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; 1135e7b3020SYonghong Song }; 1145e7b3020SYonghong Song 1157799e4d9SJoe Stringer /* BPF syscall commands, see bpf(2) man-page for more details. */ 1167799e4d9SJoe Stringer /** 1177799e4d9SJoe Stringer * DOC: eBPF Syscall Preamble 1187799e4d9SJoe Stringer * 1197799e4d9SJoe Stringer * The operation to be performed by the **bpf**\ () system call is determined 1207799e4d9SJoe Stringer * by the *cmd* argument. Each operation takes an accompanying argument, 1217799e4d9SJoe Stringer * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 1227799e4d9SJoe Stringer * below). The size argument is the size of the union pointed to by *attr*. 1237799e4d9SJoe Stringer */ 1247799e4d9SJoe Stringer /** 1257799e4d9SJoe Stringer * DOC: eBPF Syscall Commands 1267799e4d9SJoe Stringer * 1277799e4d9SJoe Stringer * BPF_MAP_CREATE 1287799e4d9SJoe Stringer * Description 1297799e4d9SJoe Stringer * Create a map and return a file descriptor that refers to the 1307799e4d9SJoe Stringer * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 1317799e4d9SJoe Stringer * is automatically enabled for the new file descriptor. 1327799e4d9SJoe Stringer * 1337799e4d9SJoe Stringer * Applying **close**\ (2) to the file descriptor returned by 1347799e4d9SJoe Stringer * **BPF_MAP_CREATE** will delete the map (but see NOTES). 1357799e4d9SJoe Stringer * 1367799e4d9SJoe Stringer * Return 1377799e4d9SJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 1387799e4d9SJoe Stringer * error occurred (in which case, *errno* is set appropriately). 1397799e4d9SJoe Stringer * 1407799e4d9SJoe Stringer * BPF_MAP_LOOKUP_ELEM 1417799e4d9SJoe Stringer * Description 1427799e4d9SJoe Stringer * Look up an element with a given *key* in the map referred to 1437799e4d9SJoe Stringer * by the file descriptor *map_fd*. 1447799e4d9SJoe Stringer * 1456690523bSJoe Stringer * The *flags* argument may be specified as one of the 1466690523bSJoe Stringer * following: 1476690523bSJoe Stringer * 1486690523bSJoe Stringer * **BPF_F_LOCK** 1496690523bSJoe Stringer * Look up the value of a spin-locked map without 1506690523bSJoe Stringer * returning the lock. This must be specified if the 1516690523bSJoe Stringer * elements contain a spinlock. 1526690523bSJoe Stringer * 1537799e4d9SJoe Stringer * Return 1547799e4d9SJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 1557799e4d9SJoe Stringer * is set appropriately. 1567799e4d9SJoe Stringer * 1577799e4d9SJoe Stringer * BPF_MAP_UPDATE_ELEM 1587799e4d9SJoe Stringer * Description 1597799e4d9SJoe Stringer * Create or update an element (key/value pair) in a specified map. 1607799e4d9SJoe Stringer * 1617799e4d9SJoe Stringer * The *flags* argument should be specified as one of the 1627799e4d9SJoe Stringer * following: 1637799e4d9SJoe Stringer * 1647799e4d9SJoe Stringer * **BPF_ANY** 1657799e4d9SJoe Stringer * Create a new element or update an existing element. 1667799e4d9SJoe Stringer * **BPF_NOEXIST** 1677799e4d9SJoe Stringer * Create a new element only if it did not exist. 1687799e4d9SJoe Stringer * **BPF_EXIST** 1697799e4d9SJoe Stringer * Update an existing element. 1706690523bSJoe Stringer * **BPF_F_LOCK** 1716690523bSJoe Stringer * Update a spin_lock-ed map element. 1727799e4d9SJoe Stringer * 1737799e4d9SJoe Stringer * Return 1747799e4d9SJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 1757799e4d9SJoe Stringer * is set appropriately. 1767799e4d9SJoe Stringer * 1777799e4d9SJoe Stringer * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 1787799e4d9SJoe Stringer * **E2BIG**, **EEXIST**, or **ENOENT**. 1797799e4d9SJoe Stringer * 1807799e4d9SJoe Stringer * **E2BIG** 1817799e4d9SJoe Stringer * The number of elements in the map reached the 1827799e4d9SJoe Stringer * *max_entries* limit specified at map creation time. 1837799e4d9SJoe Stringer * **EEXIST** 1847799e4d9SJoe Stringer * If *flags* specifies **BPF_NOEXIST** and the element 1857799e4d9SJoe Stringer * with *key* already exists in the map. 1867799e4d9SJoe Stringer * **ENOENT** 1877799e4d9SJoe Stringer * If *flags* specifies **BPF_EXIST** and the element with 1887799e4d9SJoe Stringer * *key* does not exist in the map. 1897799e4d9SJoe Stringer * 1907799e4d9SJoe Stringer * BPF_MAP_DELETE_ELEM 1917799e4d9SJoe Stringer * Description 1927799e4d9SJoe Stringer * Look up and delete an element by key in a specified map. 1937799e4d9SJoe Stringer * 1947799e4d9SJoe Stringer * Return 1957799e4d9SJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 1967799e4d9SJoe Stringer * is set appropriately. 1977799e4d9SJoe Stringer * 1987799e4d9SJoe Stringer * BPF_MAP_GET_NEXT_KEY 1997799e4d9SJoe Stringer * Description 2007799e4d9SJoe Stringer * Look up an element by key in a specified map and return the key 2017799e4d9SJoe Stringer * of the next element. Can be used to iterate over all elements 2027799e4d9SJoe Stringer * in the map. 2037799e4d9SJoe Stringer * 2047799e4d9SJoe Stringer * Return 2057799e4d9SJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 2067799e4d9SJoe Stringer * is set appropriately. 2077799e4d9SJoe Stringer * 2087799e4d9SJoe Stringer * The following cases can be used to iterate over all elements of 2097799e4d9SJoe Stringer * the map: 2107799e4d9SJoe Stringer * 2117799e4d9SJoe Stringer * * If *key* is not found, the operation returns zero and sets 2127799e4d9SJoe Stringer * the *next_key* pointer to the key of the first element. 2137799e4d9SJoe Stringer * * If *key* is found, the operation returns zero and sets the 2147799e4d9SJoe Stringer * *next_key* pointer to the key of the next element. 2157799e4d9SJoe Stringer * * If *key* is the last element, returns -1 and *errno* is set 2167799e4d9SJoe Stringer * to **ENOENT**. 2177799e4d9SJoe Stringer * 2187799e4d9SJoe Stringer * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 2197799e4d9SJoe Stringer * **EINVAL** on error. 2207799e4d9SJoe Stringer * 2217799e4d9SJoe Stringer * