentin Monnet * Requested value, or 0, if *flags* are not recognized. 43737cdec54fSDaniel Borkmann * 4374bdb7b79bSAndrii Nakryiko * long bpf_csum_level(struct sk_buff *skb, u64 level) 43757cdec54fSDaniel Borkmann * Description 43767cdec54fSDaniel Borkmann * Change the skbs checksum level by one layer up or down, or 43777cdec54fSDaniel Borkmann * reset it entirely to none in order to have the stack perform 43787cdec54fSDaniel Borkmann * checksum validation. The level is applicable to the following 43797cdec54fSDaniel Borkmann * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 43807cdec54fSDaniel Borkmann * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 43817cdec54fSDaniel Borkmann * through **bpf_skb_adjust_room**\ () helper with passing in 43827cdec54fSDaniel Borkmann * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 43837cdec54fSDaniel Borkmann * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 43847cdec54fSDaniel Borkmann * the UDP header is removed. Similarly, an encap of the latter 43857cdec54fSDaniel Borkmann * into the former could be accompanied by a helper call to 43867cdec54fSDaniel Borkmann * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 43877cdec54fSDaniel Borkmann * skb is still intended to be processed in higher layers of the 43887cdec54fSDaniel Borkmann * stack instead of just egressing at tc. 43897cdec54fSDaniel Borkmann * 43907cdec54fSDaniel Borkmann * There are three supported level settings at this time: 43917cdec54fSDaniel Borkmann * 43927cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 43937cdec54fSDaniel Borkmann * with CHECKSUM_UNNECESSARY. 43947cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 43957cdec54fSDaniel Borkmann * with CHECKSUM_UNNECESSARY. 43967cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 43977cdec54fSDaniel Borkmann * sets CHECKSUM_NONE to force checksum validation by the stack. 43987cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 43997cdec54fSDaniel Borkmann * skb->csum_level. 44007cdec54fSDaniel Borkmann * Return 44017cdec54fSDaniel Borkmann * 0 on success, or a negative error in case of failure. In the 44027cdec54fSDaniel Borkmann * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 44037cdec54fSDaniel Borkmann * is returned or the error code -EACCES in case the skb is not 44047cdec54fSDaniel Borkmann * subject to CHECKSUM_UNNECESSARY. 4405af7ec138SYonghong Song * 4406af7ec138SYonghong Song * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4407af7ec138SYonghong Song * Description 4408af7ec138SYonghong Song * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4409af7ec138SYonghong Song * Return 4410938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4411478cfbdfSYonghong Song * 4412478cfbdfSYonghong Song * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4413478cfbdfSYonghong Song * Description 4414478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4415478cfbdfSYonghong Song * Return 4416938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4417478cfbdfSYonghong Song * 4418478cfbdfSYonghong Song * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4419478cfbdfSYonghong Song * Description 4420478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4421478cfbdfSYonghong Song * Return 4422938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4423478cfbdfSYonghong Song * 4424478cfbdfSYonghong Song * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4425478cfbdfSYonghong Song * Description 4426478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4427478cfbdfSYonghong Song * Return 4428938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 44290d4fad3eSYonghong Song * 44300d4fad3eSYonghong Song * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 44310d4fad3eSYonghong Song * Description 44320d4fad3eSYonghong Song * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 44330d4fad3eSYonghong Song * Return 4434938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4435fa28dcb8SSong Liu * 4436fa28dcb8SSong Liu * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4437fa28dcb8SSong Liu * Description 4438fa28dcb8SSong Liu * Return a user or a kernel stack in bpf program provided buffer. 4439fa28dcb8SSong Liu * To achieve this, the helper needs *task*, which is a valid 4440938c3efdSQuentin Monnet * pointer to **struct task_struct**. To store the stacktrace, the 4441fa28dcb8SSong Liu * bpf program provides *buf* with a nonnegative *size*. 4442fa28dcb8SSong Liu * 4443fa28dcb8SSong Liu * The last argument, *flags*, holds the number of stack frames to 4444fa28dcb8SSong Liu * skip (from 0 to 255), masked with 4445fa28dcb8SSong Liu * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4446fa28dcb8SSong Liu * the following flags: 4447fa28dcb8SSong Liu * 4448fa28dcb8SSong Liu * **BPF_F_USER_STACK** 4449fa28dcb8SSong Liu * Collect a user space stack instead of a kernel stack. 4450fa28dcb8SSong Liu * **BPF_F_USER_BUILD_ID** 4451fa28dcb8SSong Liu * Collect buildid+offset instead of ips for user stack, 4452fa28dcb8SSong Liu * only valid if **BPF_F_USER_STACK** is also specified. 4453fa28dcb8SSong Liu * 4454fa28dcb8SSong Liu * **bpf_get_task_stack**\ () can collect up to 4455fa28dcb8SSong Liu * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4456fa28dcb8SSong Liu * to sufficient large buffer size. Note that 4457fa28dcb8SSong Liu * this limit can be controlled with the **sysctl** program, and 4458fa28dcb8SSong Liu * that it should be manually increased in order to profile long 4459fa28dcb8SSong Liu * user stacks (such as stacks for Java programs). To do so, use: 4460fa28dcb8SSong Liu * 4461fa28dcb8SSong Liu * :: 4462fa28dcb8SSong Liu * 4463fa28dcb8SSong Liu * # sysctl kernel.perf_event_max_stack=<new value> 4464fa28dcb8SSong Liu * Return 4465ee2a0988SNamhyung Kim * The non-negative copied *buf* length equal to or less than 4466ee2a0988SNamhyung Kim * *size* on success, or a negative error in case of failure. 4467fa28dcb8SSong Liu * 44680813a841SMartin KaFai Lau * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 44690813a841SMartin KaFai Lau * Description 44700813a841SMartin KaFai Lau * Load header option. Support reading a particular TCP header 4471938c3efdSQuentin Monnet * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 44720813a841SMartin KaFai Lau * 44730813a841SMartin KaFai Lau * If *flags* is 0, it will search the option from the 4474938c3efdSQuentin Monnet * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 44750813a841SMartin KaFai Lau * has details on what skb_data contains under different 4476938c3efdSQuentin Monnet * *skops*\ **->op**. 44770813a841SMartin KaFai Lau * 44780813a841SMartin KaFai Lau * The first byte of the *searchby_res* specifies the 44790813a841SMartin KaFai Lau * kind that it wants to search. 44800813a841SMartin KaFai Lau * 44810813a841SMartin KaFai Lau * If the searching kind is an experimental kind 44820813a841SMartin KaFai Lau * (i.e. 253 or 254 according to RFC6994). It also 44830813a841SMartin KaFai Lau * needs to specify the "magic" which is either 44840813a841SMartin KaFai Lau * 2 bytes or 4 bytes. It then also needs to 44850813a841SMartin KaFai Lau * specify the size of the magic by using 44860813a841SMartin KaFai Lau * the 2nd byte which is "kind-length" of a TCP 44870813a841SMartin KaFai Lau * header option and the "kind-length" also 44880813a841SMartin KaFai Lau * includes the first 2 bytes "kind" and "kind-length" 44890813a841SMartin KaFai Lau * itself as a normal TCP header option also does. 44900813a841SMartin KaFai Lau * 44910813a841SMartin KaFai Lau * For example, to search experimental kind 254 with 44920813a841SMartin KaFai Lau * 2 byte magic 0xeB9F, the searchby_res should be 44930813a841SMartin KaFai Lau * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 44940813a841SMartin KaFai Lau * 44950813a841SMartin KaFai Lau * To search for the standard window scale option (3), 4496938c3efdSQuentin Monnet * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 44970813a841SMartin KaFai Lau * Note, kind-length must be 0 for regular option. 44980813a841SMartin KaFai Lau * 44990813a841SMartin KaFai Lau * Searching for No-Op (0) and End-of-Option-List (1) are 45000813a841SMartin KaFai Lau * not supported. 45010813a841SMartin KaFai Lau * 45020813a841SMartin KaFai Lau * *len* must be at least 2 bytes which is the minimal size 45030813a841SMartin KaFai Lau * of a header option. 45040813a841SMartin KaFai Lau * 45050813a841SMartin KaFai Lau * Supported flags: 4506938c3efdSQuentin Monnet * 45070813a841SMartin KaFai Lau * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 45080813a841SMartin KaFai Lau * saved_syn packet or the just-received syn packet. 45090813a841SMartin KaFai Lau * 45100813a841SMartin KaFai Lau * Return 45110813a841SMartin KaFai Lau * > 0 when found, the header option is copied to *searchby_res*. 4512938c3efdSQuentin Monnet * The return value is the total length copied. On failure, a 4513938c3efdSQuentin Monnet * negative error code is returned: 45140813a841SMartin KaFai Lau * 4515938c3efdSQuentin Monnet * **-EINVAL** if a parameter is invalid. 45160813a841SMartin KaFai Lau * 4517938c3efdSQuentin Monnet * **-ENOMSG** if the option is not found. 45180813a841SMartin KaFai Lau * 4519938c3efdSQuentin Monnet * **-ENOENT** if no syn packet is available when 4520938c3efdSQuentin Monnet * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 45210813a841SMartin KaFai Lau * 4522938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space. Only *len* number of 45230813a841SMartin KaFai Lau * bytes are copied. 45240813a841SMartin KaFai Lau * 4525938c3efdSQuentin Monnet * **-EFAULT** on failure to parse the header options in the 4526938c3efdSQuentin Monnet * packet. 45270813a841SMartin KaFai Lau * 4528938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4529938c3efdSQuentin Monnet * *skops*\ **->op**. 45300813a841SMartin KaFai Lau * 45310813a841SMartin KaFai Lau * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 45320813a841SMartin KaFai Lau * Description 45330813a841SMartin KaFai Lau * Store header option. The data will be copied 45340813a841SMartin KaFai Lau * from buffer *from* with length *len* to the TCP header. 45350813a841SMartin KaFai Lau * 45360813a841SMartin KaFai Lau * The buffer *from* should have the whole option that 45370813a841SMartin KaFai Lau * includes the kind, kind-length, and the actual 45380813a841SMartin KaFai Lau * option data. The *len* must be at least kind-length 45390813a841SMartin KaFai Lau * long. The kind-length does not have to be 4 byte 45400813a841SMartin KaFai Lau * aligned. The kernel will take care of the padding 45410813a841SMartin KaFai Lau * and setting the 4 bytes aligned value to th->doff. 45420813a841SMartin KaFai Lau * 45430813a841SMartin KaFai Lau * This helper will check for duplicated option 45440813a841SMartin KaFai Lau * by searching the same option in the outgoing skb. 45450813a841SMartin KaFai Lau * 45460813a841SMartin KaFai Lau * This helper can only be called during 4547938c3efdSQuentin Monnet * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 45480813a841SMartin KaFai Lau * 45490813a841SMartin KaFai Lau * Return 45500813a841SMartin KaFai Lau * 0 on success, or negative error in case of failure: 45510813a841SMartin KaFai Lau * 4552938c3efdSQuentin Monnet * **-EINVAL** If param is invalid. 45530813a841SMartin KaFai Lau * 4554938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space in the header. 45550813a841SMartin KaFai Lau * Nothing has been written 45560813a841SMartin KaFai Lau * 4557938c3efdSQuentin Monnet * **-EEXIST** if the option already exists. 45580813a841SMartin KaFai Lau * 4559aa75622cSQuentin Monnet * **-EFAULT** on failure to parse the existing header options. 45600813a841SMartin KaFai Lau * 4561938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4562938c3efdSQuentin Monnet * *skops*\ **->op**. 45630813a841SMartin KaFai Lau * 45640813a841SMartin KaFai Lau * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 45650813a841SMartin KaFai Lau * Description 45660813a841SMartin KaFai Lau * Reserve *len* bytes for the bpf header option. The 4567938c3efdSQuentin Monnet * space will be used by **bpf_store_hdr_opt**\ () later in 4568938c3efdSQuentin Monnet * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 45690813a841SMartin KaFai Lau * 4570938c3efdSQuentin Monnet * If **bpf_reserve_hdr_opt**\ () is called multiple times, 45710813a841SMartin KaFai Lau * the total number of bytes will be reserved. 45720813a841SMartin KaFai Lau * 45730813a841SMartin KaFai Lau * This helper can only be called during 4574938c3efdSQuentin Monnet * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 45750813a841SMartin KaFai Lau * 45760813a841SMartin KaFai Lau * Return 45770813a841SMartin KaFai Lau * 0 on success, or negative error in case of failure: 45780813a841SMartin KaFai Lau * 4579938c3efdSQuentin Monnet * **-EINVAL** if a parameter is invalid. 45800813a841SMartin KaFai Lau * 4581938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space in the header. 45820813a841SMartin KaFai Lau * 4583938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4584938c3efdSQuentin Monnet * *skops*\ **->op**. 45856e22ab9dSJiri Olsa * 45868ea63684SKP Singh * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 45878ea63684SKP Singh * Description 45888ea63684SKP Singh * Get a bpf_local_storage from an *inode*. 45898ea63684SKP Singh * 45908ea63684SKP Singh * Logically, it could be thought of as getting the value from 45918ea63684SKP Singh * a *map* with *inode* as the **key**. From this 45928ea63684SKP Singh * perspective, the usage is not much different from 45938ea63684SKP Singh * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 45948ea63684SKP Singh * helper enforces the key must be an inode and the map must also 45958ea63684SKP Singh * be a **BPF_MAP_TYPE_INODE_STORAGE**. 45968ea63684SKP Singh * 45978ea63684SKP Singh * Underneath, the value is stored locally at *inode* instead of 45988ea63684SKP Singh * the *map*. The *map* is used as the bpf-local-storage 45998ea63684SKP Singh * "type". The bpf-local-storage "type" (i.e. the *map*) is 46008ea63684SKP Singh * searched against all bpf_local_storage residing at *inode*. 46018ea63684SKP Singh * 46028ea63684SKP Singh * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 46038ea63684SKP Singh * used such that a new bpf_local_storage will be 46048ea63684SKP Singh * created if one does not exist. *value* can be used 46058ea63684SKP Singh * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 46068ea63684SKP Singh * the initial value of a bpf_local_storage. If *value* is 46078ea63684SKP Singh * **NULL**, the new bpf_local_storage will be zero initialized. 46088ea63684SKP Singh * Return 46098ea63684SKP Singh * A bpf_local_storage pointer is returned on success. 46108ea63684SKP Singh * 46118ea63684SKP Singh * **NULL** if not found or there was an error in adding 46128ea63684SKP Singh * a new bpf_local_storage. 46138ea63684SKP Singh * 46148ea63684SKP Singh * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 46158ea63684SKP Singh * Description 46168ea63684SKP Singh * Delete a bpf_local_storage from an *inode*. 46178ea63684SKP Singh * Return 46188ea63684SKP Singh * 0 on success. 46198ea63684SKP Singh * 46208ea63684SKP Singh * **-ENOENT** if the bpf_local_storage cannot be found. 46216e22ab9dSJiri Olsa * 46226e22ab9dSJiri Olsa * long bpf_d_path(struct path *path, char *buf, u32 sz) 46236e22ab9dSJiri Olsa * Description 4624938c3efdSQuentin Monnet * Return full path for given **struct path** object, which 4625938c3efdSQuentin Monnet * needs to be the kernel BTF *path* object. The path is 4626938c3efdSQuentin Monnet * returned in the provided buffer *buf* of size *sz* and 46276e22ab9dSJiri Olsa * is zero terminated. 46286e22ab9dSJiri Olsa * 46296e22ab9dSJiri Olsa * Return 46306e22ab9dSJiri Olsa * On success, the strictly positive length of the string, 46316e22ab9dSJiri Olsa * including the trailing NUL character. On error, a negative 46326e22ab9dSJiri Olsa * value. 463307be4c4aSAlexei Starovoitov * 463407be4c4aSAlexei Starovoitov * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 463507be4c4aSAlexei Starovoitov * Description 463607be4c4aSAlexei Starovoitov * Read *size* bytes from user space address *user_ptr* and store 4637938c3efdSQuentin Monnet * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 463807be4c4aSAlexei Starovoitov * Return 463907be4c4aSAlexei Starovoitov * 0 on success, or a negative error in case of failure. 4640c4d0bfb4SAlan Maguire * 4641c4d0bfb4SAlan Maguire * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4642c4d0bfb4SAlan Maguire * Description 4643c4d0bfb4SAlan Maguire * Use BTF to store a string representation of *ptr*->ptr in *str*, 4644c4d0bfb4SAlan Maguire * using *ptr*->type_id. This value should specify the type 4645c4d0bfb4SAlan Maguire * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4646c4d0bfb4SAlan Maguire * can be used to look up vmlinux BTF type ids. Traversing the 4647c4d0bfb4SAlan Maguire * data structure using BTF, the type information and values are 4648c4d0bfb4SAlan Maguire * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4649c4d0bfb4SAlan Maguire * the pointer data is carried out to avoid kernel crashes during 4650c4d0bfb4SAlan Maguire * operation. Smaller types can use string space on the stack; 4651c4d0bfb4SAlan Maguire * larger programs can use map data to store the string 4652c4d0bfb4SAlan Maguire * representation. 4653c4d0bfb4SAlan Maguire * 4654c4d0bfb4SAlan Maguire * The string can be subsequently shared with userspace via 4655c4d0bfb4SAlan Maguire * bpf_perf_event_output() or ring buffer interfaces. 4656c4d0bfb4SAlan Maguire * bpf_trace_printk() is to be avoided as it places too small 4657c4d0bfb4SAlan Maguire * a limit on string size to be useful. 4658c4d0bfb4SAlan Maguire * 4659c4d0bfb4SAlan Maguire * *flags* is a combination of 4660c4d0bfb4SAlan Maguire * 4661c4d0bfb4SAlan Maguire * **BTF_F_COMPACT** 4662c4d0bfb4SAlan Maguire * no formatting around type information 4663c4d0bfb4SAlan Maguire * **BTF_F_NONAME** 4664c4d0bfb4SAlan Maguire * no struct/union member names/types 4665c4d0bfb4SAlan Maguire * **BTF_F_PTR_RAW** 4666c4d0bfb4SAlan Maguire * show raw (unobfuscated) pointer values; 4667c4d0bfb4SAlan Maguire * equivalent to printk specifier %px. 4668c4d0bfb4SAlan Maguire * **BTF_F_ZERO** 4669c4d0bfb4SAlan Maguire * show zero-valued struct/union members; they 4670c4d0bfb4SAlan Maguire * are not displayed by default 4671c4d0bfb4SAlan Maguire * 4672c4d0bfb4SAlan Maguire * Return 4673c4d0bfb4SAlan Maguire * The number of bytes that were written (or would have been 4674c4d0bfb4SAlan Maguire * written if output had to be truncated due to string size), 4675c4d0bfb4SAlan Maguire * or a negative error in cases of failure. 4676eb411377SAlan Maguire * 4677eb411377SAlan Maguire * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4678eb411377SAlan Maguire * Description 4679eb411377SAlan Maguire * Use BTF to write to seq_write a string representation of 4680eb411377SAlan Maguire * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4681eb411377SAlan Maguire * *flags* are identical to those used for bpf_snprintf_btf. 4682eb411377SAlan Maguire * Return 4683eb411377SAlan Maguire * 0 on success or a negative error in case of failure. 4684b426ce83SDaniel Borkmann * 4685b426ce83SDaniel Borkmann * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4686b426ce83SDaniel Borkmann * Description 4687b426ce83SDaniel Borkmann * See **bpf_get_cgroup_classid**\ () for the main description. 4688b426ce83SDaniel Borkmann * This helper differs from **bpf_get_cgroup_classid**\ () in that 4689b426ce83SDaniel Borkmann * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4690b426ce83SDaniel Borkmann * associated socket instead of the current process. 4691b426ce83SDaniel Borkmann * Return 4692b426ce83SDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 4693b4ab3141SDaniel Borkmann * 4694ba452c9eSToke Høiland-Jørgensen * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4695b4ab3141SDaniel Borkmann * Description 4696b4ab3141SDaniel Borkmann * Redirect the packet to another net device of index *ifindex* 4697b4ab3141SDaniel Borkmann * and fill in L2 addresses from neighboring subsystem. This helper 4698b4ab3141SDaniel Borkmann * is somewhat similar to **bpf_redirect**\ (), except that it 4699dd2ce6a5SDaniel Borkmann * populates L2 addresses as well, meaning, internally, the helper 4700ba452c9eSToke Høiland-Jørgensen * relies on the neighbor lookup for the L2 address of the nexthop. 4701ba452c9eSToke Høiland-Jørgensen * 4702ba452c9eSToke Høiland-Jørgensen * The helper will perform a FIB lookup based on the skb's 4703ba452c9eSToke Høiland-Jørgensen * networking header to get the address of the next hop, unless 4704ba452c9eSToke Høiland-Jørgensen * this is supplied by the caller in the *params* argument. The 4705ba452c9eSToke Høiland-Jørgensen * *plen* argument indicates the len of *params* and should be set 4706ba452c9eSToke Høiland-Jørgensen * to 0 if *params* is NULL. 4707dd2ce6a5SDaniel Borkmann * 4708b4ab3141SDaniel Borkmann * The *flags* argument is reserved and must be 0. The helper is 4709dd2ce6a5SDaniel Borkmann * currently only supported for tc BPF program types, and enabled 4710dd2ce6a5SDaniel Borkmann * for IPv4 and IPv6 protocols. 4711b4ab3141SDaniel Borkmann * Return 4712b4ab3141SDaniel Borkmann * The helper returns **TC_ACT_REDIRECT** on success or 4713b4ab3141SDaniel Borkmann * **TC_ACT_SHOT** on error. 4714eaa6bcb7SHao Luo * 4715eaa6bcb7SHao Luo * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4716eaa6bcb7SHao Luo * Description 4717eaa6bcb7SHao Luo * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4718eaa6bcb7SHao Luo * pointer to the percpu kernel variable on *cpu*. A ksym is an 4719eaa6bcb7SHao Luo * extern variable decorated with '__ksym'. For ksym, there is a 4720eaa6bcb7SHao Luo * global var (either static or global) defined of the same name 4721eaa6bcb7SHao Luo * in the kernel. The ksym is percpu if the global var is percpu. 4722eaa6bcb7SHao Luo * The returned pointer points to the global percpu var on *cpu*. 4723eaa6bcb7SHao Luo * 4724eaa6bcb7SHao Luo * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4725eaa6bcb7SHao Luo * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4726eaa6bcb7SHao Luo * happens if *cpu* is larger than nr_cpu_ids. The caller of 4727eaa6bcb7SHao Luo * bpf_per_cpu_ptr() must check the returned value. 4728eaa6bcb7SHao Luo * Return 4729eaa6bcb7SHao Luo * A pointer pointing to the kernel percpu variable on *cpu*, or 4730eaa6bcb7SHao Luo * NULL, if *cpu* is invalid. 473163d9b80dSHao Luo * 473263d9b80dSHao Luo * void *bpf_this_cpu_ptr(const void *percpu_ptr) 473363d9b80dSHao Luo * Description 473463d9b80dSHao Luo * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 473563d9b80dSHao Luo * pointer to the percpu kernel variable on this cpu. See the 473663d9b80dSHao Luo * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 473763d9b80dSHao Luo * 473863d9b80dSHao Luo * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 473963d9b80dSHao Luo * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 474063d9b80dSHao Luo * never return NULL. 474163d9b80dSHao Luo * Return 474263d9b80dSHao Luo * A pointer pointing to the kernel percpu variable on this cpu. 47439aa1206eSDaniel Borkmann * 47449aa1206eSDaniel Borkmann * long bpf_redirect_peer(u32 ifindex, u64 flags) 47459aa1206eSDaniel Borkmann * Description 47469aa1206eSDaniel Borkmann * Redirect the packet to another net device of index *ifindex*. 47479aa1206eSDaniel Borkmann * This helper is somewhat similar to **bpf_redirect**\ (), except 47489aa1206eSDaniel Borkmann * that the redirection happens to the *ifindex*' peer device and 47499aa1206eSDaniel Borkmann * the netns switch takes place from ingress to ingress without 47509aa1206eSDaniel Borkmann * going through the CPU's backlog queue. 47519aa1206eSDaniel Borkmann * 47529aa1206eSDaniel Borkmann * The *flags* argument is reserved and must be 0. The helper is 47539aa1206eSDaniel Borkmann * currently only supported for tc BPF program types at the ingress 47549aa1206eSDaniel Borkmann * hook and for veth device types. The peer device must reside in a 47559aa1206eSDaniel Borkmann * different network namespace. 47569aa1206eSDaniel Borkmann * Return 47579aa1206eSDaniel Borkmann * The helper returns **TC_ACT_REDIRECT** on success or 47589aa1206eSDaniel Borkmann * **TC_ACT_SHOT** on error. 47594cf1bc1fSKP Singh * 47604cf1bc1fSKP Singh * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 47614cf1bc1fSKP Singh * Description 47624cf1bc1fSKP Singh * Get a bpf_local_storage from the *task*. 47634cf1bc1fSKP Singh * 47644cf1bc1fSKP Singh * Logically, it could be thought of as getting the value from 47654cf1bc1fSKP Singh * a *map* with *task* as the **key**. From this 47664cf1bc1fSKP Singh * perspective, the usage is not much different from 47674cf1bc1fSKP Singh * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4768aa75622cSQuentin Monnet * helper enforces the key must be a task_struct and the map must also 47694cf1bc1fSKP Singh * be a **BPF_MAP_TYPE_TASK_STORAGE**. 47704cf1bc1fSKP Singh * 47714cf1bc1fSKP Singh * Underneath, the value is stored locally at *task* instead of 47724cf1bc1fSKP Singh * the *map*. The *map* is used as the bpf-local-storage 47734cf1bc1fSKP Singh * "type". The bpf-local-storage "type" (i.e. the *map*) is 47744cf1bc1fSKP Singh * searched against all bpf_local_storage residing at *task*. 47754cf1bc1fSKP Singh * 47764cf1bc1fSKP Singh * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 47774cf1bc1fSKP Singh * used such that a new bpf_local_storage will be 47784cf1bc1fSKP Singh * created if one does not exist. *value* can be used 47794cf1bc1fSKP Singh * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 47804cf1bc1fSKP Singh * the initial value of a bpf_local_storage. If *value* is 47814cf1bc1fSKP Singh * **NULL**, the new bpf_local_storage will be zero initialized. 47824cf1bc1fSKP Singh * Return 47834cf1bc1fSKP Singh * A bpf_local_storage pointer is returned on success. 47844cf1bc1fSKP Singh * 47854cf1bc1fSKP Singh * **NULL** if not found or there was an error in adding 47864cf1bc1fSKP Singh * a new bpf_local_storage. 47874cf1bc1fSKP Singh * 47884cf1bc1fSKP Singh * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 47894cf1bc1fSKP Singh * Description 47904cf1bc1fSKP Singh * Delete a bpf_local_storage from a *task*. 47914cf1bc1fSKP Singh * Return 47924cf1bc1fSKP Singh * 0 on success. 47934cf1bc1fSKP Singh * 47944cf1bc1fSKP Singh * **-ENOENT** if the bpf_local_storage cannot be found. 47953ca1032aSKP Singh * 47963ca1032aSKP Singh * struct task_struct *bpf_get_current_task_btf(void) 47973ca1032aSKP Singh * Description 47983ca1032aSKP Singh * Return a BTF pointer to the "current" task. 47993ca1032aSKP Singh * This pointer can also be used in helpers that accept an 48003ca1032aSKP Singh * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 48013ca1032aSKP Singh * Return 48023ca1032aSKP Singh * Pointer to the current task. 48033f6719c7SKP Singh * 48043f6719c7SKP Singh * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 48053f6719c7SKP Singh * Description 48063f6719c7SKP Singh * Set or clear certain options on *bprm*: 48073f6719c7SKP Singh * 48083f6719c7SKP Singh * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 48093f6719c7SKP Singh * which sets the **AT_SECURE** auxv for glibc. The bit 48103f6719c7SKP Singh * is cleared if the flag is not specified. 48113f6719c7SKP Singh * Return 48123f6719c7SKP Singh * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4813d0551261SDmitrii Banshchikov * 4814d0551261SDmitrii Banshchikov * u64 bpf_ktime_get_coarse_ns(void) 4815d0551261SDmitrii Banshchikov * Description 4816d0551261SDmitrii Banshchikov * Return a coarse-grained version of the time elapsed since 4817d0551261SDmitrii Banshchikov * system boot, in nanoseconds. Does not include time the system 4818d0551261SDmitrii Banshchikov * was suspended. 4819d0551261SDmitrii Banshchikov * 4820d0551261SDmitrii Banshchikov * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4821d0551261SDmitrii Banshchikov * Return 4822d0551261SDmitrii Banshchikov * Current *ktime*. 482327672f0dSKP Singh * 482427672f0dSKP Singh * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 482527672f0dSKP Singh * Description 4826aa75622cSQuentin Monnet * Returns the stored IMA hash of the *inode* (if it's available). 482727672f0dSKP Singh * If the hash is larger than *size*, then only *size* 482827672f0dSKP Singh * bytes will be copied to *dst* 482927672f0dSKP Singh * Return 483027672f0dSKP Singh * The **hash_algo** is returned on success, 483127672f0dSKP Singh * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 483227672f0dSKP Singh * invalid arguments are passed. 48334f19cab7SFlorent Revest * 48344f19cab7SFlorent Revest * struct socket *bpf_sock_from_file(struct file *file) 48354f19cab7SFlorent Revest * Description 48364f19cab7SFlorent Revest * If the given file represents a socket, returns the associated 48374f19cab7SFlorent Revest * socket. 48384f19cab7SFlorent Revest * Return 48394f19cab7SFlorent Revest * A pointer to a struct socket on success or NULL if the file is 48404f19cab7SFlorent Revest * not a socket. 484134b2021cSJesper Dangaard Brouer * 484234b2021cSJesper Dangaard Brouer * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 484334b2021cSJesper Dangaard Brouer * Description 4844e5e35e75SJesper Dangaard Brouer * Check packet size against exceeding MTU of net device (based 484534b2021cSJesper Dangaard Brouer * on *ifindex*). This helper will likely be used in combination 484634b2021cSJesper Dangaard Brouer * with helpers that adjust/change the packet size. 484734b2021cSJesper Dangaard Brouer * 484834b2021cSJesper Dangaard Brouer * The argument *len_diff* can be used for querying with a planned 484934b2021cSJesper Dangaard Brouer * size change. This allows to check MTU prior to changing packet 4850aa75622cSQuentin Monnet * ctx. Providing a *len_diff* adjustment that is larger than the 485134b2021cSJesper Dangaard Brouer * actual packet size (resulting in negative packet size) will in 4852aa75622cSQuentin Monnet * principle not exceed the MTU, which is why it is not considered 4853aa75622cSQuentin Monnet * a failure. Other BPF helpers are needed for performing the 4854aa75622cSQuentin Monnet * planned size change; therefore the responsibility for catching 4855aa75622cSQuentin Monnet * a negative packet size belongs in those helpers. 485634b2021cSJesper Dangaard Brouer * 485734b2021cSJesper Dangaard Brouer * Specifying *ifindex* zero means the MTU check is performed 485834b2021cSJesper Dangaard Brouer * against the current net device. This is practical if this isn't 485934b2021cSJesper Dangaard Brouer * used prior to redirect. 486034b2021cSJesper Dangaard Brouer * 4861e5e35e75SJesper Dangaard Brouer * On input *mtu_len* must be a valid pointer, else verifier will 4862e5e35e75SJesper Dangaard Brouer * reject BPF program. If the value *mtu_len* is initialized to 4863e5e35e75SJesper Dangaard Brouer * zero then the ctx packet size is use. When value *mtu_len* is 4864e5e35e75SJesper Dangaard Brouer * provided as input this specify the L3 length that the MTU check 4865e5e35e75SJesper Dangaard Brouer * is done against. Remember XDP and TC length operate at L2, but 4866e5e35e75SJesper Dangaard Brouer * this value is L3 as this correlate to MTU and IP-header tot_len 4867e5e35e75SJesper Dangaard Brouer * values which are L3 (similar behavior as bpf_fib_lookup). 4868e5e35e75SJesper Dangaard Brouer * 486934b2021cSJesper Dangaard Brouer * The Linux kernel route table can configure MTUs on a more 487034b2021cSJesper Dangaard Brouer * specific per route level, which is not provided by this helper. 487134b2021cSJesper Dangaard Brouer * For route level MTU checks use the **bpf_fib_lookup**\ () 487234b2021cSJesper Dangaard Brouer * helper. 487334b2021cSJesper Dangaard Brouer * 487434b2021cSJesper Dangaard Brouer * *ctx* is either **struct xdp_md** for XDP programs or 487534b2021cSJesper Dangaard Brouer * **struct sk_buff** for tc cls_act programs. 487634b2021cSJesper Dangaard Brouer * 487734b2021cSJesper Dangaard Brouer * The *flags* argument can be a combination of one or more of the 487834b2021cSJesper Dangaard Brouer * following values: 487934b2021cSJesper Dangaard Brouer * 488034b2021cSJesper Dangaard Brouer * **BPF_MTU_CHK_SEGS** 488134b2021cSJesper Dangaard Brouer * This flag will only works for *ctx* **struct sk_buff**. 488234b2021cSJesper Dangaard Brouer * If packet context contains extra packet segment buffers 488334b2021cSJesper Dangaard Brouer * (often knows as GSO skb), then MTU check is harder to 488434b2021cSJesper Dangaard Brouer * check at this point, because in transmit path it is 488534b2021cSJesper Dangaard Brouer * possible for the skb packet to get re-segmented 488634b2021cSJesper Dangaard Brouer * (depending on net device features). This could still be 488734b2021cSJesper Dangaard Brouer * a MTU violation, so this flag enables performing MTU 488834b2021cSJesper Dangaard Brouer * check against segments, with a different violation 488934b2021cSJesper Dangaard Brouer * return code to tell it apart. Check cannot use len_diff. 489034b2021cSJesper Dangaard Brouer * 489134b2021cSJesper Dangaard Brouer * On return *mtu_len* pointer contains the MTU value of the net 489234b2021cSJesper Dangaard Brouer * device. Remember the net device configured MTU is the L3 size, 4893e5e35e75SJesper Dangaard Brouer * which is returned here and XDP and TC length operate at L2. 489434b2021cSJesper Dangaard Brouer * Helper take this into account for you, but remember when using 4895e5e35e75SJesper Dangaard Brouer * MTU value in your BPF-code. 489634b2021cSJesper Dangaard Brouer * 489734b2021cSJesper Dangaard Brouer * Return 489834b2021cSJesper Dangaard Brouer * * 0 on success, and populate MTU value in *mtu_len* pointer. 489934b2021cSJesper Dangaard Brouer * 490034b2021cSJesper Dangaard Brouer * * < 0 if any input argument is invalid (*mtu_len* not updated) 490134b2021cSJesper Dangaard Brouer * 490234b2021cSJesper Dangaard Brouer * MTU violations return positive values, but also populate MTU 490334b2021cSJesper Dangaard Brouer * value in *mtu_len* pointer, as this can be needed for 490434b2021cSJesper Dangaard Brouer * implementing PMTU handing: 490534b2021cSJesper Dangaard Brouer * 490634b2021cSJesper Dangaard Brouer * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 490734b2021cSJesper Dangaard Brouer * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 490834b2021cSJesper Dangaard Brouer * 490969c087baSYonghong Song * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 491069c087baSYonghong Song * Description 491169c087baSYonghong Song * For each element in **map**, call **callback_fn** function with 491269c087baSYonghong Song * **map**, **callback_ctx** and other map-specific parameters. 491369c087baSYonghong Song * The **callback_fn** should be a static function and 491469c087baSYonghong Song * the **callback_ctx** should be a pointer to the stack. 491569c087baSYonghong Song * The **flags** is used to control certain aspects of the helper. 491669c087baSYonghong Song * Currently, the **flags** must be 0. 491769c087baSYonghong Song * 491869c087baSYonghong Song * The following are a list of supported map types and their 491969c087baSYonghong Song * respective expected callback signatures: 492069c087baSYonghong Song * 492169c087baSYonghong Song * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 492269c087baSYonghong Song * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 492369c087baSYonghong Song * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 492469c087baSYonghong Song * 492569c087baSYonghong Song * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 492669c087baSYonghong Song * 492769c087baSYonghong Song * For per_cpu maps, the map_value is the value on the cpu where the 492869c087baSYonghong Song * bpf_prog is running. 492969c087baSYonghong Song * 493069c087baSYonghong Song * If **callback_fn** return 0, the helper will continue to the next 493169c087baSYonghong Song * element. If return value is 1, the helper will skip the rest of 493269c087baSYonghong Song * elements and return. Other return values are not used now. 493369c087baSYonghong Song * 493469c087baSYonghong Song * Return 493569c087baSYonghong Song * The number of traversed map elements for success, **-EINVAL** for 493669c087baSYonghong Song * invalid **flags**. 49377b15523aSFlorent Revest * 49387b15523aSFlorent Revest * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 49397b15523aSFlorent Revest * Description 49407b15523aSFlorent Revest * Outputs a string into the **str** buffer of size **str_size** 49417b15523aSFlorent Revest * based on a format string stored in a read-only map pointed by 49427b15523aSFlorent Revest * **fmt**. 49437b15523aSFlorent Revest * 49447b15523aSFlorent Revest * Each format specifier in **fmt** corresponds to one u64 element 49457b15523aSFlorent Revest * in the **data** array. For strings and pointers where pointees 49467b15523aSFlorent Revest * are accessed, only the pointer values are stored in the *data* 4947a42effb0SDave Marchevsky * array. The *data_len* is the size of *data* in bytes - must be 4948a42effb0SDave Marchevsky * a multiple of 8. 49497b15523aSFlorent Revest * 49507b15523aSFlorent Revest * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 49517b15523aSFlorent Revest * memory. Reading kernel memory may fail due to either invalid 49527b15523aSFlorent Revest * address or valid address but requiring a major memory fault. If 49537b15523aSFlorent Revest * reading kernel memory fails, the string for **%s** will be an 49547b15523aSFlorent Revest * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 49557b15523aSFlorent Revest * Not returning error to bpf program is consistent with what 49567b15523aSFlorent Revest * **bpf_trace_printk**\ () does for now. 49577b15523aSFlorent Revest * 49587b15523aSFlorent Revest * Return 49597b15523aSFlorent Revest * The strictly positive length of the formatted string, including 49607b15523aSFlorent Revest * the trailing zero character. If the return value is greater than 49617b15523aSFlorent Revest * **str_size**, **str** contains a truncated string, guaranteed to 49627b15523aSFlorent Revest * be zero-terminated except when **str_size** is 0. 49637b15523aSFlorent Revest * 49647b15523aSFlorent Revest * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 496579a7f8bdSAlexei Starovoitov * 496679a7f8bdSAlexei Starovoitov * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 496779a7f8bdSAlexei Starovoitov * Description 496879a7f8bdSAlexei Starovoitov * Execute bpf syscall with given arguments. 496979a7f8bdSAlexei Starovoitov * Return 497079a7f8bdSAlexei Starovoitov * A syscall result. 49713d78417bSAlexei Starovoitov * 49723d78417bSAlexei Starovoitov * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 49733d78417bSAlexei Starovoitov * Description 49743d78417bSAlexei Starovoitov * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 49753d78417bSAlexei Starovoitov * Return 49763d78417bSAlexei Starovoitov * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 49773abea089SAlexei Starovoitov * 49783abea089SAlexei Starovoitov * long bpf_sys_close(u32 fd) 49793abea089SAlexei Starovoitov * Description 49803abea089SAlexei Starovoitov * Execute close syscall for given FD. 49813abea089SAlexei Starovoitov * Return 49823abea089SAlexei Starovoitov * A syscall result. 4983b00628b1SAlexei Starovoitov * 4984b00628b1SAlexei Starovoitov * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 4985b00628b1SAlexei Starovoitov * Description 4986b00628b1SAlexei Starovoitov * Initialize the timer. 4987b00628b1SAlexei Starovoitov * First 4 bits of *flags* specify clockid. 4988b00628b1SAlexei Starovoitov * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 4989b00628b1SAlexei Starovoitov * All other bits of *flags* are reserved. 4990b00628b1SAlexei Starovoitov * The verifier will reject the program if *timer* is not from 4991b00628b1SAlexei Starovoitov * the same *map*. 4992b00628b1SAlexei Starovoitov * Return 4993b00628b1SAlexei Starovoitov * 0 on success. 4994b00628b1SAlexei Starovoitov * **-EBUSY** if *timer* is already initialized. 4995b00628b1SAlexei Starovoitov * **-EINVAL** if invalid *flags* are passed. 4996b00628b1SAlexei Starovoitov * **-EPERM** if *timer* is in a map that doesn't have any user references. 4997b00628b1SAlexei Starovoitov * The user space should either hold a file descriptor to a map with timers 4998b00628b1SAlexei Starovoitov * or pin such map in bpffs. When map is unpinned or file descriptor is 4999b00628b1SAlexei Starovoitov * closed all timers in the map will be cancelled and freed. 5000b00628b1SAlexei Starovoitov * 5001b00628b1SAlexei Starovoitov * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 5002b00628b1SAlexei Starovoitov * Description 5003b00628b1SAlexei Starovoitov * Configure the timer to call *callback_fn* static function. 5004b00628b1SAlexei Starovoitov * Return 5005b00628b1SAlexei Starovoitov * 0 on success. 5006b00628b1SAlexei Starovoitov * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5007b00628b1SAlexei Starovoitov * **-EPERM** if *timer* is in a map that doesn't have any user references. 5008b00628b1SAlexei Starovoitov * The user space should either hold a file descriptor to a map with timers 5009b00628b1SAlexei Starovoitov * or pin such map in bpffs. When map is unpinned or file descriptor is 5010b00628b1SAlexei Starovoitov * closed all timers in the map will be cancelled and freed. 5011b00628b1SAlexei Starovoitov * 5012b00628b1SAlexei Starovoitov * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 5013b00628b1SAlexei Starovoitov * Description 5014b00628b1SAlexei Starovoitov * Set timer expiration N nanoseconds from the current time. The 5015b00628b1SAlexei Starovoitov * configured callback will be invoked in soft irq context on some cpu 5016b00628b1SAlexei Starovoitov * and will not repeat unless another bpf_timer_start() is made. 5017b00628b1SAlexei Starovoitov * In such case the next invocation can migrate to a different cpu. 5018b00628b1SAlexei Starovoitov * Since struct bpf_timer is a field inside map element the map 5019b00628b1SAlexei Starovoitov * owns the timer. The bpf_timer_set_callback() will increment refcnt 5020b00628b1SAlexei Starovoitov * of BPF program to make sure that callback_fn code stays valid. 5021b00628b1SAlexei Starovoitov * When user space reference to a map reaches zero all timers 5022b00628b1SAlexei Starovoitov * in a map are cancelled and corresponding program's refcnts are 5023b00628b1SAlexei Starovoitov * decremented. This is done to make sure that Ctrl-C of a user 5024b00628b1SAlexei Starovoitov * process doesn't leave any timers running. If map is pinned in 5025b00628b1SAlexei Starovoitov * bpffs the callback_fn can re-arm itself indefinitely. 5026b00628b1SAlexei Starovoitov * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 5027b00628b1SAlexei Starovoitov * cancel and free the timer in the given map element. 5028b00628b1SAlexei Starovoitov * The map can contain timers that invoke callback_fn-s from different 5029b00628b1SAlexei Starovoitov * programs. The same callback_fn can serve different timers from 5030b00628b1SAlexei Starovoitov * different maps if key/value layout matches across maps. 5031b00628b1SAlexei Starovoitov * Every bpf_timer_set_callback() can have different callback_fn. 5032b00628b1SAlexei Starovoitov * 5033f71f8530STero Kristo * *flags* can be one of: 5034f71f8530STero Kristo * 5035f71f8530STero Kristo * **BPF_F_TIMER_ABS** 5036f71f8530STero Kristo * Start the timer in absolute expire value instead of the 5037f71f8530STero Kristo * default relative one. 5038f71f8530STero Kristo * 5039b00628b1SAlexei Starovoitov * Return 5040b00628b1SAlexei Starovoitov * 0 on success. 5041b00628b1SAlexei Starovoitov * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 5042b00628b1SAlexei Starovoitov * or invalid *flags* are passed. 5043b00628b1SAlexei Starovoitov * 5044b00628b1SAlexei Starovoitov * long bpf_timer_cancel(struct bpf_timer *timer) 5045b00628b1SAlexei Starovoitov * Description 5046b00628b1SAlexei Starovoitov * Cancel the timer and wait for callback_fn to finish if it was running. 5047b00628b1SAlexei Starovoitov * Return 5048b00628b1SAlexei Starovoitov * 0 if the timer was not active. 5049b00628b1SAlexei Starovoitov * 1 if the timer was active. 5050b00628b1SAlexei Starovoitov * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5051b00628b1SAlexei Starovoitov * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 5052b00628b1SAlexei Starovoitov * own timer which would have led to a deadlock otherwise. 50539b99edcaSJiri Olsa * 50549b99edcaSJiri Olsa * u64 bpf_get_func_ip(void *ctx) 50559b99edcaSJiri Olsa * Description 50569ffd9f3fSJiri Olsa * Get address of the traced function (for tracing and kprobe programs). 50579b99edcaSJiri Olsa * Return 50589b99edcaSJiri Olsa * Address of the traced function. 50590e253f7eSJiri Olsa * 0 for kprobes placed within the function (not at the entry). 50607adfc6c9SAndrii Nakryiko * 50617adfc6c9SAndrii Nakryiko * u64 bpf_get_attach_cookie(void *ctx) 50627adfc6c9SAndrii Nakryiko * Description 50637adfc6c9SAndrii Nakryiko * Get bpf_cookie value provided (optionally) during the program 50647adfc6c9SAndrii Nakryiko * attachment. It might be different for each individual 50657adfc6c9SAndrii Nakryiko * attachment, even if BPF program itself is the same. 50667adfc6c9SAndrii Nakryiko * Expects BPF program context *ctx* as a first argument. 50677adfc6c9SAndrii Nakryiko * 50687adfc6c9SAndrii Nakryiko * Supported for the following program types: 50697adfc6c9SAndrii Nakryiko * - kprobe/uprobe; 50707adfc6c9SAndrii Nakryiko * - tracepoint; 50717adfc6c9SAndrii Nakryiko * - perf_event. 50727adfc6c9SAndrii Nakryiko * Return 50737adfc6c9SAndrii Nakryiko * Value specified by user at BPF link creation/attachment time 50747adfc6c9SAndrii Nakryiko * or 0, if it was not specified. 5075dd6e10fbSDaniel Xu * 5076dd6e10fbSDaniel Xu * long bpf_task_pt_regs(struct task_struct *task) 5077dd6e10fbSDaniel Xu * Description 5078dd6e10fbSDaniel Xu * Get the struct pt_regs associated with **task**. 5079dd6e10fbSDaniel Xu * Return 5080dd6e10fbSDaniel Xu * A pointer to struct pt_regs. 5081856c02dbSSong Liu * 5082856c02dbSSong Liu * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 5083856c02dbSSong Liu * Description 5084856c02dbSSong Liu * Get branch trace from hardware engines like Intel LBR. The 5085856c02dbSSong Liu * hardware engine is stopped shortly after the helper is 5086856c02dbSSong Liu * called. Therefore, the user need to filter branch entries 5087856c02dbSSong Liu * based on the actual use case. To capture branch trace 5088856c02dbSSong Liu * before the trigger point of the BPF program, the helper 5089856c02dbSSong Liu * should be called at the beginning of the BPF program. 5090856c02dbSSong Liu * 5091856c02dbSSong Liu * The data is stored as struct perf_branch_entry into output 5092856c02dbSSong Liu * buffer *entries*. *size* is the size of *entries* in bytes. 5093856c02dbSSong Liu * *flags* is reserved for now and must be zero. 5094856c02dbSSong Liu * 5095856c02dbSSong Liu * Return 5096856c02dbSSong Liu * On success, number of bytes written to *buf*. On error, a 5097856c02dbSSong Liu * negative value. 5098856c02dbSSong Liu * 5099856c02dbSSong Liu * **-EINVAL** if *flags* is not zero. 5100856c02dbSSong Liu * 5101856c02dbSSong Liu * **-ENOENT** if architecture does not support branch records. 510210aceb62SDave Marchevsky * 510310aceb62SDave Marchevsky * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 510410aceb62SDave Marchevsky * Description 510510aceb62SDave Marchevsky * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 510610aceb62SDave Marchevsky * to format and can handle more format args as a result. 510710aceb62SDave Marchevsky * 510810aceb62SDave Marchevsky * Arguments are to be used as in **bpf_seq_printf**\ () helper. 510910aceb62SDave Marchevsky * Return 511010aceb62SDave Marchevsky * The number of bytes written to the buffer, or a negative error 511110aceb62SDave Marchevsky * in case of failure. 51129eeb3aa3SHengqi Chen * 51139eeb3aa3SHengqi Chen * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 51149eeb3aa3SHengqi Chen * Description 51159eeb3aa3SHengqi Chen * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 51169eeb3aa3SHengqi Chen * Return 51179eeb3aa3SHengqi Chen * *sk* if casting is valid, or **NULL** otherwise. 5118d6aef08aSKumar Kartikeya Dwivedi * 5119d6aef08aSKumar Kartikeya Dwivedi * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 5120d6aef08aSKumar Kartikeya Dwivedi * Description 5121d6aef08aSKumar Kartikeya Dwivedi * Get the address of a kernel symbol, returned in *res*. *res* is 5122d6aef08aSKumar Kartikeya Dwivedi * set to 0 if the symbol is not found. 5123d6aef08aSKumar Kartikeya Dwivedi * Return 5124d6aef08aSKumar Kartikeya Dwivedi * On success, zero. On error, a negative value. 5125d6aef08aSKumar Kartikeya Dwivedi * 5126d6aef08aSKumar Kartikeya Dwivedi * **-EINVAL** if *flags* is not zero. 5127d6aef08aSKumar Kartikeya Dwivedi * 5128d6aef08aSKumar Kartikeya Dwivedi * **-EINVAL** if string *name* is not the same size as *name_sz*. 5129d6aef08aSKumar Kartikeya Dwivedi * 5130d6aef08aSKumar Kartikeya Dwivedi * **-ENOENT** if symbol is not found. 5131d6aef08aSKumar Kartikeya Dwivedi * 5132d6aef08aSKumar Kartikeya Dwivedi * **-EPERM** if caller does not have permission to obtain kernel address. 51337c7e3d31SSong Liu * 51347c7e3d31SSong Liu * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 51357c7e3d31SSong Liu * Description 51367c7e3d31SSong Liu * Find vma of *task* that contains *addr*, call *callback_fn* 51377c7e3d31SSong Liu * function with *task*, *vma*, and *callback_ctx*. 51387c7e3d31SSong Liu * The *callback_fn* should be a static function and 51397c7e3d31SSong Liu * the *callback_ctx* should be a pointer to the stack. 51407c7e3d31SSong Liu * The *flags* is used to control certain aspects of the helper. 51417c7e3d31SSong Liu * Currently, the *flags* must be 0. 51427c7e3d31SSong Liu * 51437c7e3d31SSong Liu * The expected callback signature is 51447c7e3d31SSong Liu * 51457c7e3d31SSong Liu * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 51467c7e3d31SSong Liu * 51477c7e3d31SSong Liu * Return 51487c7e3d31SSong Liu * 0 on success. 51497c7e3d31SSong Liu * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 51507c7e3d31SSong Liu * **-EBUSY** if failed to try lock mmap_lock. 51517c7e3d31SSong Liu * **-EINVAL** for invalid **flags**. 5152e6f2dd0fSJoanne Koong * 5153e6f2dd0fSJoanne Koong * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 5154e6f2dd0fSJoanne Koong * Description 5155e6f2dd0fSJoanne Koong * For **nr_loops**, call **callback_fn** function 5156e6f2dd0fSJoanne Koong * with **callback_ctx** as the context parameter. 5157e6f2dd0fSJoanne Koong * The **callback_fn** should be a static function and 5158e6f2dd0fSJoanne Koong * the **callback_ctx** should be a pointer to the stack. 5159e6f2dd0fSJoanne Koong * The **flags** is used to control certain aspects of the helper. 5160e6f2dd0fSJoanne Koong * Currently, the **flags** must be 0. Currently, nr_loops is 5161e6f2dd0fSJoanne Koong * limited to 1 << 23 (~8 million) loops. 5162e6f2dd0fSJoanne Koong * 5163e6f2dd0fSJoanne Koong * long (\*callback_fn)(u32 index, void \*ctx); 5164e6f2dd0fSJoanne Koong * 5165e6f2dd0fSJoanne Koong * where **index** is the current index in the loop. The index 5166e6f2dd0fSJoanne Koong * is zero-indexed. 5167e6f2dd0fSJoanne Koong * 5168e6f2dd0fSJoanne Koong * If **callback_fn** returns 0, the helper will continue to the next 5169e6f2dd0fSJoanne Koong * loop. If return value is 1, the helper will skip the rest of 5170e6f2dd0fSJoanne Koong * the loops and return. Other return values are not used now, 5171e6f2dd0fSJoanne Koong * and will be rejected by the verifier. 5172e6f2dd0fSJoanne Koong * 5173e6f2dd0fSJoanne Koong * Return 5174e6f2dd0fSJoanne Koong * The number of loops performed, **-EINVAL** for invalid **flags**, 5175e6f2dd0fSJoanne Koong * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5176c5fb1993SHou Tao * 5177c5fb1993SHou Tao * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5178c5fb1993SHou Tao * Description 5179c5fb1993SHou Tao * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5180c5fb1993SHou Tao * to be null-terminated and **s1_sz** is the maximum storage 5181c5fb1993SHou Tao * size of **s1**. **s2** must be a read-only string. 5182c5fb1993SHou Tao * Return 5183c5fb1993SHou Tao * An integer less than, equal to, or greater than zero 5184c5fb1993SHou Tao * if the first **s1_sz** bytes of **s1** is found to be 5185c5fb1993SHou Tao * less than, to match, or be greater than **s2**. 5186f92c1e18SJiri Olsa * 5187f92c1e18SJiri Olsa * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5188f92c1e18SJiri Olsa * Description 518927ed9353SYonghong Song * Get **n**-th argument register (zero based) of the traced function (for tracing programs) 5190f92c1e18SJiri Olsa * returned in **value**. 5191f92c1e18SJiri Olsa * 5192f92c1e18SJiri Olsa * Return 5193f92c1e18SJiri Olsa * 0 on success. 519427ed9353SYonghong Song * **-EINVAL** if n >= argument register count of traced function. 5195f92c1e18SJiri Olsa * 5196f92c1e18SJiri Olsa * long bpf_get_func_ret(void *ctx, u64 *value) 5197f92c1e18SJiri Olsa * Description 5198f92c1e18SJiri Olsa * Get return value of the traced function (for tracing programs) 5199f92c1e18SJiri Olsa * in **value**. 5200f92c1e18SJiri Olsa * 5201f92c1e18SJiri Olsa * Return 5202f92c1e18SJiri Olsa * 0 on success. 5203f92c1e18SJiri Olsa * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5204f92c1e18SJiri Olsa * 5205f92c1e18SJiri Olsa * long bpf_get_func_arg_cnt(void *ctx) 5206f92c1e18SJiri Olsa * Description 520727ed9353SYonghong Song * Get number of registers of the traced function (for tracing programs) where 520827ed9353SYonghong Song * function arguments are stored in these registers. 5209f92c1e18SJiri Olsa * 5210f92c1e18SJiri Olsa * Return 521127ed9353SYonghong Song * The number of argument registers of the traced function. 5212b44123b4SYiFei Zhu * 5213b44123b4SYiFei Zhu * int bpf_get_retval(void) 5214b44123b4SYiFei Zhu * Description 52152172fb80SStanislav Fomichev * Get the BPF program's return value that will be returned to the upper layers. 5216b44123b4SYiFei Zhu * 52172172fb80SStanislav Fomichev * This helper is currently supported by cgroup programs and only by the hooks 52182172fb80SStanislav Fomichev * where BPF program's return value is returned to the userspace via errno. 5219b44123b4SYiFei Zhu * Return 52202172fb80SStanislav Fomichev * The BPF program's return value. 5221b44123b4SYiFei Zhu * 5222b44123b4SYiFei Zhu * int bpf_set_retval(int retval) 5223b44123b4SYiFei Zhu * Description 52242172fb80SStanislav Fomichev * Set the BPF program's return value that will be returned to the upper layers. 5225b44123b4SYiFei Zhu * 52262172fb80SStanislav Fomichev * This helper is currently supported by cgroup programs and only by the hooks 52272172fb80SStanislav Fomichev * where BPF program's return value is returned to the userspace via errno. 52282172fb80SStanislav Fomichev * 52292172fb80SStanislav Fomichev * Note that there is the following corner case where the program exports an error 52302172fb80SStanislav Fomichev * via bpf_set_retval but signals success via 'return 1': 52312172fb80SStanislav Fomichev * 52322172fb80SStanislav Fomichev * bpf_set_retval(-EPERM); 52332172fb80SStanislav Fomichev * return 1; 52342172fb80SStanislav Fomichev * 52352172fb80SStanislav Fomichev * In this case, the BPF program's return value will use helper's -EPERM. This 52362172fb80SStanislav Fomichev * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case. 52372172fb80SStanislav Fomichev * 5238b44123b4SYiFei Zhu * Return 5239b44123b4SYiFei Zhu * 0 on success, or a negative error in case of failure. 52400165cc81SLorenzo Bianconi * 52410165cc81SLorenzo Bianconi * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 52420165cc81SLorenzo Bianconi * Description 52430165cc81SLorenzo Bianconi * Get the total size of a given xdp buff (linear and paged area) 52440165cc81SLorenzo Bianconi * Return 52450165cc81SLorenzo Bianconi * The total size of a given xdp buffer. 52463f364222SLorenzo Bianconi * 52473f364222SLorenzo Bianconi * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 52483f364222SLorenzo Bianconi * Description 52493f364222SLorenzo Bianconi * This helper is provided as an easy way to load data from a 52503f364222SLorenzo Bianconi * xdp buffer. It can be used to load *len* bytes from *offset* from 52513f364222SLorenzo Bianconi * the frame associated to *xdp_md*, into the buffer pointed by 52523f364222SLorenzo Bianconi * *buf*. 52533f364222SLorenzo Bianconi * Return 52543f364222SLorenzo Bianconi * 0 on success, or a negative error in case of failure. 52553f364222SLorenzo Bianconi * 52563f364222SLorenzo Bianconi * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 52573f364222SLorenzo Bianconi * Description 52583f364222SLorenzo Bianconi * Store *len* bytes from buffer *buf* into the frame 52593f364222SLorenzo Bianconi * associated to *xdp_md*, at *offset*. 52603f364222SLorenzo Bianconi * Return 52613f364222SLorenzo Bianconi * 0 on success, or a negative error in case of failure. 5262376040e4SKenny Yu * 5263376040e4SKenny Yu * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5264376040e4SKenny Yu * Description 5265376040e4SKenny Yu * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5266376040e4SKenny Yu * address space, and stores the data in *dst*. *flags* is not 5267376040e4SKenny Yu * used yet and is provided for future extensibility. This helper 5268376040e4SKenny Yu * can only be used by sleepable programs. 5269376040e4SKenny Yu * Return 5270376040e4SKenny Yu * 0 on success, or a negative error in case of failure. On error 5271376040e4SKenny Yu * *dst* buffer is zeroed out. 52728d21ec0eSMartin KaFai Lau * 52739bb984f2SMartin KaFai Lau * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 52748d21ec0eSMartin KaFai Lau * Description 52759bb984f2SMartin KaFai Lau * Change the __sk_buff->tstamp_type to *tstamp_type* 52769bb984f2SMartin KaFai Lau * and set *tstamp* to the __sk_buff->tstamp together. 52778d21ec0eSMartin KaFai Lau * 52789bb984f2SMartin KaFai Lau * If there is no need to change the __sk_buff->tstamp_type, 52799bb984f2SMartin KaFai Lau * the tstamp value can be directly written to __sk_buff->tstamp 52808d21ec0eSMartin KaFai Lau * instead. 52818d21ec0eSMartin KaFai Lau * 52829bb984f2SMartin KaFai Lau * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 52839bb984f2SMartin KaFai Lau * will be kept during bpf_redirect_*(). A non zero 52849bb984f2SMartin KaFai Lau * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 52859bb984f2SMartin KaFai Lau * *tstamp_type*. 52869bb984f2SMartin KaFai Lau * 52879bb984f2SMartin KaFai Lau * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 52889bb984f2SMartin KaFai Lau * with a zero *tstamp*. 52898d21ec0eSMartin KaFai Lau * 52908d21ec0eSMartin KaFai Lau * Only IPv4 and IPv6 skb->protocol are supported. 52918d21ec0eSMartin KaFai Lau * 52928d21ec0eSMartin KaFai Lau * This function is most useful when it needs to set a 52938d21ec0eSMartin KaFai Lau * mono delivery time to __sk_buff->tstamp and then 52948d21ec0eSMartin KaFai Lau * bpf_redirect_*() to the egress of an iface. For example, 52958d21ec0eSMartin KaFai Lau * changing the (rcv) timestamp in __sk_buff->tstamp at 52968d21ec0eSMartin KaFai Lau * ingress to a mono delivery time and then bpf_redirect_*() 52978d21ec0eSMartin KaFai Lau * to sch_fq@phy-dev. 52988d21ec0eSMartin KaFai Lau * Return 52998d21ec0eSMartin KaFai Lau * 0 on success. 53008d21ec0eSMartin KaFai Lau * **-EINVAL** for invalid input 53019bb984f2SMartin KaFai Lau * **-EOPNOTSUPP** for unsupported protocol 5302174b1694SRoberto Sassu * 5303174b1694SRoberto Sassu * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5304174b1694SRoberto Sassu * Description 5305174b1694SRoberto Sassu * Returns a calculated IMA hash of the *file*. 5306174b1694SRoberto Sassu * If the hash is larger than *size*, then only *size* 5307174b1694SRoberto Sassu * bytes will be copied to *dst* 5308174b1694SRoberto Sassu * Return 5309174b1694SRoberto Sassu * The **hash_algo** is returned on success, 5310174b1694SRoberto Sassu * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if 5311174b1694SRoberto Sassu * invalid arguments are passed. 5312c0a5a21cSKumar Kartikeya Dwivedi * 5313c0a5a21cSKumar Kartikeya Dwivedi * void *bpf_kptr_xchg(void *map_value, void *ptr) 5314c0a5a21cSKumar Kartikeya Dwivedi * Description 5315c0a5a21cSKumar Kartikeya Dwivedi * Exchange kptr at pointer *map_value* with *ptr*, and return the 5316c0a5a21cSKumar Kartikeya Dwivedi * old value. *ptr* can be NULL, otherwise it must be a referenced 5317c0a5a21cSKumar Kartikeya Dwivedi * pointer which will be released when this helper is called. 5318c0a5a21cSKumar Kartikeya Dwivedi * Return 5319c0a5a21cSKumar Kartikeya Dwivedi * The old value of kptr (which can be NULL). The returned pointer 5320c0a5a21cSKumar Kartikeya Dwivedi * if not NULL, is a reference which must be released using its 5321c0a5a21cSKumar Kartikeya Dwivedi * corresponding release function, or moved into a BPF map before 5322c0a5a21cSKumar Kartikeya Dwivedi * program exit. 532307343110SFeng Zhou * 532407343110SFeng Zhou * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu) 532507343110SFeng Zhou * Description 532607343110SFeng Zhou * Perform a lookup in *percpu map* for an entry associated to 532707343110SFeng Zhou * *key* on *cpu*. 532807343110SFeng Zhou * Return 532907343110SFeng Zhou * Map value associated to *key* on *cpu*, or **NULL** if no entry 533007343110SFeng Zhou * was found or *cpu* is invalid. 53313bc253c2SGeliang Tang * 53323bc253c2SGeliang Tang * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk) 53333bc253c2SGeliang Tang * Description 53343bc253c2SGeliang Tang * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 53353bc253c2SGeliang Tang * Return 53363bc253c2SGeliang Tang * *sk* if casting is valid, or **NULL** otherwise. 5337263ae152SJoanne Koong * 5338263ae152SJoanne Koong * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr) 5339263ae152SJoanne Koong * Description 5340263ae152SJoanne Koong * Get a dynptr to local memory *data*. 5341263ae152SJoanne Koong * 5342263ae152SJoanne Koong * *data* must be a ptr to a map value. 5343263ae152SJoanne Koong * The maximum *size* supported is DYNPTR_MAX_SIZE. 5344263ae152SJoanne Koong * *flags* is currently unused. 5345263ae152SJoanne Koong * Return 5346263ae152SJoanne Koong * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 5347263ae152SJoanne Koong * -EINVAL if flags is not 0. 5348bc34dee6SJoanne Koong * 5349bc34dee6SJoanne Koong * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr) 5350bc34dee6SJoanne Koong * Description 5351bc34dee6SJoanne Koong * Reserve *size* bytes of payload in a ring buffer *ringbuf* 5352bc34dee6SJoanne Koong * through the dynptr interface. *flags* must be 0. 5353bc34dee6SJoanne Koong * 5354bc34dee6SJoanne Koong * Please note that a corresponding bpf_ringbuf_submit_dynptr or 5355bc34dee6SJoanne Koong * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 5356bc34dee6SJoanne Koong * reservation fails. This is enforced by the verifier. 5357bc34dee6SJoanne Koong * Return 5358bc34dee6SJoanne Koong * 0 on success, or a negative error in case of failure. 5359bc34dee6SJoanne Koong * 5360bc34dee6SJoanne Koong * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags) 5361bc34dee6SJoanne Koong * Description 5362bc34dee6SJoanne Koong * Submit reserved ring buffer sample, pointed to by *data*, 5363bc34dee6SJoanne Koong * through the dynptr interface. This is a no-op if the dynptr is 5364bc34dee6SJoanne Koong * invalid/null. 5365bc34dee6SJoanne Koong * 5366bc34dee6SJoanne Koong * For more information on *flags*, please see 5367bc34dee6SJoanne Koong * 'bpf_ringbuf_submit'. 5368bc34dee6SJoanne Koong * Return 5369bc34dee6SJoanne Koong * Nothing. Always succeeds. 5370bc34dee6SJoanne Koong * 5371bc34dee6SJoanne Koong * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags) 5372bc34dee6SJoanne Koong * Description 5373bc34dee6SJoanne Koong * Discard reserved ring buffer sample through the dynptr 5374bc34dee6SJoanne Koong * interface. This is a no-op if the dynptr is invalid/null. 5375bc34dee6SJoanne Koong * 5376bc34dee6SJoanne Koong * For more information on *flags*, please see 5377bc34dee6SJoanne Koong * 'bpf_ringbuf_discard'. 5378bc34dee6SJoanne Koong * Return 5379bc34dee6SJoanne Koong * Nothing. Always succeeds. 538013bbbfbeSJoanne Koong * 538127060531SKumar Kartikeya Dwivedi * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags) 538213bbbfbeSJoanne Koong * Description 538313bbbfbeSJoanne Koong * Read *len* bytes from *src* into *dst*, starting from *offset* 538413bbbfbeSJoanne Koong * into *src*. 5385f8d3da4eSJoanne Koong * *flags* is currently unused. 538613bbbfbeSJoanne Koong * Return 538713bbbfbeSJoanne Koong * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5388f8d3da4eSJoanne Koong * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 5389f8d3da4eSJoanne Koong * *flags* is not 0. 539013bbbfbeSJoanne Koong * 539127060531SKumar Kartikeya Dwivedi * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags) 539213bbbfbeSJoanne Koong * Description 539313bbbfbeSJoanne Koong * Write *len* bytes from *src* into *dst*, starting from *offset* 539413bbbfbeSJoanne Koong * into *dst*. 5395b5964b96SJoanne Koong * 5396b5964b96SJoanne Koong * *flags* must be 0 except for skb-type dynptrs. 5397b5964b96SJoanne Koong * 5398b5964b96SJoanne Koong * For skb-type dynptrs: 539966e3a13eSJoanne Koong * * All data slices of the dynptr are automatically 540066e3a13eSJoanne Koong * invalidated after **bpf_dynptr_write**\ (). This is 540166e3a13eSJoanne Koong * because writing may pull the skb and change the 540266e3a13eSJoanne Koong * underlying packet buffer. 540366e3a13eSJoanne Koong * 5404b5964b96SJoanne Koong * * For *flags*, please see the flags accepted by 5405b5964b96SJoanne Koong * **bpf_skb_store_bytes**\ (). 540613bbbfbeSJoanne Koong * Return 540713bbbfbeSJoanne Koong * 0 on success, -E2BIG if *offset* + *len* exceeds the length 540813bbbfbeSJoanne Koong * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 5409b5964b96SJoanne Koong * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs, 5410b5964b96SJoanne Koong * other errors correspond to errors returned by **bpf_skb_store_bytes**\ (). 541134d4ef57SJoanne Koong * 541227060531SKumar Kartikeya Dwivedi * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len) 541334d4ef57SJoanne Koong * Description 541434d4ef57SJoanne Koong * Get a pointer to the underlying dynptr data. 541534d4ef57SJoanne Koong * 541634d4ef57SJoanne Koong * *len* must be a statically known value. The returned data slice 541734d4ef57SJoanne Koong * is invalidated whenever the dynptr is invalidated. 5418b5964b96SJoanne Koong * 541905421aecSJoanne Koong * skb and xdp type dynptrs may not use bpf_dynptr_data. They should 5420b5964b96SJoanne Koong * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr. 542134d4ef57SJoanne Koong * Return 542234d4ef57SJoanne Koong * Pointer to the underlying dynptr data, NULL if the dynptr is 542334d4ef57SJoanne Koong * read-only, if the dynptr is invalid, or if the offset and length 542434d4ef57SJoanne Koong * is out of bounds. 542533bf9885SMaxim Mikityanskiy * 542633bf9885SMaxim Mikityanskiy * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len) 542733bf9885SMaxim Mikityanskiy * Description 542833bf9885SMaxim Mikityanskiy * Try to issue a SYN cookie for the packet with corresponding 542933bf9885SMaxim Mikityanskiy * IPv4/TCP headers, *iph* and *th*, without depending on a 543033bf9885SMaxim Mikityanskiy * listening socket. 543133bf9885SMaxim Mikityanskiy * 543233bf9885SMaxim Mikityanskiy * *iph* points to the IPv4 header. 543333bf9885SMaxim Mikityanskiy * 543433bf9885SMaxim Mikityanskiy * *th* points to the start of the TCP header, while *th_len* 543533bf9885SMaxim Mikityanskiy * contains the length of the TCP header (at least 543633bf9885SMaxim Mikityanskiy * **sizeof**\ (**struct tcphdr**)). 543733bf9885SMaxim Mikityanskiy * Return 543833bf9885SMaxim Mikityanskiy * On success, lower 32 bits hold the generated SYN cookie in 543933bf9885SMaxim Mikityanskiy * followed by 16 bits which hold the MSS value for that cookie, 544033bf9885SMaxim Mikityanskiy * and the top 16 bits are unused. 544133bf9885SMaxim Mikityanskiy * 544233bf9885SMaxim Mikityanskiy * On failure, the returned value is one of the following: 544333bf9885SMaxim Mikityanskiy * 544433bf9885SMaxim Mikityanskiy * **-EINVAL** if *th_len* is invalid. 544533bf9885SMaxim Mikityanskiy * 544633bf9885SMaxim Mikityanskiy * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len) 544733bf9885SMaxim Mikityanskiy * Description 544833bf9885SMaxim Mikityanskiy * Try to issue a SYN cookie for the packet with corresponding 544933bf9885SMaxim Mikityanskiy * IPv6/TCP headers, *iph* and *th*, without depending on a 545033bf9885SMaxim Mikityanskiy * listening socket. 545133bf9885SMaxim Mikityanskiy * 545233bf9885SMaxim Mikityanskiy * *iph* points to the IPv6 header. 545333bf9885SMaxim Mikityanskiy * 545433bf9885SMaxim Mikityanskiy * *th* points to the start of the TCP header, while *th_len* 545533bf9885SMaxim Mikityanskiy * contains the length of the TCP header (at least 545633bf9885SMaxim Mikityanskiy * **sizeof**\ (**struct tcphdr**)). 545733bf9885SMaxim Mikityanskiy * Return 545833bf9885SMaxim Mikityanskiy * On success, lower 32 bits hold the generated SYN cookie in 545933bf9885SMaxim Mikityanskiy * followed by 16 bits which hold the MSS value for that cookie, 546033bf9885SMaxim Mikityanskiy * and the top 16 bits are unused. 546133bf9885SMaxim Mikityanskiy * 546233bf9885SMaxim Mikityanskiy * On failure, the returned value is one of the following: 546333bf9885SMaxim Mikityanskiy * 546433bf9885SMaxim Mikityanskiy * **-EINVAL** if *th_len* is invalid. 546533bf9885SMaxim Mikityanskiy * 546633bf9885SMaxim Mikityanskiy * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 546733bf9885SMaxim Mikityanskiy * 546833bf9885SMaxim Mikityanskiy * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th) 546933bf9885SMaxim Mikityanskiy * Description 547033bf9885SMaxim Mikityanskiy * Check whether *iph* and *th* contain a valid SYN cookie ACK 547133bf9885SMaxim Mikityanskiy * without depending on a listening socket. 547233bf9885SMaxim Mikityanskiy * 547333bf9885SMaxim Mikityanskiy * *iph* points to the IPv4 header. 547433bf9885SMaxim Mikityanskiy * 547533bf9885SMaxim Mikityanskiy * *th* points to the TCP header. 547633bf9885SMaxim Mikityanskiy * Return 547733bf9885SMaxim Mikityanskiy * 0 if *iph* and *th* are a valid SYN cookie ACK. 547833bf9885SMaxim Mikityanskiy * 547933bf9885SMaxim Mikityanskiy * On failure, the returned value is one of the following: 548033bf9885SMaxim Mikityanskiy * 548133bf9885SMaxim Mikityanskiy * **-EACCES** if the SYN cookie is not valid. 548233bf9885SMaxim Mikityanskiy * 548333bf9885SMaxim Mikityanskiy * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th) 548433bf9885SMaxim Mikityanskiy * Description 548533bf9885SMaxim Mikityanskiy * Check whether *iph* and *th* contain a valid SYN cookie ACK 548633bf9885SMaxim Mikityanskiy * without depending on a listening socket. 548733bf9885SMaxim Mikityanskiy * 548833bf9885SMaxim Mikityanskiy * *iph* points to the IPv6 header. 548933bf9885SMaxim Mikityanskiy * 549033bf9885SMaxim Mikityanskiy * *th* points to the TCP header. 549133bf9885SMaxim Mikityanskiy * Return 549233bf9885SMaxim Mikityanskiy * 0 if *iph* and *th* are a valid SYN cookie ACK. 549333bf9885SMaxim Mikityanskiy * 549433bf9885SMaxim Mikityanskiy * On failure, the returned value is one of the following: 549533bf9885SMaxim Mikityanskiy * 549633bf9885SMaxim Mikityanskiy * **-EACCES** if the SYN cookie is not valid. 549733bf9885SMaxim Mikityanskiy * 549833bf9885SMaxim Mikityanskiy * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5499c8996c98SJesper Dangaard Brouer * 5500c8996c98SJesper Dangaard Brouer * u64 bpf_ktime_get_tai_ns(void) 5501c8996c98SJesper Dangaard Brouer * Description 5502c8996c98SJesper Dangaard Brouer * A nonsettable system-wide clock derived from wall-clock time but 5503c8996c98SJesper Dangaard Brouer * ignoring leap seconds. This clock does not experience 5504c8996c98SJesper Dangaard Brouer * discontinuities and backwards jumps caused by NTP inserting leap 5505c8996c98SJesper Dangaard Brouer * seconds as CLOCK_REALTIME does. 5506c8996c98SJesper Dangaard Brouer * 5507c8996c98SJesper Dangaard Brouer * See: **clock_gettime**\ (**CLOCK_TAI**) 5508c8996c98SJesper Dangaard Brouer * Return 5509c8996c98SJesper Dangaard Brouer * Current *ktime*. 5510c8996c98SJesper Dangaard Brouer * 551120571567SDavid Vernet * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags) 551220571567SDavid Vernet * Description 551320571567SDavid Vernet * Drain samples from the specified user ring buffer, and invoke 551420571567SDavid Vernet * the provided callback for each such sample: 551520571567SDavid Vernet * 551627060531SKumar Kartikeya Dwivedi * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx); 551720571567SDavid Vernet * 551820571567SDavid Vernet * If **callback_fn** returns 0, the helper will continue to try 551920571567SDavid Vernet * and drain the next sample, up to a maximum of 552020571567SDavid Vernet * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1, 552120571567SDavid Vernet * the helper will skip the rest of the samples and return. Other 552220571567SDavid Vernet * return values are not used now, and will be rejected by the 552320571567SDavid Vernet * verifier. 552420571567SDavid Vernet * Return 552520571567SDavid Vernet * The number of drained samples if no error was encountered while 552620571567SDavid Vernet * draining samples, or 0 if no samples were present in the ring 552720571567SDavid Vernet * buffer. If a user-space producer was epoll-waiting on this map, 552820571567SDavid Vernet * and at least one sample was drained, they will receive an event 552920571567SDavid Vernet * notification notifying them of available space in the ring 553020571567SDavid Vernet * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this 553120571567SDavid Vernet * function, no wakeup notification will be sent. If the 553220571567SDavid Vernet * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will 553320571567SDavid Vernet * be sent even if no sample was drained. 553420571567SDavid Vernet * 553520571567SDavid Vernet * On failure, the returned value is one of the following: 553620571567SDavid Vernet * 553720571567SDavid Vernet * **-EBUSY** if the ring buffer is contended, and another calling 553820571567SDavid Vernet * context was concurrently draining the ring buffer. 553920571567SDavid Vernet * 554020571567SDavid Vernet * **-EINVAL** if user-space is not properly tracking the ring 554120571567SDavid Vernet * buffer due to the producer position not being aligned to 8 554220571567SDavid Vernet * bytes, a sample not being aligned to 8 bytes, or the producer 554320571567SDavid Vernet * position not matching the advertised length of a sample. 554420571567SDavid Vernet * 554520571567SDavid Vernet * **-E2BIG** if user-space has tried to publish a sample which is 554620571567SDavid Vernet * larger than the size of the ring buffer, or which cannot fit 554720571567SDavid Vernet * within a struct bpf_dynptr. 5548c4bcfb38SYonghong Song * 5549c4bcfb38SYonghong Song * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags) 5550c4bcfb38SYonghong Song * Description 5551c4bcfb38SYonghong Song * Get a bpf_local_storage from the *cgroup*. 5552c4bcfb38SYonghong Song * 5553c4bcfb38SYonghong Song * Logically, it could be thought of as getting the value from 5554c4bcfb38SYonghong Song * a *map* with *cgroup* as the **key**. From this 5555c4bcfb38SYonghong Song * perspective, the usage is not much different from 5556c4bcfb38SYonghong Song * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this 5557c4bcfb38SYonghong Song * helper enforces the key must be a cgroup struct and the map must also 5558c4bcfb38SYonghong Song * be a **BPF_MAP_TYPE_CGRP_STORAGE**. 5559c4bcfb38SYonghong Song * 5560c4bcfb38SYonghong Song * In reality, the local-storage value is embedded directly inside of the 5561c4bcfb38SYonghong Song * *cgroup* object itself, rather than being located in the 5562c4bcfb38SYonghong Song * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is 5563c4bcfb38SYonghong Song * queried for some *map* on a *cgroup* object, the kernel will perform an 5564c4bcfb38SYonghong Song * O(n) iteration over all of the live local-storage values for that 5565c4bcfb38SYonghong Song * *cgroup* object until the local-storage value for the *map* is found. 5566c4bcfb38SYonghong Song * 5567c4bcfb38SYonghong Song * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 5568c4bcfb38SYonghong Song * used such that a new bpf_local_storage will be 5569c4bcfb38SYonghong Song * created if one does not exist. *value* can be used 5570c4bcfb38SYonghong Song * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 5571c4bcfb38SYonghong Song * the initial value of a bpf_local_storage. If *value* is 5572c4bcfb38SYonghong Song * **NULL**, the new bpf_local_storage will be zero initialized. 5573c4bcfb38SYonghong Song * Return 5574c4bcfb38SYonghong Song * A bpf_local_storage pointer is returned on success. 5575c4bcfb38SYonghong Song * 5576c4bcfb38SYonghong Song * **NULL** if not found or there was an error in adding 5577c4bcfb38SYonghong Song * a new bpf_local_storage. 5578c4bcfb38SYonghong Song * 5579c4bcfb38SYonghong Song * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup) 5580c4bcfb38SYonghong Song * Description 5581c4bcfb38SYonghong Song * Delete a bpf_local_storage from a *cgroup*. 5582c4bcfb38SYonghong Song * Return 5583c4bcfb38SYonghong Song * 0 on success. 5584c4bcfb38SYonghong Song * 5585c4bcfb38SYonghong Song * **-ENOENT** if the bpf_local_storage cannot be found. 55867a4b28c6SDaniel Borkmann */ 55878a76145aSAndrii Nakryiko #define ___BPF_FUNC_MAPPER(FN, ctx...) \ 55888a76145aSAndrii Nakryiko FN(unspec, 0, ##ctx) \ 55898a76145aSAndrii Nakryiko FN(map_lookup_elem, 1, ##ctx) \ 55908a76145aSAndrii Nakryiko FN(map_update_elem, 2, ##ctx) \ 55918a76145aSAndrii Nakryiko FN(map_delete_elem, 3, ##ctx) \ 55928a76145aSAndrii Nakryiko FN(probe_read, 4, ##ctx) \ 55938a76145aSAndrii Nakryiko FN(ktime_get_ns, 5, ##ctx) \ 55948a76145aSAndrii Nakryiko FN(trace_printk, 6, ##ctx) \ 55958a76145aSAndrii Nakryiko FN(get_prandom_u32, 7, ##ctx) \ 55968a76145aSAndrii Nakryiko FN(get_smp_processor_id, 8, ##ctx) \ 55978a76145aSAndrii Nakryiko FN(skb_store_bytes, 9, ##ctx) \ 55988a76145aSAndrii Nakryiko FN(l3_csum_replace, 10, ##ctx) \ 55998a76145aSAndrii Nakryiko FN(l4_csum_replace, 11, ##ctx) \ 56008a76145aSAndrii Nakryiko FN(tail_call, 12, ##ctx) \ 56018a76145aSAndrii Nakryiko FN(clone_redirect, 13, ##ctx) \ 56028a76145aSAndrii Nakryiko FN(get_current_pid_tgid, 14, ##ctx) \ 56038a76145aSAndrii Nakryiko FN(get_current_uid_gid, 15, ##ctx) \ 56048a76145aSAndrii Nakryiko FN(get_current_comm, 16, ##ctx) \ 56058a76145aSAndrii Nakryiko FN(get_cgroup_classid, 17, ##ctx) \ 56068a76145aSAndrii Nakryiko FN(skb_vlan_push, 18, ##ctx) \ 56078a76145aSAndrii Nakryiko FN(skb_vlan_pop, 19, ##ctx) \ 56088a76145aSAndrii Nakryiko FN(skb_get_tunnel_key, 20, ##ctx) \ 56098a76145aSAndrii Nakryiko FN(skb_set_tunnel_key, 21, ##ctx) \ 56108a76145aSAndrii Nakryiko FN(perf_event_read, 22, ##ctx) \ 56118a76145aSAndrii Nakryiko FN(redirect, 23, ##ctx) \ 56128a76145aSAndrii Nakryiko FN(get_route_realm, 24, ##ctx) \ 56138a76145aSAndrii Nakryiko FN(perf_event_output, 25, ##ctx) \ 56148a76145aSAndrii Nakryiko FN(skb_load_bytes, 26, ##ctx) \ 56158a76145aSAndrii Nakryiko FN(get_stackid, 27, ##ctx) \ 56168a76145aSAndrii Nakryiko FN(csum_diff, 28, ##ctx) \ 56178a76145aSAndrii Nakryiko FN(skb_get_tunnel_opt, 29, ##ctx) \ 56188a76145aSAndrii Nakryiko FN(skb_set_tunnel_opt, 30, ##ctx) \ 56198a76145aSAndrii Nakryiko FN(skb_change_proto, 31, ##ctx) \ 56208a76145aSAndrii Nakryiko FN(skb_change_type, 32, ##ctx) \ 56218a76145aSAndrii Nakryiko FN(skb_under_cgroup, 33, ##ctx) \ 56228a76145aSAndrii Nakryiko FN(get_hash_recalc, 34, ##ctx) \ 56238a76145aSAndrii Nakryiko FN(get_current_task, 35, ##ctx) \ 56248a76145aSAndrii Nakryiko FN(probe_write_user, 36, ##ctx) \ 56258a76145aSAndrii Nakryiko FN(current_task_under_cgroup, 37, ##ctx) \ 56268a76145aSAndrii Nakryiko FN(skb_change_tail, 38, ##ctx) \ 56278a76145aSAndrii Nakryiko FN(skb_pull_data, 39, ##ctx) \ 56288a76145aSAndrii Nakryiko FN(csum_update, 40, ##ctx) \ 56298a76145aSAndrii Nakryiko FN(set_hash_invalid, 41, ##ctx) \ 56308a76145aSAndrii Nakryiko FN(get_numa_node_id, 42, ##ctx) \ 56318a76145aSAndrii Nakryiko FN(skb_change_head, 43, ##ctx) \ 56328a76145aSAndrii Nakryiko FN(xdp_adjust_head, 44, ##ctx) \ 56338a76145aSAndrii Nakryiko FN(probe_read_str, 45, ##ctx) \ 56348a76145aSAndrii Nakryiko FN(get_socket_cookie, 46, ##ctx) \ 56358a76145aSAndrii Nakryiko FN(get_socket_uid, 47, ##ctx) \ 56368a76145aSAndrii Nakryiko FN(set_hash, 48, ##ctx) \ 56378a76145aSAndrii Nakryiko FN(setsockopt, 49, ##ctx) \ 56388a76145aSAndrii Nakryiko FN(skb_adjust_room, 50, ##ctx) \ 56398a76145aSAndrii Nakryiko FN(redirect_map, 51, ##ctx) \ 56408a76145aSAndrii Nakryiko FN(sk_redirect_map, 52, ##ctx) \ 56418a76145aSAndrii Nakryiko FN(sock_map_update, 53, ##ctx) \ 56428a76145aSAndrii Nakryiko FN(xdp_adjust_meta, 54, ##ctx) \ 56438a76145aSAndrii Nakryiko FN(perf_event_read_value, 55, ##ctx) \ 56448a76145aSAndrii Nakryiko FN(perf_prog_read_value, 56, ##ctx) \ 56458a76145aSAndrii Nakryiko FN(getsockopt, 57, ##ctx) \ 56468a76145aSAndrii Nakryiko FN(override_return, 58, ##ctx) \ 56478a76145aSAndrii Nakryiko FN(sock_ops_cb_flags_set, 59, ##ctx) \ 56488a76145aSAndrii Nakryiko FN(msg_redirect_map, 60, ##ctx) \ 56498a76145aSAndrii Nakryiko FN(msg_apply_bytes, 61, ##ctx) \ 56508a76145aSAndrii Nakryiko FN(msg_cork_bytes, 62, ##ctx) \ 56518a76145aSAndrii Nakryiko FN(msg_pull_data, 63, ##ctx) \ 56528a76145aSAndrii Nakryiko FN(bind, 64, ##ctx) \ 56538a76145aSAndrii Nakryiko FN(xdp_adjust_tail, 65, ##ctx) \ 56548a76145aSAndrii Nakryiko FN(skb_get_xfrm_state, 66, ##ctx) \ 56558a76145aSAndrii Nakryiko FN(get_stack, 67, ##ctx) \ 56568a76145aSAndrii Nakryiko FN(skb_load_bytes_relative, 68, ##ctx) \ 56578a76145aSAndrii Nakryiko FN(fib_lookup, 69, ##ctx) \ 56588a76145aSAndrii Nakryiko FN(sock_hash_update, 70, ##ctx) \ 56598a76145aSAndrii Nakryiko FN(msg_redirect_hash, 71, ##ctx) \ 56608a76145aSAndrii Nakryiko FN(sk_redirect_hash, 72, ##ctx) \ 56618a76145aSAndrii Nakryiko FN(lwt_push_encap, 73, ##ctx) \ 56628a76145aSAndrii Nakryiko FN(lwt_seg6_store_bytes, 74, ##ctx) \ 56638a76145aSAndrii Nakryiko FN(lwt_seg6_adjust_srh, 75, ##ctx) \ 56648a76145aSAndrii Nakryiko FN(lwt_seg6_action, 76, ##ctx) \ 56658a76145aSAndrii Nakryiko FN(rc_repeat, 77, ##ctx) \ 56668a76145aSAndrii Nakryiko FN(rc_keydown, 78, ##ctx) \ 56678a76145aSAndrii Nakryiko FN(skb_cgroup_id, 79, ##ctx) \ 56688a76145aSAndrii Nakryiko FN(get_current_cgroup_id, 80, ##ctx) \ 56698a76145aSAndrii Nakryiko FN(get_local_storage, 81, ##ctx) \ 56708a76145aSAndrii Nakryiko FN(sk_select_reuseport, 82, ##ctx) \ 56718a76145aSAndrii Nakryiko FN(skb_ancestor_cgroup_id, 83, ##ctx) \ 56728a76145aSAndrii Nakryiko FN(sk_lookup_tcp, 84, ##ctx) \ 56738a76145aSAndrii Nakryiko FN(sk_lookup_udp, 85, ##ctx) \ 56748a76145aSAndrii Nakryiko FN(sk_release, 86, ##ctx) \ 56758a76145aSAndrii Nakryiko FN(map_push_elem, 87, ##ctx) \ 56768a76145aSAndrii Nakryiko FN(map_pop_elem, 88, ##ctx) \ 56778a76145aSAndrii Nakryiko FN(map_peek_elem, 89, ##ctx) \ 56788a76145aSAndrii Nakryiko FN(msg_push_data, 90, ##ctx) \ 56798a76145aSAndrii Nakryiko FN(msg_pop_data, 91, ##ctx) \ 56808a76145aSAndrii Nakryiko FN(rc_pointer_rel, 92, ##ctx) \ 56818a76145aSAndrii Nakryiko FN(spin_lock, 93, ##ctx) \ 56828a76145aSAndrii Nakryiko FN(spin_unlock, 94, ##ctx) \ 56838a76145aSAndrii Nakryiko FN(sk_fullsock, 95, ##ctx) \ 56848a76145aSAndrii Nakryiko FN(tcp_sock, 96, ##ctx) \ 56858a76145aSAndrii Nakryiko FN(skb_ecn_set_ce, 97, ##ctx) \ 56868a76145aSAndrii Nakryiko FN(get_listener_sock, 98, ##ctx) \ 56878a76145aSAndrii Nakryiko FN(skc_lookup_tcp, 99, ##ctx) \ 56888a76145aSAndrii Nakryiko FN(tcp_check_syncookie, 100, ##ctx) \ 56898a76145aSAndrii Nakryiko FN(sysctl_get_name, 101, ##ctx) \ 56908a76145aSAndrii Nakryiko FN(sysctl_get_current_value, 102, ##ctx) \ 56918a76145aSAndrii Nakryiko FN(sysctl_get_new_value, 103, ##ctx) \ 56928a76145aSAndrii Nakryiko FN(sysctl_set_new_value, 104, ##ctx) \ 56938a76145aSAndrii Nakryiko FN(strtol, 105, ##ctx) \ 56948a76145aSAndrii Nakryiko FN(strtoul, 106, ##ctx) \ 56958a76145aSAndrii Nakryiko FN(sk_storage_get, 107, ##ctx) \ 56968a76145aSAndrii Nakryiko FN(sk_storage_delete, 108, ##ctx) \ 56978a76145aSAndrii Nakryiko FN(send_signal, 109, ##ctx) \ 56988a76145aSAndrii Nakryiko FN(tcp_gen_syncookie, 110, ##ctx) \ 56998a76145aSAndrii Nakryiko FN(skb_output, 111, ##ctx) \ 57008a76145aSAndrii Nakryiko FN(probe_read_user, 112, ##ctx) \ 57018a76145aSAndrii Nakryiko FN(probe_read_kernel, 113, ##ctx) \ 57028a76145aSAndrii Nakryiko FN(probe_read_user_str, 114, ##ctx) \ 57038a76145aSAndrii Nakryiko FN(probe_read_kernel_str, 115, ##ctx) \ 57048a76145aSAndrii Nakryiko FN(tcp_send_ack, 116, ##ctx) \ 57058a76145aSAndrii Nakryiko FN(send_signal_thread, 117, ##ctx) \ 57068a76145aSAndrii Nakryiko FN(jiffies64, 118, ##ctx) \ 57078a76145aSAndrii Nakryiko FN(read_branch_records, 119, ##ctx) \ 57088a76145aSAndrii Nakryiko FN(get_ns_current_pid_tgid, 120, ##ctx) \ 57098a76145aSAndrii Nakryiko FN(xdp_output, 121, ##ctx) \ 57108a76145aSAndrii Nakryiko FN(get_netns_cookie, 122, ##ctx) \ 57118a76145aSAndrii Nakryiko FN(get_current_ancestor_cgroup_id, 123, ##ctx) \ 57128a76145aSAndrii Nakryiko FN(sk_assign, 124, ##ctx) \ 57138a76145aSAndrii Nakryiko FN(ktime_get_boot_ns, 125, ##ctx) \ 57148a76145aSAndrii Nakryiko FN(seq_printf, 126, ##ctx) \ 57158a76145aSAndrii Nakryiko FN(seq_write, 127, ##ctx) \ 57168a76145aSAndrii Nakryiko FN(sk_cgroup_id, 128, ##ctx) \ 57178a76145aSAndrii Nakryiko FN(sk_ancestor_cgroup_id, 129, ##ctx) \ 57188a76145aSAndrii Nakryiko FN(ringbuf_output, 130, ##ctx) \ 57198a76145aSAndrii Nakryiko FN(ringbuf_reserve, 131, ##ctx) \ 57208a76145aSAndrii Nakryiko FN(ringbuf_submit, 132, ##ctx) \ 57218a76145aSAndrii Nakryiko FN(ringbuf_discard, 133, ##ctx) \ 57228a76145aSAndrii Nakryiko FN(ringbuf_query, 134, ##ctx) \ 57238a76145aSAndrii Nakryiko FN(csum_level, 135, ##ctx) \ 57248a76145aSAndrii Nakryiko FN(skc_to_tcp6_sock, 136, ##ctx) \ 57258a76145aSAndrii Nakryiko FN(skc_to_tcp_sock, 137, ##ctx) \ 57268a76145aSAndrii Nakryiko FN(skc_to_tcp_timewait_sock, 138, ##ctx) \ 57278a76145aSAndrii Nakryiko FN(skc_to_tcp_request_sock, 139, ##ctx) \ 57288a76145aSAndrii Nakryiko FN(skc_to_udp6_sock, 140, ##ctx) \ 57298a76145aSAndrii Nakryiko FN(get_task_stack, 141, ##ctx) \ 57308a76145aSAndrii Nakryiko FN(load_hdr_opt, 142, ##ctx) \ 57318a76145aSAndrii Nakryiko FN(store_hdr_opt, 143, ##ctx) \ 57328a76145aSAndrii Nakryiko FN(reserve_hdr_opt, 144, ##ctx) \ 57338a76145aSAndrii Nakryiko FN(inode_storage_get, 145, ##ctx) \ 57348a76145aSAndrii Nakryiko FN(inode_storage_delete, 146, ##ctx) \ 57358a76145aSAndrii Nakryiko FN(d_path, 147, ##ctx) \ 57368a76145aSAndrii Nakryiko FN(copy_from_user, 148, ##ctx) \ 57378a76145aSAndrii Nakryiko FN(snprintf_btf, 149, ##ctx) \ 57388a76145aSAndrii Nakryiko FN(seq_printf_btf, 150, ##ctx) \ 57398a76145aSAndrii Nakryiko FN(skb_cgroup_classid, 151, ##ctx) \ 57408a76145aSAndrii Nakryiko FN(redirect_neigh, 152, ##ctx) \ 57418a76145aSAndrii Nakryiko FN(per_cpu_ptr, 153, ##ctx) \ 57428a76145aSAndrii Nakryiko FN(this_cpu_ptr, 154, ##ctx) \ 57438a76145aSAndrii Nakryiko FN(redirect_peer, 155, ##ctx) \ 57448a76145aSAndrii Nakryiko FN(task_storage_get, 156, ##ctx) \ 57458a76145aSAndrii Nakryiko FN(task_storage_delete, 157, ##ctx) \ 57468a76145aSAndrii Nakryiko FN(get_current_task_btf, 158, ##ctx) \ 57478a76145aSAndrii Nakryiko FN(bprm_opts_set, 159, ##ctx) \ 57488a76145aSAndrii Nakryiko FN(ktime_get_coarse_ns, 160, ##ctx) \ 57498a76145aSAndrii Nakryiko FN(ima_inode_hash, 161, ##ctx) \ 57508a76145aSAndrii Nakryiko FN(sock_from_file, 162, ##ctx) \ 57518a76145aSAndrii Nakryiko FN(check_mtu, 163, ##ctx) \ 57528a76145aSAndrii Nakryiko FN(for_each_map_elem, 164, ##ctx) \ 57538a76145aSAndrii Nakryiko FN(snprintf, 165, ##ctx) \ 57548a76145aSAndrii Nakryiko FN(sys_bpf, 166, ##ctx) \ 57558a76145aSAndrii Nakryiko FN(btf_find_by_name_kind, 167, ##ctx) \ 57568a76145aSAndrii Nakryiko FN(sys_close, 168, ##ctx) \ 57578a76145aSAndrii Nakryiko FN(timer_init, 169, ##ctx) \ 57588a76145aSAndrii Nakryiko FN(timer_set_callback, 170, ##ctx) \ 57598a76145aSAndrii Nakryiko FN(timer_start, 171, ##ctx) \ 57608a76145aSAndrii Nakryiko FN(timer_cancel, 172, ##ctx) \ 57618a76145aSAndrii Nakryiko FN(get_func_ip, 173, ##ctx) \ 57628a76145aSAndrii Nakryiko FN(get_attach_cookie, 174, ##ctx) \ 57638a76145aSAndrii Nakryiko FN(task_pt_regs, 175, ##ctx) \ 57648a76145aSAndrii Nakryiko FN(get_branch_snapshot, 176, ##ctx) \ 57658a76145aSAndrii Nakryiko FN(trace_vprintk, 177, ##ctx) \ 57668a76145aSAndrii Nakryiko FN(skc_to_unix_sock, 178, ##ctx) \ 57678a76145aSAndrii Nakryiko FN(kallsyms_lookup_name, 179, ##ctx) \ 57688a76145aSAndrii Nakryiko FN(find_vma, 180, ##ctx) \ 57698a76145aSAndrii Nakryiko FN(loop, 181, ##ctx) \ 57708a76145aSAndrii Nakryiko FN(strncmp, 182, ##ctx) \ 57718a76145aSAndrii Nakryiko FN(get_func_arg, 183, ##ctx) \ 57728a76145aSAndrii Nakryiko FN(get_func_ret, 184, ##ctx) \ 57738a76145aSAndrii Nakryiko FN(get_func_arg_cnt, 185, ##ctx) \ 57748a76145aSAndrii Nakryiko FN(get_retval, 186, ##ctx) \ 57758a76145aSAndrii Nakryiko FN(set_retval, 187, ##ctx) \ 57768a76145aSAndrii Nakryiko FN(xdp_get_buff_len, 188, ##ctx) \ 57778a76145aSAndrii Nakryiko FN(xdp_load_bytes, 189, ##ctx) \ 57788a76145aSAndrii Nakryiko FN(xdp_store_bytes, 190, ##ctx) \ 57798a76145aSAndrii Nakryiko FN(copy_from_user_task, 191, ##ctx) \ 57808a76145aSAndrii Nakryiko FN(skb_set_tstamp, 192, ##ctx) \ 57818a76145aSAndrii Nakryiko FN(ima_file_hash, 193, ##ctx) \ 57828a76145aSAndrii Nakryiko FN(kptr_xchg, 194, ##ctx) \ 57838a76145aSAndrii Nakryiko FN(map_lookup_percpu_elem, 195, ##ctx) \ 57848a76145aSAndrii Nakryiko FN(skc_to_mptcp_sock, 196, ##ctx) \ 57858a76145aSAndrii Nakryiko FN(dynptr_from_mem, 197, ##ctx) \ 57868a76145aSAndrii Nakryiko FN(ringbuf_reserve_dynptr, 198, ##ctx) \ 57878a76145aSAndrii Nakryiko FN(ringbuf_submit_dynptr, 199, ##ctx) \ 57888a76145aSAndrii Nakryiko FN(ringbuf_discard_dynptr, 200, ##ctx) \ 57898a76145aSAndrii Nakryiko FN(dynptr_read, 201, ##ctx) \ 57908a76145aSAndrii Nakryiko FN(dynptr_write, 202, ##ctx) \ 57918a76145aSAndrii Nakryiko FN(dynptr_data, 203, ##ctx) \ 57928a76145aSAndrii Nakryiko FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \ 57938a76145aSAndrii Nakryiko FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \ 57948a76145aSAndrii Nakryiko FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \ 57958a76145aSAndrii Nakryiko FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \ 57968a76145aSAndrii Nakryiko FN(ktime_get_tai_ns, 208, ##ctx) \ 57978a76145aSAndrii Nakryiko FN(user_ringbuf_drain, 209, ##ctx) \ 5798c4bcfb38SYonghong Song FN(cgrp_storage_get, 210, ##ctx) \ 5799c4bcfb38SYonghong Song FN(cgrp_storage_delete, 211, ##ctx) \ 5800fa28dcb8SSong Liu /* */ 58017a4b28c6SDaniel Borkmann 58028a76145aSAndrii Nakryiko /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't 58038a76145aSAndrii Nakryiko * know or care about integer value that is now passed as second argument 58048a76145aSAndrii Nakryiko */ 58058a76145aSAndrii Nakryiko #define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name), 58068a76145aSAndrii Nakryiko #define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN) 58078a76145aSAndrii Nakryiko 5808ebb676daSThomas Graf /* integer value in 'imm' field of BPF_CALL instruction selects which helper 5809ebb676daSThomas Graf * function eBPF program intends to call 58102d0e30c3SDaniel Borkmann */ 58118a76145aSAndrii Nakryiko #define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y, 5812ebb676daSThomas Graf enum bpf_func_id { 58138a76145aSAndrii Nakryiko ___BPF_FUNC_MAPPER(__BPF_ENUM_FN) 581409756af4SAlexei Starovoitov __BPF_FUNC_MAX_ID, 581509756af4SAlexei Starovoitov }; 5816ebb676daSThomas Graf #undef __BPF_ENUM_FN 581709756af4SAlexei Starovoitov 5818781c53bcSDaniel Borkmann /* All flags used by eBPF helper functions, placed here. */ 5819781c53bcSDaniel Borkmann 5820781c53bcSDaniel Borkmann /* BPF_FUNC_skb_store_bytes flags. */ 58211aae4bddSAndrii Nakryiko enum { 58221aae4bddSAndrii Nakryiko BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 58231aae4bddSAndrii Nakryiko BPF_F_INVALIDATE_HASH = (1ULL << 1), 58241aae4bddSAndrii Nakryiko }; 5825781c53bcSDaniel Borkmann 5826781c53bcSDaniel Borkmann /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 5827781c53bcSDaniel Borkmann * First 4 bits are for passing the header field size. 5828781c53bcSDaniel Borkmann */ 58291aae4bddSAndrii Nakryiko enum { 58301aae4bddSAndrii Nakryiko BPF_F_HDR_FIELD_MASK = 0xfULL, 58311aae4bddSAndrii Nakryiko }; 5832781c53bcSDaniel Borkmann 5833781c53bcSDaniel Borkmann /* BPF_FUNC_l4_csum_replace flags. */ 58341aae4bddSAndrii Nakryiko enum { 58351aae4bddSAndrii Nakryiko BPF_F_PSEUDO_HDR = (1ULL << 4), 58361aae4bddSAndrii Nakryiko BPF_F_MARK_MANGLED_0 = (1ULL << 5), 58371aae4bddSAndrii Nakryiko BPF_F_MARK_ENFORCE = (1ULL << 6), 58381aae4bddSAndrii Nakryiko }; 5839781c53bcSDaniel Borkmann 5840781c53bcSDaniel Borkmann /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 58411aae4bddSAndrii Nakryiko enum { 58421aae4bddSAndrii Nakryiko BPF_F_INGRESS = (1ULL << 0), 58431aae4bddSAndrii Nakryiko }; 5844781c53bcSDaniel Borkmann 5845c6c33454SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 58461aae4bddSAndrii Nakryiko enum { 58471aae4bddSAndrii Nakryiko BPF_F_TUNINFO_IPV6 = (1ULL << 0), 58481aae4bddSAndrii Nakryiko }; 5849c6c33454SDaniel Borkmann 5850c195651eSYonghong Song /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 58511aae4bddSAndrii Nakryiko enum { 58521aae4bddSAndrii Nakryiko BPF_F_SKIP_FIELD_MASK = 0xffULL, 58531aae4bddSAndrii Nakryiko BPF_F_USER_STACK = (1ULL << 8), 5854c195651eSYonghong Song /* flags used by BPF_FUNC_get_stackid only. */ 58551aae4bddSAndrii Nakryiko BPF_F_FAST_STACK_CMP = (1ULL << 9), 58561aae4bddSAndrii Nakryiko BPF_F_REUSE_STACKID = (1ULL << 10), 5857c195651eSYonghong Song /* flags used by BPF_FUNC_get_stack only. */ 58581aae4bddSAndrii Nakryiko BPF_F_USER_BUILD_ID = (1ULL << 11), 58591aae4bddSAndrii Nakryiko }; 5860d5a3b1f6SAlexei Starovoitov 58612da897e5SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key flags. */ 58621aae4bddSAndrii Nakryiko enum { 58631aae4bddSAndrii Nakryiko BPF_F_ZERO_CSUM_TX = (1ULL << 1), 58641aae4bddSAndrii Nakryiko BPF_F_DONT_FRAGMENT = (1ULL << 2), 58651aae4bddSAndrii Nakryiko BPF_F_SEQ_NUMBER = (1ULL << 3), 5866e26aa600SChristian Ehrig BPF_F_NO_TUNNEL_KEY = (1ULL << 4), 58671aae4bddSAndrii Nakryiko }; 58682da897e5SDaniel Borkmann 586944c51472SShmulik Ladkani /* BPF_FUNC_skb_get_tunnel_key flags. */ 587044c51472SShmulik Ladkani enum { 587144c51472SShmulik Ladkani BPF_F_TUNINFO_FLAGS = (1ULL << 4), 587244c51472SShmulik Ladkani }; 587344c51472SShmulik Ladkani 5874908432caSYonghong Song /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 5875908432caSYonghong Song * BPF_FUNC_perf_event_read_value flags. 5876908432caSYonghong Song */ 58771aae4bddSAndrii Nakryiko enum { 58781aae4bddSAndrii Nakryiko BPF_F_INDEX_MASK = 0xffffffffULL, 58791aae4bddSAndrii Nakryiko BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 5880555c8a86SDaniel Borkmann /* BPF_FUNC_perf_event_output for sk_buff input context. */ 58811aae4bddSAndrii Nakryiko BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 58821aae4bddSAndrii Nakryiko }; 58831e33759cSDaniel Borkmann 5884f71c6143SJoe Stringer /* Current network namespace */ 58851aae4bddSAndrii Nakryiko enum { 58861aae4bddSAndrii Nakryiko BPF_F_CURRENT_NETNS = (-1L), 58871aae4bddSAndrii Nakryiko }; 5888f71c6143SJoe Stringer 58897cdec54fSDaniel Borkmann /* BPF_FUNC_csum_level level values. */ 58907cdec54fSDaniel Borkmann enum { 58917cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_QUERY, 58927cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_INC, 58937cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_DEC, 58947cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_RESET, 58957cdec54fSDaniel Borkmann }; 58967cdec54fSDaniel Borkmann 58972278f6ccSWillem de Bruijn /* BPF_FUNC_skb_adjust_room flags. */ 58981aae4bddSAndrii Nakryiko enum { 58991aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 59001aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 59011aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 59021aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 59031aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 5904836e66c2SDaniel Borkmann BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 5905d01b59c9SXuesen Huang BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 5906d219df60SZiyang Xuan BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7), 5907d219df60SZiyang Xuan BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8), 59081aae4bddSAndrii Nakryiko }; 59092278f6ccSWillem de Bruijn 59101aae4bddSAndrii Nakryiko enum { 59111aae4bddSAndrii Nakryiko BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 59121aae4bddSAndrii Nakryiko BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 59131aae4bddSAndrii Nakryiko }; 591458dfc900SAlan Maguire 591558dfc900SAlan Maguire #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 591658dfc900SAlan Maguire BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 591758dfc900SAlan Maguire << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 5918868d5235SWillem de Bruijn 5919808649fbSAndrey Ignatov /* BPF_FUNC_sysctl_get_name flags. */ 59201aae4bddSAndrii Nakryiko enum { 59211aae4bddSAndrii Nakryiko BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 59221aae4bddSAndrii Nakryiko }; 5923808649fbSAndrey Ignatov 5924f836a56eSKP Singh /* BPF_FUNC_<kernel_obj>_storage_get flags */ 59251aae4bddSAndrii Nakryiko enum { 5926f836a56eSKP Singh BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 5927f836a56eSKP Singh /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 5928f836a56eSKP Singh * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 5929f836a56eSKP Singh */ 5930f836a56eSKP Singh BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 59311aae4bddSAndrii Nakryiko }; 59326ac99e8fSMartin KaFai Lau 5933fff7b643SDaniel Xu /* BPF_FUNC_read_branch_records flags. */ 59341aae4bddSAndrii Nakryiko enum { 59351aae4bddSAndrii Nakryiko BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 59361aae4bddSAndrii Nakryiko }; 5937fff7b643SDaniel Xu 5938457f4436SAndrii Nakryiko /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 5939457f4436SAndrii Nakryiko * BPF_FUNC_bpf_ringbuf_output flags. 5940457f4436SAndrii Nakryiko */ 5941457f4436SAndrii Nakryiko enum { 5942457f4436SAndrii Nakryiko BPF_RB_NO_WAKEUP = (1ULL << 0), 5943457f4436SAndrii Nakryiko BPF_RB_FORCE_WAKEUP = (1ULL << 1), 5944457f4436SAndrii Nakryiko }; 5945457f4436SAndrii Nakryiko 5946457f4436SAndrii Nakryiko /* BPF_FUNC_bpf_ringbuf_query flags */ 5947457f4436SAndrii Nakryiko enum { 5948457f4436SAndrii Nakryiko BPF_RB_AVAIL_DATA = 0, 5949457f4436SAndrii Nakryiko BPF_RB_RING_SIZE = 1, 5950457f4436SAndrii Nakryiko BPF_RB_CONS_POS = 2, 5951457f4436SAndrii Nakryiko BPF_RB_PROD_POS = 3, 5952457f4436SAndrii Nakryiko }; 5953457f4436SAndrii Nakryiko 5954457f4436SAndrii Nakryiko /* BPF ring buffer constants */ 5955457f4436SAndrii Nakryiko enum { 5956457f4436SAndrii Nakryiko BPF_RINGBUF_BUSY_BIT = (1U << 31), 5957457f4436SAndrii Nakryiko BPF_RINGBUF_DISCARD_BIT = (1U << 30), 5958457f4436SAndrii Nakryiko BPF_RINGBUF_HDR_SZ = 8, 5959457f4436SAndrii Nakryiko }; 5960457f4436SAndrii Nakryiko 5961e9ddbb77SJakub Sitnicki /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 5962e9ddbb77SJakub Sitnicki enum { 5963e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 5964e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 5965e9ddbb77SJakub Sitnicki }; 5966e9ddbb77SJakub Sitnicki 59672be7e212SDaniel Borkmann /* Mode for BPF_FUNC_skb_adjust_room helper. */ 59682be7e212SDaniel Borkmann enum bpf_adj_room_mode { 59692be7e212SDaniel Borkmann BPF_ADJ_ROOM_NET, 597014aa3192SWillem de Bruijn BPF_ADJ_ROOM_MAC, 59712be7e212SDaniel Borkmann }; 59722be7e212SDaniel Borkmann 59734e1ec56cSDaniel Borkmann /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 59744e1ec56cSDaniel Borkmann enum bpf_hdr_start_off { 59754e1ec56cSDaniel Borkmann BPF_HDR_START_MAC, 59764e1ec56cSDaniel Borkmann BPF_HDR_START_NET, 59774e1ec56cSDaniel Borkmann }; 59784e1ec56cSDaniel Borkmann 5979fe94cc29SMathieu Xhonneux /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 5980fe94cc29SMathieu Xhonneux enum bpf_lwt_encap_mode { 5981fe94cc29SMathieu Xhonneux BPF_LWT_ENCAP_SEG6, 59823e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_SEG6_INLINE, 59833e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_IP, 5984fe94cc29SMathieu Xhonneux }; 5985fe94cc29SMathieu Xhonneux 59863f6719c7SKP Singh /* Flags for bpf_bprm_opts_set helper */ 59873f6719c7SKP Singh enum { 59883f6719c7SKP Singh BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 59893f6719c7SKP Singh }; 59903f6719c7SKP Singh 5991e624d4edSHangbin Liu /* Flags for bpf_redirect_map helper */ 5992e624d4edSHangbin Liu enum { 5993e624d4edSHangbin Liu BPF_F_BROADCAST = (1ULL << 3), 5994e624d4edSHangbin Liu BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 5995e624d4edSHangbin Liu }; 5996e624d4edSHangbin Liu 5997b7df9adaSDaniel Borkmann #define __bpf_md_ptr(type, name) \ 5998b7df9adaSDaniel Borkmann union { \ 5999b7df9adaSDaniel Borkmann type name; \ 6000b7df9adaSDaniel Borkmann __u64 :64; \ 6001b7df9adaSDaniel Borkmann } __attribute__((aligned(8))) 6002b7df9adaSDaniel Borkmann 60038d21ec0eSMartin KaFai Lau enum { 60049bb984f2SMartin KaFai Lau BPF_SKB_TSTAMP_UNSPEC, 60059bb984f2SMartin KaFai Lau BPF_SKB_TSTAMP_DELIVERY_MONO, /* tstamp has mono delivery time */ 60069bb984f2SMartin KaFai Lau /* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle, 60079bb984f2SMartin KaFai Lau * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC 60089bb984f2SMartin KaFai Lau * and try to deduce it by ingress, egress or skb->sk->sk_clockid. 60099bb984f2SMartin KaFai Lau */ 60108d21ec0eSMartin KaFai Lau }; 60118d21ec0eSMartin KaFai Lau 60129bac3d6dSAlexei Starovoitov /* user accessible mirror of in-kernel sk_buff. 60139bac3d6dSAlexei Starovoitov * new fields can only be added to the end of this structure 60149bac3d6dSAlexei Starovoitov */ 60159bac3d6dSAlexei Starovoitov struct __sk_buff { 60169bac3d6dSAlexei Starovoitov __u32 len; 60179bac3d6dSAlexei Starovoitov __u32 pkt_type; 60189bac3d6dSAlexei Starovoitov __u32 mark; 60199bac3d6dSAlexei Starovoitov __u32 queue_mapping; 6020c2497395SAlexei Starovoitov __u32 protocol; 6021c2497395SAlexei Starovoitov __u32 vlan_present; 6022c2497395SAlexei Starovoitov __u32 vlan_tci; 602327cd5452SMichal Sekletar __u32 vlan_proto; 6024bcad5718SDaniel Borkmann __u32 priority; 602537e82c2fSAlexei Starovoitov __u32 ingress_ifindex; 602637e82c2fSAlexei Starovoitov __u32 ifindex; 6027d691f9e8SAlexei Starovoitov __u32 tc_index; 6028d691f9e8SAlexei Starovoitov __u32 cb[5]; 6029ba7591d8SDaniel Borkmann __u32 hash; 6030045efa82SDaniel Borkmann __u32 tc_classid; 6031969bf05eSAlexei Starovoitov __u32 data; 6032969bf05eSAlexei Starovoitov __u32 data_end; 6033b1d9fc41SDaniel Borkmann __u32 napi_id; 60348a31db56SJohn Fastabend 6035de8f3a83SDaniel Borkmann /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 60368a31db56SJohn Fastabend __u32 family; 60378a31db56SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 60388a31db56SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 60398a31db56SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 60408a31db56SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 60418a31db56SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 60428a31db56SJohn Fastabend __u32 local_port; /* stored in host byte order */ 6043de8f3a83SDaniel Borkmann /* ... here. */ 6044de8f3a83SDaniel Borkmann 6045de8f3a83SDaniel Borkmann __u32 data_meta; 6046b7df9adaSDaniel Borkmann __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 6047f11216b2SVlad Dumitrescu __u64 tstamp; 6048e3da08d0SPetar Penkov __u32 wire_len; 6049d9ff286aSEric Dumazet __u32 gso_segs; 605046f8bc92SMartin KaFai Lau __bpf_md_ptr(struct bpf_sock *, sk); 6051cf62089bSWillem de Bruijn __u32 gso_size; 60529bb984f2SMartin KaFai Lau __u8 tstamp_type; 60538d21ec0eSMartin KaFai Lau __u32 :24; /* Padding, future use. */ 6054f64c4aceSVadim Fedorenko __u64 hwtstamp; 60559bac3d6dSAlexei Starovoitov }; 60569bac3d6dSAlexei Starovoitov 6057d3aa45ceSAlexei Starovoitov struct bpf_tunnel_key { 6058d3aa45ceSAlexei Starovoitov __u32 tunnel_id; 6059c6c33454SDaniel Borkmann union { 6060d3aa45ceSAlexei Starovoitov __u32 remote_ipv4; 6061c6c33454SDaniel Borkmann __u32 remote_ipv6[4]; 6062c6c33454SDaniel Borkmann }; 6063c6c33454SDaniel Borkmann __u8 tunnel_tos; 6064c6c33454SDaniel Borkmann __u8 tunnel_ttl; 606544c51472SShmulik Ladkani union { 606644c51472SShmulik Ladkani __u16 tunnel_ext; /* compat */ 606744c51472SShmulik Ladkani __be16 tunnel_flags; 606844c51472SShmulik Ladkani }; 60694018ab18SDaniel Borkmann __u32 tunnel_label; 607026101f5aSKaixi Fan union { 607126101f5aSKaixi Fan __u32 local_ipv4; 607226101f5aSKaixi Fan __u32 local_ipv6[4]; 607326101f5aSKaixi Fan }; 6074d3aa45ceSAlexei Starovoitov }; 6075d3aa45ceSAlexei Starovoitov 607612bed760SEyal Birger /* user accessible mirror of in-kernel xfrm_state. 607712bed760SEyal Birger * new fields can only be added to the end of this structure 607812bed760SEyal Birger */ 607912bed760SEyal Birger struct bpf_xfrm_state { 608012bed760SEyal Birger __u32 reqid; 608112bed760SEyal Birger __u32 spi; /* Stored in network byte order */ 608212bed760SEyal Birger __u16 family; 60831fbc2e0cSDaniel Borkmann __u16 ext; /* Padding, future use. */ 608412bed760SEyal Birger union { 608512bed760SEyal Birger __u32 remote_ipv4; /* Stored in network byte order */ 608612bed760SEyal Birger __u32 remote_ipv6[4]; /* Stored in network byte order */ 608712bed760SEyal Birger }; 608812bed760SEyal Birger }; 608912bed760SEyal Birger 60903a0af8fdSThomas Graf /* Generic BPF return codes which all BPF program types may support. 60913a0af8fdSThomas Graf * The values are binary compatible with their TC_ACT_* counter-part to 60923a0af8fdSThomas Graf * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 60933a0af8fdSThomas Graf * programs. 60943a0af8fdSThomas Graf * 60953a0af8fdSThomas Graf * XDP is handled seprately, see XDP_*. 60963a0af8fdSThomas Graf */ 60973a0af8fdSThomas Graf enum bpf_ret_code { 60983a0af8fdSThomas Graf BPF_OK = 0, 60993a0af8fdSThomas Graf /* 1 reserved */ 61003a0af8fdSThomas Graf BPF_DROP = 2, 61013a0af8fdSThomas Graf /* 3-6 reserved */ 61023a0af8fdSThomas Graf BPF_REDIRECT = 7, 61033e0bd37cSPeter Oskolkov /* >127 are reserved for prog type specific return codes. 61043e0bd37cSPeter Oskolkov * 61053e0bd37cSPeter Oskolkov * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 61063e0bd37cSPeter Oskolkov * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 61073e0bd37cSPeter Oskolkov * changed and should be routed based on its new L3 header. 61083e0bd37cSPeter Oskolkov * (This is an L3 redirect, as opposed to L2 redirect 61093e0bd37cSPeter Oskolkov * represented by BPF_REDIRECT above). 61103e0bd37cSPeter Oskolkov */ 61113e0bd37cSPeter Oskolkov BPF_LWT_REROUTE = 128, 611291350fe1SShmulik Ladkani /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR 611391350fe1SShmulik Ladkani * to indicate that no custom dissection was performed, and 611491350fe1SShmulik Ladkani * fallback to standard dissector is requested. 611591350fe1SShmulik Ladkani */ 611691350fe1SShmulik Ladkani BPF_FLOW_DISSECTOR_CONTINUE = 129, 61173a0af8fdSThomas Graf }; 61183a0af8fdSThomas Graf 611961023658SDavid Ahern struct bpf_sock { 612061023658SDavid Ahern __u32 bound_dev_if; 6121aa4c1037SDavid Ahern __u32 family; 6122aa4c1037SDavid Ahern __u32 type; 6123aa4c1037SDavid Ahern __u32 protocol; 6124482dca93SDavid Ahern __u32 mark; 6125482dca93SDavid Ahern __u32 priority; 6126aa65d696SMartin KaFai Lau /* IP address also allows 1 and 2 bytes access */ 6127aa65d696SMartin KaFai Lau __u32 src_ip4; 6128aa65d696SMartin KaFai Lau __u32 src_ip6[4]; 6129aa65d696SMartin KaFai Lau __u32 src_port; /* host byte order */ 61304421a582SJakub Sitnicki __be16 dst_port; /* network byte order */ 61314421a582SJakub Sitnicki __u16 :16; /* zero padding */ 6132aa65d696SMartin KaFai Lau __u32 dst_ip4; 6133aa65d696SMartin KaFai Lau __u32 dst_ip6[4]; 6134aa65d696SMartin KaFai Lau __u32 state; 6135c3c16f2eSAmritha Nambiar __s32 rx_queue_mapping; 613661023658SDavid Ahern }; 613761023658SDavid Ahern 6138655a51e5SMartin KaFai Lau struct bpf_tcp_sock { 6139655a51e5SMartin KaFai Lau __u32 snd_cwnd; /* Sending congestion window */ 6140655a51e5SMartin KaFai Lau __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 6141655a51e5SMartin KaFai Lau __u32 rtt_min; 6142655a51e5SMartin KaFai Lau __u32 snd_ssthresh; /* Slow start size threshold */ 6143655a51e5SMartin KaFai Lau __u32 rcv_nxt; /* What we want to receive next */ 6144655a51e5SMartin KaFai Lau __u32 snd_nxt; /* Next sequence we send */ 6145655a51e5SMartin KaFai Lau __u32 snd_una; /* First byte we want an ack for */ 6146655a51e5SMartin KaFai Lau __u32 mss_cache; /* Cached effective mss, not including SACKS */ 6147655a51e5SMartin KaFai Lau __u32 ecn_flags; /* ECN status bits. */ 6148655a51e5SMartin KaFai Lau __u32 rate_delivered; /* saved rate sample: packets delivered */ 6149655a51e5SMartin KaFai Lau __u32 rate_interval_us; /* saved rate sample: time elapsed */ 6150655a51e5SMartin KaFai Lau __u32 packets_out; /* Packets which are "in flight" */ 6151655a51e5SMartin KaFai Lau __u32 retrans_out; /* Retransmitted packets out */ 6152655a51e5SMartin KaFai Lau __u32 total_retrans; /* Total retransmits for entire connection */ 6153655a51e5SMartin KaFai Lau __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 6154655a51e5SMartin KaFai Lau * total number of segments in. 6155655a51e5SMartin KaFai Lau */ 6156655a51e5SMartin KaFai Lau __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 6157655a51e5SMartin KaFai Lau * total number of data segments in. 6158655a51e5SMartin KaFai Lau */ 6159655a51e5SMartin KaFai Lau __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 6160655a51e5SMartin KaFai Lau * The total number of segments sent. 6161655a51e5SMartin KaFai Lau */ 6162655a51e5SMartin KaFai Lau __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 6163655a51e5SMartin KaFai Lau * total number of data segments sent. 6164655a51e5SMartin KaFai Lau */ 6165655a51e5SMartin KaFai Lau __u32 lost_out; /* Lost packets */ 6166655a51e5SMartin KaFai Lau __u32 sacked_out; /* SACK'd packets */ 6167655a51e5SMartin KaFai Lau __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 6168655a51e5SMartin KaFai Lau * sum(delta(rcv_nxt)), or how many bytes 6169655a51e5SMartin KaFai Lau * were acked. 6170655a51e5SMartin KaFai Lau */ 6171655a51e5SMartin KaFai Lau __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 6172655a51e5SMartin KaFai Lau * sum(delta(snd_una)), or how many bytes 6173655a51e5SMartin KaFai Lau * were acked. 6174655a51e5SMartin KaFai Lau */ 61750357746dSStanislav Fomichev __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 61760357746dSStanislav Fomichev * total number of DSACK blocks received 61770357746dSStanislav Fomichev */ 61780357746dSStanislav Fomichev __u32 delivered; /* Total data packets delivered incl. rexmits */ 61790357746dSStanislav Fomichev __u32 delivered_ce; /* Like the above but only ECE marked packets */ 6180c2cb5e82SStanislav Fomichev __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 6181655a51e5SMartin KaFai Lau }; 6182655a51e5SMartin KaFai Lau 61836acc9b43SJoe Stringer struct bpf_sock_tuple { 61846acc9b43SJoe Stringer union { 61856acc9b43SJoe Stringer struct { 61866acc9b43SJoe Stringer __be32 saddr; 61876acc9b43SJoe Stringer __be32 daddr; 61886acc9b43SJoe Stringer __be16 sport; 61896acc9b43SJoe Stringer __be16 dport; 61906acc9b43SJoe Stringer } ipv4; 61916acc9b43SJoe Stringer struct { 61926acc9b43SJoe Stringer __be32 saddr[4]; 61936acc9b43SJoe Stringer __be32 daddr[4]; 61946acc9b43SJoe Stringer __be16 sport; 61956acc9b43SJoe Stringer __be16 dport; 61966acc9b43SJoe Stringer } ipv6; 61976acc9b43SJoe Stringer }; 61986acc9b43SJoe Stringer }; 61996acc9b43SJoe Stringer 6200fada7fdcSJonathan Lemon struct bpf_xdp_sock { 6201fada7fdcSJonathan Lemon __u32 queue_id; 6202fada7fdcSJonathan Lemon }; 6203fada7fdcSJonathan Lemon 620417bedab2SMartin KaFai Lau #define XDP_PACKET_HEADROOM 256 620517bedab2SMartin KaFai Lau 62066a773a15SBrenden Blanco /* User return codes for XDP prog type. 62076a773a15SBrenden Blanco * A valid XDP program must return one of these defined values. All other 62089beb8bedSDaniel Borkmann * return codes are reserved for future use. Unknown return codes will 62099beb8bedSDaniel Borkmann * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 62106a773a15SBrenden Blanco */ 62116a773a15SBrenden Blanco enum xdp_action { 62126a773a15SBrenden Blanco XDP_ABORTED = 0, 62136a773a15SBrenden Blanco XDP_DROP, 62146a773a15SBrenden Blanco XDP_PASS, 62156ce96ca3SBrenden Blanco XDP_TX, 6216814abfabSJohn Fastabend XDP_REDIRECT, 62176a773a15SBrenden Blanco }; 62186a773a15SBrenden Blanco 62196a773a15SBrenden Blanco /* user accessible metadata for XDP packet hook 62206a773a15SBrenden Blanco * new fields must be added to the end of this structure 62216a773a15SBrenden Blanco */ 62226a773a15SBrenden Blanco struct xdp_md { 62236a773a15SBrenden Blanco __u32 data; 62246a773a15SBrenden Blanco __u32 data_end; 6225de8f3a83SDaniel Borkmann __u32 data_meta; 6226daaf24c6SJesper Dangaard Brouer /* Below access go through struct xdp_rxq_info */ 622702dd3291SJesper Dangaard Brouer __u32 ingress_ifindex; /* rxq->dev->ifindex */ 622802dd3291SJesper Dangaard Brouer __u32 rx_queue_index; /* rxq->queue_index */ 622964b59025SDavid Ahern 623064b59025SDavid Ahern __u32 egress_ifindex; /* txq->dev->ifindex */ 62316a773a15SBrenden Blanco }; 62326a773a15SBrenden Blanco 6233281920b7SJesper Dangaard Brouer /* DEVMAP map-value layout 6234281920b7SJesper Dangaard Brouer * 6235281920b7SJesper Dangaard Brouer * The struct data-layout of map-value is a configuration interface. 6236281920b7SJesper Dangaard Brouer * New members can only be added to the end of this structure. 6237281920b7SJesper Dangaard Brouer */ 6238281920b7SJesper Dangaard Brouer struct bpf_devmap_val { 6239281920b7SJesper Dangaard Brouer __u32 ifindex; /* device index */ 6240281920b7SJesper Dangaard Brouer union { 6241281920b7SJesper Dangaard Brouer int fd; /* prog fd on map write */ 6242281920b7SJesper Dangaard Brouer __u32 id; /* prog id on map read */ 6243281920b7SJesper Dangaard Brouer } bpf_prog; 6244281920b7SJesper Dangaard Brouer }; 6245281920b7SJesper Dangaard Brouer 6246644bfe51SLorenzo Bianconi /* CPUMAP map-value layout 6247644bfe51SLorenzo Bianconi * 6248644bfe51SLorenzo Bianconi * The struct data-layout of map-value is a configuration interface. 6249644bfe51SLorenzo Bianconi * New members can only be added to the end of this structure. 6250644bfe51SLorenzo Bianconi */ 6251644bfe51SLorenzo Bianconi struct bpf_cpumap_val { 6252644bfe51SLorenzo Bianconi __u32 qsize; /* queue size to remote target CPU */ 625392164774SLorenzo Bianconi union { 625492164774SLorenzo Bianconi int fd; /* prog fd on map write */ 625592164774SLorenzo Bianconi __u32 id; /* prog id on map read */ 625692164774SLorenzo Bianconi } bpf_prog; 6257644bfe51SLorenzo Bianconi }; 6258644bfe51SLorenzo Bianconi 6259174a79ffSJohn Fastabend enum sk_action { 6260bfa64075SJohn Fastabend SK_DROP = 0, 6261bfa64075SJohn Fastabend SK_PASS, 6262174a79ffSJohn Fastabend }; 6263174a79ffSJohn Fastabend 62644f738adbSJohn Fastabend /* user accessible metadata for SK_MSG packet hook, new fields must 62654f738adbSJohn Fastabend * be added to the end of this structure 62664f738adbSJohn Fastabend */ 62674f738adbSJohn Fastabend struct sk_msg_md { 6268b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 6269b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 6270303def35SJohn Fastabend 6271303def35SJohn Fastabend __u32 family; 6272303def35SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 6273303def35SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 6274303def35SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 6275303def35SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 6276303def35SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 6277303def35SJohn Fastabend __u32 local_port; /* stored in host byte order */ 62783bdbd022SJohn Fastabend __u32 size; /* Total size of sk_msg */ 627913d70f5aSJohn Fastabend 628013d70f5aSJohn Fastabend __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 62814f738adbSJohn Fastabend }; 62824f738adbSJohn Fastabend 62832dbb9b9eSMartin KaFai Lau struct sk_reuseport_md { 62842dbb9b9eSMartin KaFai Lau /* 62852dbb9b9eSMartin KaFai Lau * Start of directly accessible data. It begins from 62862dbb9b9eSMartin KaFai Lau * the tcp/udp header. 62872dbb9b9eSMartin KaFai Lau */ 6288b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 6289b7df9adaSDaniel Borkmann /* End of directly accessible data */ 6290b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 62912dbb9b9eSMartin KaFai Lau /* 62922dbb9b9eSMartin KaFai Lau * Total length of packet (starting from the tcp/udp header). 62932dbb9b9eSMartin KaFai Lau * Note that the directly accessible bytes (data_end - data) 62942dbb9b9eSMartin KaFai Lau * could be less than this "len". Those bytes could be 62952dbb9b9eSMartin KaFai Lau * indirectly read by a helper "bpf_skb_load_bytes()". 62962dbb9b9eSMartin KaFai Lau */ 62972dbb9b9eSMartin KaFai Lau __u32 len; 62982dbb9b9eSMartin KaFai Lau /* 62992dbb9b9eSMartin KaFai Lau * Eth protocol in the mac header (network byte order). e.g. 63002dbb9b9eSMartin KaFai Lau * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 63012dbb9b9eSMartin KaFai Lau */ 63022dbb9b9eSMartin KaFai Lau __u32 eth_protocol; 63032dbb9b9eSMartin KaFai Lau __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 63042dbb9b9eSMartin KaFai Lau __u32 bind_inany; /* Is sock bound to an INANY address? */ 63052dbb9b9eSMartin KaFai Lau __u32 hash; /* A hash of the packet 4 tuples */ 6306d5e4ddaeSKuniyuki Iwashima /* When reuse->migrating_sk is NULL, it is selecting a sk for the 6307d5e4ddaeSKuniyuki Iwashima * new incoming connection request (e.g. selecting a listen sk for 6308d5e4ddaeSKuniyuki Iwashima * the received SYN in the TCP case). reuse->sk is one of the sk 6309d5e4ddaeSKuniyuki Iwashima * in the reuseport group. The bpf prog can use reuse->sk to learn 6310d5e4ddaeSKuniyuki Iwashima * the local listening ip/port without looking into the skb. 6311d5e4ddaeSKuniyuki Iwashima * 6312d5e4ddaeSKuniyuki Iwashima * When reuse->migrating_sk is not NULL, reuse->sk is closed and 6313d5e4ddaeSKuniyuki Iwashima * reuse->migrating_sk is the socket that needs to be migrated 6314d5e4ddaeSKuniyuki Iwashima * to another listening socket. migrating_sk could be a fullsock 6315d5e4ddaeSKuniyuki Iwashima * sk that is fully established or a reqsk that is in-the-middle 6316d5e4ddaeSKuniyuki Iwashima * of 3-way handshake. 6317d5e4ddaeSKuniyuki Iwashima */ 6318e0610476SKuniyuki Iwashima __bpf_md_ptr(struct bpf_sock *, sk); 6319d5e4ddaeSKuniyuki Iwashima __bpf_md_ptr(struct bpf_sock *, migrating_sk); 63202dbb9b9eSMartin KaFai Lau }; 63212dbb9b9eSMartin KaFai Lau 63221e270976SMartin KaFai Lau #define BPF_TAG_SIZE 8 63231e270976SMartin KaFai Lau 63241e270976SMartin KaFai Lau struct bpf_prog_info { 63251e270976SMartin KaFai Lau __u32 type; 63261e270976SMartin KaFai Lau __u32 id; 63271e270976SMartin KaFai Lau __u8 tag[BPF_TAG_SIZE]; 63281e270976SMartin KaFai Lau __u32 jited_prog_len; 63291e270976SMartin KaFai Lau __u32 xlated_prog_len; 63301e270976SMartin KaFai Lau __aligned_u64 jited_prog_insns; 63311e270976SMartin KaFai Lau __aligned_u64 xlated_prog_insns; 6332cb4d2b3fSMartin KaFai Lau __u64 load_time; /* ns since boottime */ 6333cb4d2b3fSMartin KaFai Lau __u32 created_by_uid; 6334cb4d2b3fSMartin KaFai Lau __u32 nr_map_ids; 6335cb4d2b3fSMartin KaFai Lau __aligned_u64 map_ids; 6336067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 6337675fc275SJakub Kicinski __u32 ifindex; 6338b85fab0eSJiri Olsa __u32 gpl_compatible:1; 63390472301aSBaruch Siach __u32 :31; /* alignment pad */ 6340675fc275SJakub Kicinski __u64 netns_dev; 6341675fc275SJakub Kicinski __u64 netns_ino; 6342dbecd738SSandipan Das __u32 nr_jited_ksyms; 6343815581c1SSandipan Das __u32 nr_jited_func_lens; 6344dbecd738SSandipan Das __aligned_u64 jited_ksyms; 6345815581c1SSandipan Das __aligned_u64 jited_func_lens; 6346838e9690SYonghong Song __u32 btf_id; 6347838e9690SYonghong Song __u32 func_info_rec_size; 6348838e9690SYonghong Song __aligned_u64 func_info; 634911d8b82dSYonghong Song __u32 nr_func_info; 635011d8b82dSYonghong Song __u32 nr_line_info; 6351c454a46bSMartin KaFai Lau __aligned_u64 line_info; 6352c454a46bSMartin KaFai Lau __aligned_u64 jited_line_info; 635311d8b82dSYonghong Song __u32 nr_jited_line_info; 6354c454a46bSMartin KaFai Lau __u32 line_info_rec_size; 6355c454a46bSMartin KaFai Lau __u32 jited_line_info_rec_size; 6356c872bdb3SSong Liu __u32 nr_prog_tags; 6357c872bdb3SSong Liu __aligned_u64 prog_tags; 63585f8f8b93SAlexei Starovoitov __u64 run_time_ns; 63595f8f8b93SAlexei Starovoitov __u64 run_cnt; 63609ed9e9baSAlexei Starovoitov __u64 recursion_misses; 6361aba64c7dSDave Marchevsky __u32 verified_insns; 6362b79c9fc9SStanislav Fomichev __u32 attach_btf_obj_id; 6363b79c9fc9SStanislav Fomichev __u32 attach_btf_id; 63641e270976SMartin KaFai Lau } __attribute__((aligned(8))); 63651e270976SMartin KaFai Lau 63661e270976SMartin KaFai Lau struct bpf_map_info { 63671e270976SMartin KaFai Lau __u32 type; 63681e270976SMartin KaFai Lau __u32 id; 63691e270976SMartin KaFai Lau __u32 key_size; 63701e270976SMartin KaFai Lau __u32 value_size; 63711e270976SMartin KaFai Lau __u32 max_entries; 63721e270976SMartin KaFai Lau __u32 map_flags; 6373067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 637452775b33SJakub Kicinski __u32 ifindex; 637585d33df3SMartin KaFai Lau __u32 btf_vmlinux_value_type_id; 637652775b33SJakub Kicinski __u64 netns_dev; 637752775b33SJakub Kicinski __u64 netns_ino; 637878958fcaSMartin KaFai Lau __u32 btf_id; 63799b2cf328SMartin KaFai Lau __u32 btf_key_type_id; 63809b2cf328SMartin KaFai Lau __u32 btf_value_type_id; 63818845b468SJoanne Koong __u32 :32; /* alignment pad */ 63829330986cSJoanne Koong __u64 map_extra; 63831e270976SMartin KaFai Lau } __attribute__((aligned(8))); 63841e270976SMartin KaFai Lau 638562dab84cSMartin KaFai Lau struct bpf_btf_info { 638662dab84cSMartin KaFai Lau __aligned_u64 btf; 638762dab84cSMartin KaFai Lau __u32 btf_size; 638862dab84cSMartin KaFai Lau __u32 id; 638953297220SAndrii Nakryiko __aligned_u64 name; 639053297220SAndrii Nakryiko __u32 name_len; 639153297220SAndrii Nakryiko __u32 kernel_btf; 639262dab84cSMartin KaFai Lau } __attribute__((aligned(8))); 639362dab84cSMartin KaFai Lau 6394f2e10bffSAndrii Nakryiko struct bpf_link_info { 6395f2e10bffSAndrii Nakryiko __u32 type; 6396f2e10bffSAndrii Nakryiko __u32 id; 6397f2e10bffSAndrii Nakryiko __u32 prog_id; 6398f2e10bffSAndrii Nakryiko union { 6399f2e10bffSAndrii Nakryiko struct { 6400f2e10bffSAndrii Nakryiko __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 6401f2e10bffSAndrii Nakryiko __u32 tp_name_len; /* in/out: tp_name buffer len */ 6402f2e10bffSAndrii Nakryiko } raw_tracepoint; 6403f2e10bffSAndrii Nakryiko struct { 6404f2e10bffSAndrii Nakryiko __u32 attach_type; 6405441e8c66SToke Høiland-Jørgensen __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 6406441e8c66SToke Høiland-Jørgensen __u32 target_btf_id; /* BTF type id inside the object */ 6407f2e10bffSAndrii Nakryiko } tracing; 6408f2e10bffSAndrii Nakryiko struct { 6409f2e10bffSAndrii Nakryiko __u64 cgroup_id; 6410f2e10bffSAndrii Nakryiko __u32 attach_type; 6411f2e10bffSAndrii Nakryiko } cgroup; 64127f045a49SJakub Sitnicki struct { 64136b0a249aSYonghong Song __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 64146b0a249aSYonghong Song __u32 target_name_len; /* in/out: target_name buffer len */ 6415d4ccaf58SHao Luo 6416d4ccaf58SHao Luo /* If the iter specific field is 32 bits, it can be put 6417d4ccaf58SHao Luo * in the first or second union. Otherwise it should be 6418d4ccaf58SHao Luo * put in the second union. 6419d4ccaf58SHao Luo */ 64206b0a249aSYonghong Song union { 6421b0c9eb37SYonghong Song struct { 64226b0a249aSYonghong Song __u32 map_id; 64236b0a249aSYonghong Song } map; 6424b0c9eb37SYonghong Song }; 6425d4ccaf58SHao Luo union { 6426d4ccaf58SHao Luo struct { 6427d4ccaf58SHao Luo __u64 cgroup_id; 6428d4ccaf58SHao Luo __u32 order; 6429d4ccaf58SHao Luo } cgroup; 643021fb6f2aSKui-Feng Lee struct { 643121fb6f2aSKui-Feng Lee __u32 tid; 643221fb6f2aSKui-Feng Lee __u32 pid; 643321fb6f2aSKui-Feng Lee } task; 6434d4ccaf58SHao Luo }; 64356b0a249aSYonghong Song } iter; 64366b0a249aSYonghong Song struct { 64377f045a49SJakub Sitnicki __u32 netns_ino; 64387f045a49SJakub Sitnicki __u32 attach_type; 64397f045a49SJakub Sitnicki } netns; 6440c1931c97SAndrii Nakryiko struct { 6441c1931c97SAndrii Nakryiko __u32 ifindex; 6442c1931c97SAndrii Nakryiko } xdp; 644368b04864SKui-Feng Lee struct { 644468b04864SKui-Feng Lee __u32 map_id; 644568b04864SKui-Feng Lee } struct_ops; 644684601d6eSFlorian Westphal struct { 644784601d6eSFlorian Westphal __u32 pf; 644884601d6eSFlorian Westphal __u32 hooknum; 644984601d6eSFlorian Westphal __s32 priority; 645084601d6eSFlorian Westphal __u32 flags; 645184601d6eSFlorian Westphal } netfilter; 64527ac8d0d2SYafang Shao struct { 64537ac8d0d2SYafang Shao __aligned_u64 addrs; 64547ac8d0d2SYafang Shao __u32 count; /* in/out: kprobe_multi function count */ 64557ac8d0d2SYafang Shao __u32 flags; 64567ac8d0d2SYafang Shao } kprobe_multi; 6457*1b715e1bSYafang Shao struct { 6458*1b715e1bSYafang Shao __u32 type; /* enum bpf_perf_event_type */ 6459*1b715e1bSYafang Shao __u32 :32; 6460*1b715e1bSYafang Shao union { 6461*1b715e1bSYafang Shao struct { 6462*1b715e1bSYafang Shao __aligned_u64 file_name; /* in/out */ 6463*1b715e1bSYafang Shao __u32 name_len; 6464*1b715e1bSYafang Shao __u32 offset; /* offset from file_name */ 6465*1b715e1bSYafang Shao } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */ 6466*1b715e1bSYafang Shao struct { 6467*1b715e1bSYafang Shao __aligned_u64 func_name; /* in/out */ 6468*1b715e1bSYafang Shao __u32 name_len; 6469*1b715e1bSYafang Shao __u32 offset; /* offset from func_name */ 6470*1b715e1bSYafang Shao __u64 addr; 6471*1b715e1bSYafang Shao } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */ 6472*1b715e1bSYafang Shao struct { 6473*1b715e1bSYafang Shao __aligned_u64 tp_name; /* in/out */ 6474*1b715e1bSYafang Shao __u32 name_len; 6475*1b715e1bSYafang Shao } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */ 6476*1b715e1bSYafang Shao struct { 6477*1b715e1bSYafang Shao __u64 config; 6478*1b715e1bSYafang Shao __u32 type; 6479*1b715e1bSYafang Shao } event; /* BPF_PERF_EVENT_EVENT */ 6480*1b715e1bSYafang Shao }; 6481*1b715e1bSYafang Shao } perf_event; 6482f2e10bffSAndrii Nakryiko }; 6483f2e10bffSAndrii Nakryiko } __attribute__((aligned(8))); 6484f2e10bffSAndrii Nakryiko 64854fbac77dSAndrey Ignatov /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 64864fbac77dSAndrey Ignatov * by user and intended to be used by socket (e.g. to bind to, depends on 6487bfdfa517SRandy Dunlap * attach type). 64884fbac77dSAndrey Ignatov */ 64894fbac77dSAndrey Ignatov struct bpf_sock_addr { 64904fbac77dSAndrey Ignatov __u32 user_family; /* Allows 4-byte read, but no write. */ 64914fbac77dSAndrey Ignatov __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 64924fbac77dSAndrey Ignatov * Stored in network byte order. 64934fbac77dSAndrey Ignatov */ 6494d4ecfeb1SStanislav Fomichev __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 64954fbac77dSAndrey Ignatov * Stored in network byte order. 64964fbac77dSAndrey Ignatov */ 64977aebfa1bSAndrey Ignatov __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 64984fbac77dSAndrey Ignatov * Stored in network byte order 64994fbac77dSAndrey Ignatov */ 65004fbac77dSAndrey Ignatov __u32 family; /* Allows 4-byte read, but no write */ 65014fbac77dSAndrey Ignatov __u32 type; /* Allows 4-byte read, but no write */ 65024fbac77dSAndrey Ignatov __u32 protocol; /* Allows 4-byte read, but no write */ 6503600c70baSStanislav Fomichev __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 65041cedee13SAndrey Ignatov * Stored in network byte order. 65051cedee13SAndrey Ignatov */ 6506d4ecfeb1SStanislav Fomichev __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 65071cedee13SAndrey Ignatov * Stored in network byte order. 65081cedee13SAndrey Ignatov */ 6509fb85c4a7SStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 65104fbac77dSAndrey Ignatov }; 65114fbac77dSAndrey Ignatov 651240304b2aSLawrence Brakmo /* User bpf_sock_ops struct to access socket values and specify request ops 651340304b2aSLawrence Brakmo * and their replies. 651440304b2aSLawrence Brakmo * Some of this fields are in network (bigendian) byte order and may need 651540304b2aSLawrence Brakmo * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 651640304b2aSLawrence Brakmo * New fields can only be added at the end of this structure 651740304b2aSLawrence Brakmo */ 651840304b2aSLawrence Brakmo struct bpf_sock_ops { 651940304b2aSLawrence Brakmo __u32 op; 652040304b2aSLawrence Brakmo union { 6521de525be2SLawrence Brakmo __u32 args[4]; /* Optionally passed to bpf program */ 6522de525be2SLawrence Brakmo __u32 reply; /* Returned by bpf program */ 6523de525be2SLawrence Brakmo __u32 replylong[4]; /* Optionally returned by bpf prog */ 652440304b2aSLawrence Brakmo }; 652540304b2aSLawrence Brakmo __u32 family; 652640304b2aSLawrence Brakmo __u32 remote_ip4; /* Stored in network byte order */ 652740304b2aSLawrence Brakmo __u32 local_ip4; /* Stored in network byte order */ 652840304b2aSLawrence Brakmo __u32 remote_ip6[4]; /* Stored in network byte order */ 652940304b2aSLawrence Brakmo __u32 local_ip6[4]; /* Stored in network byte order */ 653040304b2aSLawrence Brakmo __u32 remote_port; /* Stored in network byte order */ 653140304b2aSLawrence Brakmo __u32 local_port; /* stored in host byte order */ 6532f19397a5SLawrence Brakmo __u32 is_fullsock; /* Some TCP fields are only valid if 6533f19397a5SLawrence Brakmo * there is a full socket. If not, the 6534f19397a5SLawrence Brakmo * fields read as zero. 6535f19397a5SLawrence Brakmo */ 6536f19397a5SLawrence Brakmo __u32 snd_cwnd; 6537f19397a5SLawrence Brakmo __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 6538b13d8807SLawrence Brakmo __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 653944f0e430SLawrence Brakmo __u32 state; 654044f0e430SLawrence Brakmo __u32 rtt_min; 654144f0e430SLawrence Brakmo __u32 snd_ssthresh; 654244f0e430SLawrence Brakmo __u32 rcv_nxt; 654344f0e430SLawrence Brakmo __u32 snd_nxt; 654444f0e430SLawrence Brakmo __u32 snd_una; 654544f0e430SLawrence Brakmo __u32 mss_cache; 654644f0e430SLawrence Brakmo __u32 ecn_flags; 654744f0e430SLawrence Brakmo __u32 rate_delivered; 654844f0e430SLawrence Brakmo __u32 rate_interval_us; 654944f0e430SLawrence Brakmo __u32 packets_out; 655044f0e430SLawrence Brakmo __u32 retrans_out; 655144f0e430SLawrence Brakmo __u32 total_retrans; 655244f0e430SLawrence Brakmo __u32 segs_in; 655344f0e430SLawrence Brakmo __u32 data_segs_in; 655444f0e430SLawrence Brakmo __u32 segs_out; 655544f0e430SLawrence Brakmo __u32 data_segs_out; 655644f0e430SLawrence Brakmo __u32 lost_out; 655744f0e430SLawrence Brakmo __u32 sacked_out; 655844f0e430SLawrence Brakmo __u32 sk_txhash; 655944f0e430SLawrence Brakmo __u64 bytes_received; 656044f0e430SLawrence Brakmo __u64 bytes_acked; 65611314ef56SStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 65620813a841SMartin KaFai Lau /* [skb_data, skb_data_end) covers the whole TCP header. 65630813a841SMartin KaFai Lau * 65640813a841SMartin KaFai Lau * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 65650813a841SMartin KaFai Lau * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 65660813a841SMartin KaFai Lau * header has not been written. 65670813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 65680813a841SMartin KaFai Lau * been written so far. 65690813a841SMartin KaFai Lau * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 65700813a841SMartin KaFai Lau * the 3WHS. 65710813a841SMartin KaFai Lau * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 65720813a841SMartin KaFai Lau * the 3WHS. 65730813a841SMartin KaFai Lau * 65740813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also be used to read a particular option. 65750813a841SMartin KaFai Lau */ 65760813a841SMartin KaFai Lau __bpf_md_ptr(void *, skb_data); 65770813a841SMartin KaFai Lau __bpf_md_ptr(void *, skb_data_end); 65780813a841SMartin KaFai Lau __u32 skb_len; /* The total length of a packet. 65790813a841SMartin KaFai Lau * It includes the header, options, 65800813a841SMartin KaFai Lau * and payload. 65810813a841SMartin KaFai Lau */ 65820813a841SMartin KaFai Lau __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 65830813a841SMartin KaFai Lau * an easy way to check for tcp_flags 65840813a841SMartin KaFai Lau * without parsing skb_data. 65850813a841SMartin KaFai Lau * 65860813a841SMartin KaFai Lau * In particular, the skb_tcp_flags 65870813a841SMartin KaFai Lau * will still be available in 65880813a841SMartin KaFai Lau * BPF_SOCK_OPS_HDR_OPT_LEN even though 65890813a841SMartin KaFai Lau * the outgoing header has not 65900813a841SMartin KaFai Lau * been written yet. 65910813a841SMartin KaFai Lau */ 65929bb05349SMartin KaFai Lau __u64 skb_hwtstamp; 659340304b2aSLawrence Brakmo }; 659440304b2aSLawrence Brakmo 6595b13d8807SLawrence Brakmo /* Definitions for bpf_sock_ops_cb_flags */ 65961aae4bddSAndrii Nakryiko enum { 65971aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 65981aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 65991aae4bddSAndrii Nakryiko BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 66001aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 66010813a841SMartin KaFai Lau /* Call bpf for all received TCP headers. The bpf prog will be 66020813a841SMartin KaFai Lau * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 66030813a841SMartin KaFai Lau * 66040813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 66050813a841SMartin KaFai Lau * for the header option related helpers that will be useful 66060813a841SMartin KaFai Lau * to the bpf programs. 66070813a841SMartin KaFai Lau * 66080813a841SMartin KaFai Lau * It could be used at the client/active side (i.e. connect() side) 66090813a841SMartin KaFai Lau * when the server told it that the server was in syncookie 66100813a841SMartin KaFai Lau * mode and required the active side to resend the bpf-written 66110813a841SMartin KaFai Lau * options. The active side can keep writing the bpf-options until 66120813a841SMartin KaFai Lau * it received a valid packet from the server side to confirm 66130813a841SMartin KaFai Lau * the earlier packet (and options) has been received. The later 66140813a841SMartin KaFai Lau * example patch is using it like this at the active side when the 66150813a841SMartin KaFai Lau * server is in syncookie mode. 66160813a841SMartin KaFai Lau * 66170813a841SMartin KaFai Lau * The bpf prog will usually turn this off in the common cases. 66180813a841SMartin KaFai Lau */ 661900d211a4SMartin KaFai Lau BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 66200813a841SMartin KaFai Lau /* Call bpf when kernel has received a header option that 66210813a841SMartin KaFai Lau * the kernel cannot handle. The bpf prog will be called under 66220813a841SMartin KaFai Lau * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 66230813a841SMartin KaFai Lau * 66240813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 66250813a841SMartin KaFai Lau * for the header option related helpers that will be useful 66260813a841SMartin KaFai Lau * to the bpf programs. 66270813a841SMartin KaFai Lau */ 662800d211a4SMartin KaFai Lau BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 66290813a841SMartin KaFai Lau /* Call bpf when the kernel is writing header options for the 66300813a841SMartin KaFai Lau * outgoing packet. The bpf prog will first be called 66310813a841SMartin KaFai Lau * to reserve space in a skb under 66320813a841SMartin KaFai Lau * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 66330813a841SMartin KaFai Lau * the bpf prog will be called to write the header option(s) 66340813a841SMartin KaFai Lau * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 66350813a841SMartin KaFai Lau * 66360813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 66370813a841SMartin KaFai Lau * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 66380813a841SMartin KaFai Lau * related helpers that will be useful to the bpf programs. 66390813a841SMartin KaFai Lau * 66400813a841SMartin KaFai Lau * The kernel gets its chance to reserve space and write 66410813a841SMartin KaFai Lau * options first before the BPF program does. 66420813a841SMartin KaFai Lau */ 6643331fca43SMartin KaFai Lau BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 66441aae4bddSAndrii Nakryiko /* Mask of all currently supported cb flags */ 6645331fca43SMartin KaFai Lau BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 66461aae4bddSAndrii Nakryiko }; 6647b13d8807SLawrence Brakmo 664840304b2aSLawrence Brakmo /* List of known BPF sock_ops operators. 664940304b2aSLawrence Brakmo * New entries can only be added at the end 665040304b2aSLawrence Brakmo */ 665140304b2aSLawrence Brakmo enum { 665240304b2aSLawrence Brakmo BPF_SOCK_OPS_VOID, 66538550f328SLawrence Brakmo BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 66548550f328SLawrence Brakmo * -1 if default value should be used 66558550f328SLawrence Brakmo */ 665613d3b1ebSLawrence Brakmo BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 665713d3b1ebSLawrence Brakmo * window (in packets) or -1 if default 665813d3b1ebSLawrence Brakmo * value should be used 665913d3b1ebSLawrence Brakmo */ 66609872a4bdSLawrence Brakmo BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 66619872a4bdSLawrence Brakmo * active connection is initialized 66629872a4bdSLawrence Brakmo */ 66639872a4bdSLawrence Brakmo BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 66649872a4bdSLawrence Brakmo * active connection is 66659872a4bdSLawrence Brakmo * established 66669872a4bdSLawrence Brakmo */ 66679872a4bdSLawrence Brakmo BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 66689872a4bdSLawrence Brakmo * passive connection is 66699872a4bdSLawrence Brakmo * established 66709872a4bdSLawrence Brakmo */ 667191b5b21cSLawrence Brakmo BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 667291b5b21cSLawrence Brakmo * needs ECN 667391b5b21cSLawrence Brakmo */ 6674e6546ef6SLawrence Brakmo BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 6675e6546ef6SLawrence Brakmo * based on the path and may be 6676e6546ef6SLawrence Brakmo * dependent on the congestion control 6677e6546ef6SLawrence Brakmo * algorithm. In general it indicates 6678e6546ef6SLawrence Brakmo * a congestion threshold. RTTs above 6679e6546ef6SLawrence Brakmo * this indicate congestion 6680e6546ef6SLawrence Brakmo */ 6681f89013f6SLawrence Brakmo BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 6682f89013f6SLawrence Brakmo * Arg1: value of icsk_retransmits 6683f89013f6SLawrence Brakmo * Arg2: value of icsk_rto 6684f89013f6SLawrence Brakmo * Arg3: whether RTO has expired 6685f89013f6SLawrence Brakmo */ 6686a31ad29eSLawrence Brakmo BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6687a31ad29eSLawrence Brakmo * Arg1: sequence number of 1st byte 6688a31ad29eSLawrence Brakmo * Arg2: # segments 6689a31ad29eSLawrence Brakmo * Arg3: return value of 6690a31ad29eSLawrence Brakmo * tcp_transmit_skb (0 => success) 6691a31ad29eSLawrence Brakmo */ 6692d4487491SLawrence Brakmo BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6693d4487491SLawrence Brakmo * Arg1: old_state 6694d4487491SLawrence Brakmo * Arg2: new_state 6695d4487491SLawrence Brakmo */ 6696f333ee0cSAndrey Ignatov BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6697f333ee0cSAndrey Ignatov * socket transition to LISTEN state. 6698f333ee0cSAndrey Ignatov */ 669923729ff2SStanislav Fomichev BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 670023729ff2SStanislav Fomichev */ 67010813a841SMartin KaFai Lau BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 67020813a841SMartin KaFai Lau * It will be called to handle 67030813a841SMartin KaFai Lau * the packets received at 67040813a841SMartin KaFai Lau * an already established 67050813a841SMartin KaFai Lau * connection. 67060813a841SMartin KaFai Lau * 67070813a841SMartin KaFai Lau * sock_ops->skb_data: 67080813a841SMartin KaFai Lau * Referring to the received skb. 67090813a841SMartin KaFai Lau * It covers the TCP header only. 67100813a841SMartin KaFai Lau * 67110813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also 67120813a841SMartin KaFai Lau * be used to search for a 67130813a841SMartin KaFai Lau * particular option. 67140813a841SMartin KaFai Lau */ 67150813a841SMartin KaFai Lau BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 67160813a841SMartin KaFai Lau * header option later in 67170813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 67180813a841SMartin KaFai Lau * Arg1: bool want_cookie. (in 67190813a841SMartin KaFai Lau * writing SYNACK only) 67200813a841SMartin KaFai Lau * 67210813a841SMartin KaFai Lau * sock_ops->skb_data: 67220813a841SMartin KaFai Lau * Not available because no header has 67230813a841SMartin KaFai Lau * been written yet. 67240813a841SMartin KaFai Lau * 67250813a841SMartin KaFai Lau * sock_ops->skb_tcp_flags: 67260813a841SMartin KaFai Lau * The tcp_flags of the 67270813a841SMartin KaFai Lau * outgoing skb. (e.g. SYN, ACK, FIN). 67280813a841SMartin KaFai Lau * 67290813a841SMartin KaFai Lau * bpf_reserve_hdr_opt() should 67300813a841SMartin KaFai Lau * be used to reserve space. 67310813a841SMartin KaFai Lau */ 67320813a841SMartin KaFai Lau BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 67330813a841SMartin KaFai Lau * Arg1: bool want_cookie. (in 67340813a841SMartin KaFai Lau * writing SYNACK only) 67350813a841SMartin KaFai Lau * 67360813a841SMartin KaFai Lau * sock_ops->skb_data: 67370813a841SMartin KaFai Lau * Referring to the outgoing skb. 67380813a841SMartin KaFai Lau * It covers the TCP header 67390813a841SMartin KaFai Lau * that has already been written 67400813a841SMartin KaFai Lau * by the kernel and the 67410813a841SMartin KaFai Lau * earlier bpf-progs. 67420813a841SMartin KaFai Lau * 67430813a841SMartin KaFai Lau * sock_ops->skb_tcp_flags: 67440813a841SMartin KaFai Lau * The tcp_flags of the outgoing 67450813a841SMartin KaFai Lau * skb. (e.g. SYN, ACK, FIN). 67460813a841SMartin KaFai Lau * 67470813a841SMartin KaFai Lau * bpf_store_hdr_opt() should 67480813a841SMartin KaFai Lau * be used to write the 67490813a841SMartin KaFai Lau * option. 67500813a841SMartin KaFai Lau * 67510813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also 67520813a841SMartin KaFai Lau * be used to search for a 67530813a841SMartin KaFai Lau * particular option that 67540813a841SMartin KaFai Lau * has already been written 67550813a841SMartin KaFai Lau * by the kernel or the 67560813a841SMartin KaFai Lau * earlier bpf-progs. 67570813a841SMartin KaFai Lau */ 6758d4487491SLawrence Brakmo }; 6759d4487491SLawrence Brakmo 6760d4487491SLawrence Brakmo /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 6761d4487491SLawrence Brakmo * changes between the TCP and BPF versions. Ideally this should never happen. 6762d4487491SLawrence Brakmo * If it does, we need to add code to convert them before calling 6763d4487491SLawrence Brakmo * the BPF sock_ops function. 6764d4487491SLawrence Brakmo */ 6765d4487491SLawrence Brakmo enum { 6766d4487491SLawrence Brakmo BPF_TCP_ESTABLISHED = 1, 6767d4487491SLawrence Brakmo BPF_TCP_SYN_SENT, 6768d4487491SLawrence Brakmo BPF_TCP_SYN_RECV, 6769d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT1, 6770d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT2, 6771d4487491SLawrence Brakmo BPF_TCP_TIME_WAIT, 6772d4487491SLawrence Brakmo BPF_TCP_CLOSE, 6773d4487491SLawrence Brakmo BPF_TCP_CLOSE_WAIT, 6774d4487491SLawrence Brakmo BPF_TCP_LAST_ACK, 6775d4487491SLawrence Brakmo BPF_TCP_LISTEN, 6776d4487491SLawrence Brakmo BPF_TCP_CLOSING, /* Now a valid state */ 6777d4487491SLawrence Brakmo BPF_TCP_NEW_SYN_RECV, 6778d4487491SLawrence Brakmo 6779d4487491SLawrence Brakmo BPF_TCP_MAX_STATES /* Leave at the end! */ 678040304b2aSLawrence Brakmo }; 678140304b2aSLawrence Brakmo 67821aae4bddSAndrii Nakryiko enum { 67831aae4bddSAndrii Nakryiko TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 67841aae4bddSAndrii Nakryiko TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 67852b8ee4f0SMartin KaFai Lau TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 6786ca584ba0SMartin KaFai Lau TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 67870813a841SMartin KaFai Lau /* Copy the SYN pkt to optval 67880813a841SMartin KaFai Lau * 67890813a841SMartin KaFai Lau * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 67900813a841SMartin KaFai Lau * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 67910813a841SMartin KaFai Lau * to only getting from the saved_syn. It can either get the 67920813a841SMartin KaFai Lau * syn packet from: 67930813a841SMartin KaFai Lau * 67940813a841SMartin KaFai Lau * 1. the just-received SYN packet (only available when writing the 67950813a841SMartin KaFai Lau * SYNACK). It will be useful when it is not necessary to 67960813a841SMartin KaFai Lau * save the SYN packet for latter use. It is also the only way 67970813a841SMartin KaFai Lau * to get the SYN during syncookie mode because the syn 67980813a841SMartin KaFai Lau * packet cannot be saved during syncookie. 67990813a841SMartin KaFai Lau * 68000813a841SMartin KaFai Lau * OR 68010813a841SMartin KaFai Lau * 68020813a841SMartin KaFai Lau * 2. the earlier saved syn which was done by 68030813a841SMartin KaFai Lau * bpf_setsockopt(TCP_SAVE_SYN). 68040813a841SMartin KaFai Lau * 68050813a841SMartin KaFai Lau * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 68060813a841SMartin KaFai Lau * SYN packet is obtained. 68070813a841SMartin KaFai Lau * 68080813a841SMartin KaFai Lau * If the bpf-prog does not need the IP[46] header, the 68090813a841SMartin KaFai Lau * bpf-prog can avoid parsing the IP header by using 68100813a841SMartin KaFai Lau * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 68110813a841SMartin KaFai Lau * IP[46] and TCP header by using TCP_BPF_SYN_IP. 68120813a841SMartin KaFai Lau * 68130813a841SMartin KaFai Lau * >0: Total number of bytes copied 68140813a841SMartin KaFai Lau * -ENOSPC: Not enough space in optval. Only optlen number of 68150813a841SMartin KaFai Lau * bytes is copied. 68160813a841SMartin KaFai Lau * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 68170813a841SMartin KaFai Lau * is not saved by setsockopt(TCP_SAVE_SYN). 68180813a841SMartin KaFai Lau */ 68190813a841SMartin KaFai Lau TCP_BPF_SYN = 1005, /* Copy the TCP header */ 68200813a841SMartin KaFai Lau TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 6821267cf9faSMartin KaFai Lau TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 68220813a841SMartin KaFai Lau }; 68230813a841SMartin KaFai Lau 68240813a841SMartin KaFai Lau enum { 68250813a841SMartin KaFai Lau BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 68260813a841SMartin KaFai Lau }; 68270813a841SMartin KaFai Lau 68280813a841SMartin KaFai Lau /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 68290813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 68300813a841SMartin KaFai Lau */ 68310813a841SMartin KaFai Lau enum { 68320813a841SMartin KaFai Lau BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 68330813a841SMartin KaFai Lau * total option spaces 68340813a841SMartin KaFai Lau * required for an established 68350813a841SMartin KaFai Lau * sk in order to calculate the 68360813a841SMartin KaFai Lau * MSS. No skb is actually 68370813a841SMartin KaFai Lau * sent. 68380813a841SMartin KaFai Lau */ 68390813a841SMartin KaFai Lau BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 68400813a841SMartin KaFai Lau * when sending a SYN. 68410813a841SMartin KaFai Lau */ 68421aae4bddSAndrii Nakryiko }; 6843fc747810SLawrence Brakmo 6844908432caSYonghong Song struct bpf_perf_event_value { 6845908432caSYonghong Song __u64 counter; 6846908432caSYonghong Song __u64 enabled; 6847908432caSYonghong Song __u64 running; 6848908432caSYonghong Song }; 6849908432caSYonghong Song 68501aae4bddSAndrii Nakryiko enum { 68511aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 68521aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_READ = (1ULL << 1), 68531aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_WRITE = (1ULL << 2), 68541aae4bddSAndrii Nakryiko }; 6855ebc614f6SRoman Gushchin 68561aae4bddSAndrii Nakryiko enum { 68571aae4bddSAndrii Nakryiko BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 68581aae4bddSAndrii Nakryiko BPF_DEVCG_DEV_CHAR = (1ULL << 1), 68591aae4bddSAndrii Nakryiko }; 6860ebc614f6SRoman Gushchin 6861ebc614f6SRoman Gushchin struct bpf_cgroup_dev_ctx { 686206ef0ccbSYonghong Song /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 686306ef0ccbSYonghong Song __u32 access_type; 6864ebc614f6SRoman Gushchin __u32 major; 6865ebc614f6SRoman Gushchin __u32 minor; 6866ebc614f6SRoman Gushchin }; 6867ebc614f6SRoman Gushchin 6868c4f6699dSAlexei Starovoitov struct bpf_raw_tracepoint_args { 6869c4f6699dSAlexei Starovoitov __u64 args[0]; 6870c4f6699dSAlexei Starovoitov }; 6871c4f6699dSAlexei Starovoitov 687287f5fc7eSDavid Ahern /* DIRECT: Skip the FIB rules and go to FIB table associated with device 687387f5fc7eSDavid Ahern * OUTPUT: Do lookup from egress perspective; default is ingress 687487f5fc7eSDavid Ahern */ 68751aae4bddSAndrii Nakryiko enum { 68761aae4bddSAndrii Nakryiko BPF_FIB_LOOKUP_DIRECT = (1U << 0), 68771aae4bddSAndrii Nakryiko BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 687831de4105SMartin KaFai Lau BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2), 68798ad77e72SLouis DeLosSantos BPF_FIB_LOOKUP_TBID = (1U << 3), 68801aae4bddSAndrii Nakryiko }; 688187f5fc7eSDavid Ahern 68824c79579bSDavid Ahern enum { 68834c79579bSDavid Ahern BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 68844c79579bSDavid Ahern BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 68854c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 68864c79579bSDavid Ahern BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 68874c79579bSDavid Ahern BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 68884c79579bSDavid Ahern BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 68894c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 68904c79579bSDavid Ahern BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 68914c79579bSDavid Ahern BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 68924c79579bSDavid Ahern }; 68934c79579bSDavid Ahern 689487f5fc7eSDavid Ahern struct bpf_fib_lookup { 6895fa898d76SDavid Ahern /* input: network family for lookup (AF_INET, AF_INET6) 6896fa898d76SDavid Ahern * output: network family of egress nexthop 6897fa898d76SDavid Ahern */ 6898fa898d76SDavid Ahern __u8 family; 689987f5fc7eSDavid Ahern 690087f5fc7eSDavid Ahern /* set if lookup is to consider L4 data - e.g., FIB rules */ 690187f5fc7eSDavid Ahern __u8 l4_protocol; 690287f5fc7eSDavid Ahern __be16 sport; 690387f5fc7eSDavid Ahern __be16 dport; 690487f5fc7eSDavid Ahern 6905e1850ea9SJesper Dangaard Brouer union { /* used for MTU check */ 6906e1850ea9SJesper Dangaard Brouer /* input to lookup */ 6907e1850ea9SJesper Dangaard Brouer __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 69084c79579bSDavid Ahern 6909e1850ea9SJesper Dangaard Brouer /* output: MTU value */ 6910e1850ea9SJesper Dangaard Brouer __u16 mtu_result; 6911e1850ea9SJesper Dangaard Brouer }; 69124c79579bSDavid Ahern /* input: L3 device index for lookup 69134c79579bSDavid Ahern * output: device index from FIB lookup 69144c79579bSDavid Ahern */ 69154c79579bSDavid Ahern __u32 ifindex; 691687f5fc7eSDavid Ahern 691787f5fc7eSDavid Ahern union { 691887f5fc7eSDavid Ahern /* inputs to lookup */ 691987f5fc7eSDavid Ahern __u8 tos; /* AF_INET */ 6920bd3a08aaSDavid Ahern __be32 flowinfo; /* AF_INET6, flow_label + priority */ 692187f5fc7eSDavid Ahern 6922fa898d76SDavid Ahern /* output: metric of fib result (IPv4/IPv6 only) */ 692387f5fc7eSDavid Ahern __u32 rt_metric; 692487f5fc7eSDavid Ahern }; 692587f5fc7eSDavid Ahern 692687f5fc7eSDavid Ahern union { 692787f5fc7eSDavid Ahern __be32 ipv4_src; 692887f5fc7eSDavid Ahern __u32 ipv6_src[4]; /* in6_addr; network order */ 692987f5fc7eSDavid Ahern }; 693087f5fc7eSDavid Ahern 6931fa898d76SDavid Ahern /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 6932fa898d76SDavid Ahern * network header. output: bpf_fib_lookup sets to gateway address 6933fa898d76SDavid Ahern * if FIB lookup returns gateway route 693487f5fc7eSDavid Ahern */ 693587f5fc7eSDavid Ahern union { 693687f5fc7eSDavid Ahern __be32 ipv4_dst; 693787f5fc7eSDavid Ahern __u32 ipv6_dst[4]; /* in6_addr; network order */ 693887f5fc7eSDavid Ahern }; 693987f5fc7eSDavid Ahern 69408ad77e72SLouis DeLosSantos union { 69418ad77e72SLouis DeLosSantos struct { 694287f5fc7eSDavid Ahern /* output */ 694387f5fc7eSDavid Ahern __be16 h_vlan_proto; 694487f5fc7eSDavid Ahern __be16 h_vlan_TCI; 69458ad77e72SLouis DeLosSantos }; 69468ad77e72SLouis DeLosSantos /* input: when accompanied with the 69478ad77e72SLouis DeLosSantos * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a 69488ad77e72SLouis DeLosSantos * specific routing table to use for the fib lookup. 69498ad77e72SLouis DeLosSantos */ 69508ad77e72SLouis DeLosSantos __u32 tbid; 69518ad77e72SLouis DeLosSantos }; 69528ad77e72SLouis DeLosSantos 695387f5fc7eSDavid Ahern __u8 smac[6]; /* ETH_ALEN */ 695487f5fc7eSDavid Ahern __u8 dmac[6]; /* ETH_ALEN */ 695587f5fc7eSDavid Ahern }; 695687f5fc7eSDavid Ahern 6957ba452c9eSToke Høiland-Jørgensen struct bpf_redir_neigh { 6958ba452c9eSToke Høiland-Jørgensen /* network family for lookup (AF_INET, AF_INET6) */ 6959ba452c9eSToke Høiland-Jørgensen __u32 nh_family; 6960ba452c9eSToke Høiland-Jørgensen /* network address of nexthop; skips fib lookup to find gateway */ 6961ba452c9eSToke Høiland-Jørgensen union { 6962ba452c9eSToke Høiland-Jørgensen __be32 ipv4_nh; 6963ba452c9eSToke Høiland-Jørgensen __u32 ipv6_nh[4]; /* in6_addr; network order */ 6964ba452c9eSToke Høiland-Jørgensen }; 6965ba452c9eSToke Høiland-Jørgensen }; 6966ba452c9eSToke Høiland-Jørgensen 696734b2021cSJesper Dangaard Brouer /* bpf_check_mtu flags*/ 696834b2021cSJesper Dangaard Brouer enum bpf_check_mtu_flags { 696934b2021cSJesper Dangaard Brouer BPF_MTU_CHK_SEGS = (1U << 0), 697034b2021cSJesper Dangaard Brouer }; 697134b2021cSJesper Dangaard Brouer 697234b2021cSJesper Dangaard Brouer enum bpf_check_mtu_ret { 697334b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 697434b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 697534b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 697634b2021cSJesper Dangaard Brouer }; 697734b2021cSJesper Dangaard Brouer 697841bdc4b4SYonghong Song enum bpf_task_fd_type { 697941bdc4b4SYonghong Song BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 698041bdc4b4SYonghong Song BPF_FD_TYPE_TRACEPOINT, /* tp name */ 698141bdc4b4SYonghong Song BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 698241bdc4b4SYonghong Song BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 698341bdc4b4SYonghong Song BPF_FD_TYPE_UPROBE, /* filename + offset */ 698441bdc4b4SYonghong Song BPF_FD_TYPE_URETPROBE, /* filename + offset */ 698541bdc4b4SYonghong Song }; 698641bdc4b4SYonghong Song 69871aae4bddSAndrii Nakryiko enum { 69881aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 69891aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 69901aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 69911aae4bddSAndrii Nakryiko }; 6992086f9568SStanislav Fomichev 6993d58e468bSPetar Penkov struct bpf_flow_keys { 6994d58e468bSPetar Penkov __u16 nhoff; 6995d58e468bSPetar Penkov __u16 thoff; 6996d58e468bSPetar Penkov __u16 addr_proto; /* ETH_P_* of valid addrs */ 6997d58e468bSPetar Penkov __u8 is_frag; 6998d58e468bSPetar Penkov __u8 is_first_frag; 6999d58e468bSPetar Penkov __u8 is_encap; 7000d58e468bSPetar Penkov __u8 ip_proto; 7001d58e468bSPetar Penkov __be16 n_proto; 7002d58e468bSPetar Penkov __be16 sport; 7003d58e468bSPetar Penkov __be16 dport; 7004d58e468bSPetar Penkov union { 7005d58e468bSPetar Penkov struct { 7006d58e468bSPetar Penkov __be32 ipv4_src; 7007d58e468bSPetar Penkov __be32 ipv4_dst; 7008d58e468bSPetar Penkov }; 7009d58e468bSPetar Penkov struct { 7010d58e468bSPetar Penkov __u32 ipv6_src[4]; /* in6_addr; network order */ 7011d58e468bSPetar Penkov __u32 ipv6_dst[4]; /* in6_addr; network order */ 7012d58e468bSPetar Penkov }; 7013d58e468bSPetar Penkov }; 7014086f9568SStanislav Fomichev __u32 flags; 701571c99e32SStanislav Fomichev __be32 flow_label; 7016d58e468bSPetar Penkov }; 7017d58e468bSPetar Penkov 7018838e9690SYonghong Song struct bpf_func_info { 7019d30d42e0SMartin KaFai Lau __u32 insn_off; 7020838e9690SYonghong Song __u32 type_id; 7021838e9690SYonghong Song }; 7022838e9690SYonghong Song 7023c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 7024c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 7025c454a46bSMartin KaFai Lau 7026c454a46bSMartin KaFai Lau struct bpf_line_info { 7027c454a46bSMartin KaFai Lau __u32 insn_off; 7028c454a46bSMartin KaFai Lau __u32 file_name_off; 7029c454a46bSMartin KaFai Lau __u32 line_off; 7030c454a46bSMartin KaFai Lau __u32 line_col; 7031c454a46bSMartin KaFai Lau }; 7032c454a46bSMartin KaFai Lau 7033d83525caSAlexei Starovoitov struct bpf_spin_lock { 7034d83525caSAlexei Starovoitov __u32 val; 7035d83525caSAlexei Starovoitov }; 70367b146cebSAndrey Ignatov 7037b00628b1SAlexei Starovoitov struct bpf_timer { 7038b00628b1SAlexei Starovoitov __u64 :64; 7039b00628b1SAlexei Starovoitov __u64 :64; 7040b00628b1SAlexei Starovoitov } __attribute__((aligned(8))); 7041b00628b1SAlexei Starovoitov 704297e03f52SJoanne Koong struct bpf_dynptr { 704397e03f52SJoanne Koong __u64 :64; 704497e03f52SJoanne Koong __u64 :64; 704597e03f52SJoanne Koong } __attribute__((aligned(8))); 704697e03f52SJoanne Koong 7047f0c5941fSKumar Kartikeya Dwivedi struct bpf_list_head { 7048f0c5941fSKumar Kartikeya Dwivedi __u64 :64; 7049f0c5941fSKumar Kartikeya Dwivedi __u64 :64; 7050f0c5941fSKumar Kartikeya Dwivedi } __attribute__((aligned(8))); 7051f0c5941fSKumar Kartikeya Dwivedi 7052f0c5941fSKumar Kartikeya Dwivedi struct bpf_list_node { 7053f0c5941fSKumar Kartikeya Dwivedi __u64 :64; 7054f0c5941fSKumar Kartikeya Dwivedi __u64 :64; 7055f0c5941fSKumar Kartikeya Dwivedi } __attribute__((aligned(8))); 7056f0c5941fSKumar Kartikeya Dwivedi 70579c395c1bSDave Marchevsky struct bpf_rb_root { 70589c395c1bSDave Marchevsky __u64 :64; 70599c395c1bSDave Marchevsky __u64 :64; 70609c395c1bSDave Marchevsky } __attribute__((aligned(8))); 70619c395c1bSDave Marchevsky 70629c395c1bSDave Marchevsky struct bpf_rb_node { 70639c395c1bSDave Marchevsky __u64 :64; 70649c395c1bSDave Marchevsky __u64 :64; 70659c395c1bSDave Marchevsky __u64 :64; 70669c395c1bSDave Marchevsky } __attribute__((aligned(8))); 70679c395c1bSDave Marchevsky 7068d54730b5SDave Marchevsky struct bpf_refcount { 7069d54730b5SDave Marchevsky __u32 :32; 7070d54730b5SDave Marchevsky } __attribute__((aligned(4))); 7071d54730b5SDave Marchevsky 70727b146cebSAndrey Ignatov struct bpf_sysctl { 70737b146cebSAndrey Ignatov __u32 write; /* Sysctl is being read (= 0) or written (= 1). 70747b146cebSAndrey Ignatov * Allows 1,2,4-byte read, but no write. 70757b146cebSAndrey Ignatov */ 7076e1550bfeSAndrey Ignatov __u32 file_pos; /* Sysctl file position to read from, write to. 7077e1550bfeSAndrey Ignatov * Allows 1,2,4-byte read an 4-byte write. 7078e1550bfeSAndrey Ignatov */ 70797b146cebSAndrey Ignatov }; 70807b146cebSAndrey Ignatov 70810d01da6aSStanislav Fomichev struct bpf_sockopt { 70820d01da6aSStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 70830d01da6aSStanislav Fomichev __bpf_md_ptr(void *, optval); 70840d01da6aSStanislav Fomichev __bpf_md_ptr(void *, optval_end); 70850d01da6aSStanislav Fomichev 70860d01da6aSStanislav Fomichev __s32 level; 70870d01da6aSStanislav Fomichev __s32 optname; 70880d01da6aSStanislav Fomichev __s32 optlen; 70890d01da6aSStanislav Fomichev __s32 retval; 70900d01da6aSStanislav Fomichev }; 70910d01da6aSStanislav Fomichev 7092b4490c5cSCarlos Neira struct bpf_pidns_info { 7093b4490c5cSCarlos Neira __u32 pid; 7094b4490c5cSCarlos Neira __u32 tgid; 7095b4490c5cSCarlos Neira }; 7096e9ddbb77SJakub Sitnicki 7097e9ddbb77SJakub Sitnicki /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 7098e9ddbb77SJakub Sitnicki struct bpf_sk_lookup { 70997c32e8f8SLorenz Bauer union { 7100e9ddbb77SJakub Sitnicki __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 71017c32e8f8SLorenz Bauer __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 71027c32e8f8SLorenz Bauer }; 7103e9ddbb77SJakub Sitnicki 7104e9ddbb77SJakub Sitnicki __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 7105e9ddbb77SJakub Sitnicki __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 7106e9ddbb77SJakub Sitnicki __u32 remote_ip4; /* Network byte order */ 7107e9ddbb77SJakub Sitnicki __u32 remote_ip6[4]; /* Network byte order */ 71089a69e2b3SJakub Sitnicki __be16 remote_port; /* Network byte order */ 71099a69e2b3SJakub Sitnicki __u16 :16; /* Zero padding */ 7110e9ddbb77SJakub Sitnicki __u32 local_ip4; /* Network byte order */ 7111e9ddbb77SJakub Sitnicki __u32 local_ip6[4]; /* Network byte order */ 7112e9ddbb77SJakub Sitnicki __u32 local_port; /* Host byte order */ 7113f8931565SMark Pashmfouroush __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 7114e9ddbb77SJakub Sitnicki }; 7115e9ddbb77SJakub Sitnicki 7116c4d0bfb4SAlan Maguire /* 7117c4d0bfb4SAlan Maguire * struct btf_ptr is used for typed pointer representation; the 7118c4d0bfb4SAlan Maguire * type id is used to render the pointer data as the appropriate type 7119c4d0bfb4SAlan Maguire * via the bpf_snprintf_btf() helper described above. A flags field - 7120c4d0bfb4SAlan Maguire * potentially to specify additional details about the BTF pointer 7121c4d0bfb4SAlan Maguire * (rather than its mode of display) - is included for future use. 7122c4d0bfb4SAlan Maguire * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 7123c4d0bfb4SAlan Maguire */ 7124c4d0bfb4SAlan Maguire struct btf_ptr { 7125c4d0bfb4SAlan Maguire void *ptr; 7126c4d0bfb4SAlan Maguire __u32 type_id; 7127c4d0bfb4SAlan Maguire __u32 flags; /* BTF ptr flags; unused at present. */ 7128c4d0bfb4SAlan Maguire }; 7129c4d0bfb4SAlan Maguire 7130c4d0bfb4SAlan Maguire /* 7131c4d0bfb4SAlan Maguire * Flags to control bpf_snprintf_btf() behaviour. 7132c4d0bfb4SAlan Maguire * - BTF_F_COMPACT: no formatting around type information 7133c4d0bfb4SAlan Maguire * - BTF_F_NONAME: no struct/union member names/types 7134c4d0bfb4SAlan Maguire * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 7135c4d0bfb4SAlan Maguire * equivalent to %px. 7136c4d0bfb4SAlan Maguire * - BTF_F_ZERO: show zero-valued struct/union members; they 7137c4d0bfb4SAlan Maguire * are not displayed by default 7138c4d0bfb4SAlan Maguire */ 7139c4d0bfb4SAlan Maguire enum { 7140c4d0bfb4SAlan Maguire BTF_F_COMPACT = (1ULL << 0), 7141c4d0bfb4SAlan Maguire BTF_F_NONAME = (1ULL << 1), 7142c4d0bfb4SAlan Maguire BTF_F_PTR_RAW = (1ULL << 2), 7143c4d0bfb4SAlan Maguire BTF_F_ZERO = (1ULL << 3), 7144c4d0bfb4SAlan Maguire }; 7145c4d0bfb4SAlan Maguire 714646334a0cSAlexei Starovoitov /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 714746334a0cSAlexei Starovoitov * has to be adjusted by relocations. It is emitted by llvm and passed to 714846334a0cSAlexei Starovoitov * libbpf and later to the kernel. 714946334a0cSAlexei Starovoitov */ 715046334a0cSAlexei Starovoitov enum bpf_core_relo_kind { 715146334a0cSAlexei Starovoitov BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 715246334a0cSAlexei Starovoitov BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 715346334a0cSAlexei Starovoitov BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 715446334a0cSAlexei Starovoitov BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 715546334a0cSAlexei Starovoitov BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 715646334a0cSAlexei Starovoitov BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 715746334a0cSAlexei Starovoitov BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 715846334a0cSAlexei Starovoitov BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 715946334a0cSAlexei Starovoitov BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 716046334a0cSAlexei Starovoitov BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 716146334a0cSAlexei Starovoitov BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 716246334a0cSAlexei Starovoitov BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 71633c660a5dSDaniel Müller BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */ 716446334a0cSAlexei Starovoitov }; 716546334a0cSAlexei Starovoitov 7166fbd94c7aSAlexei Starovoitov /* 7167fbd94c7aSAlexei Starovoitov * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 7168fbd94c7aSAlexei Starovoitov * and from libbpf to the kernel. 7169fbd94c7aSAlexei Starovoitov * 7170fbd94c7aSAlexei Starovoitov * CO-RE relocation captures the following data: 7171fbd94c7aSAlexei Starovoitov * - insn_off - instruction offset (in bytes) within a BPF program that needs 7172fbd94c7aSAlexei Starovoitov * its insn->imm field to be relocated with actual field info; 7173fbd94c7aSAlexei Starovoitov * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 7174fbd94c7aSAlexei Starovoitov * type or field; 7175fbd94c7aSAlexei Starovoitov * - access_str_off - offset into corresponding .BTF string section. String 7176fbd94c7aSAlexei Starovoitov * interpretation depends on specific relocation kind: 7177fbd94c7aSAlexei Starovoitov * - for field-based relocations, string encodes an accessed field using 7178fbd94c7aSAlexei Starovoitov * a sequence of field and array indices, separated by colon (:). It's 7179fbd94c7aSAlexei Starovoitov * conceptually very close to LLVM's getelementptr ([0]) instruction's 7180fbd94c7aSAlexei Starovoitov * arguments for identifying offset to a field. 7181fbd94c7aSAlexei Starovoitov * - for type-based relocations, strings is expected to be just "0"; 7182fbd94c7aSAlexei Starovoitov * - for enum value-based relocations, string contains an index of enum 7183fbd94c7aSAlexei Starovoitov * value within its enum type; 7184fbd94c7aSAlexei Starovoitov * - kind - one of enum bpf_core_relo_kind; 7185fbd94c7aSAlexei Starovoitov * 7186fbd94c7aSAlexei Starovoitov * Example: 7187fbd94c7aSAlexei Starovoitov * struct sample { 7188fbd94c7aSAlexei Starovoitov * int a; 7189fbd94c7aSAlexei Starovoitov * struct { 7190fbd94c7aSAlexei Starovoitov * int b[10]; 7191fbd94c7aSAlexei Starovoitov * }; 7192fbd94c7aSAlexei Starovoitov * }; 7193fbd94c7aSAlexei Starovoitov * 7194fbd94c7aSAlexei Starovoitov * struct sample *s = ...; 7195fbd94c7aSAlexei Starovoitov * int *x = &s->a; // encoded as "0:0" (a is field #0) 7196fbd94c7aSAlexei Starovoitov * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 7197fbd94c7aSAlexei Starovoitov * // b is field #0 inside anon struct, accessing elem #5) 7198fbd94c7aSAlexei Starovoitov * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 7199fbd94c7aSAlexei Starovoitov * 7200fbd94c7aSAlexei Starovoitov * type_id for all relocs in this example will capture BTF type id of 7201fbd94c7aSAlexei Starovoitov * `struct sample`. 7202fbd94c7aSAlexei Starovoitov * 7203fbd94c7aSAlexei Starovoitov * Such relocation is emitted when using __builtin_preserve_access_index() 7204fbd94c7aSAlexei Starovoitov * Clang built-in, passing expression that captures field address, e.g.: 7205fbd94c7aSAlexei Starovoitov * 7206fbd94c7aSAlexei Starovoitov * bpf_probe_read(&dst, sizeof(dst), 7207fbd94c7aSAlexei Starovoitov * __builtin_preserve_access_index(&src->a.b.c)); 7208fbd94c7aSAlexei Starovoitov * 7209fbd94c7aSAlexei Starovoitov * In this case Clang will emit field relocation recording necessary data to 7210fbd94c7aSAlexei Starovoitov * be able to find offset of embedded `a.b.c` field within `src` struct. 7211fbd94c7aSAlexei Starovoitov * 7212fbd94c7aSAlexei Starovoitov * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 7213fbd94c7aSAlexei Starovoitov */ 7214fbd94c7aSAlexei Starovoitov struct bpf_core_relo { 7215fbd94c7aSAlexei Starovoitov __u32 insn_off; 7216fbd94c7aSAlexei Starovoitov __u32 type_id; 7217fbd94c7aSAlexei Starovoitov __u32 access_str_off; 7218fbd94c7aSAlexei Starovoitov enum bpf_core_relo_kind kind; 7219fbd94c7aSAlexei Starovoitov }; 7220fbd94c7aSAlexei Starovoitov 7221f71f8530STero Kristo /* 7222f71f8530STero Kristo * Flags to control bpf_timer_start() behaviour. 7223f71f8530STero Kristo * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is 7224f71f8530STero Kristo * relative to current time. 7225f71f8530STero Kristo */ 7226f71f8530STero Kristo enum { 7227f71f8530STero Kristo BPF_F_TIMER_ABS = (1ULL << 0), 7228f71f8530STero Kristo }; 7229f71f8530STero Kristo 72306018e1f4SAndrii Nakryiko /* BPF numbers iterator state */ 72316018e1f4SAndrii Nakryiko struct bpf_iter_num { 72326018e1f4SAndrii Nakryiko /* opaque iterator state; having __u64 here allows to preserve correct 72336018e1f4SAndrii Nakryiko * alignment requirements in vmlinux.h, generated from BTF 72346018e1f4SAndrii Nakryiko */ 72356018e1f4SAndrii Nakryiko __u64 __opaque[1]; 72366018e1f4SAndrii Nakryiko } __attribute__((aligned(8))); 72376018e1f4SAndrii Nakryiko 7238daedfb22SAlexei Starovoitov #endif /* _UAPI__LINUX_BPF_H__ */ 7239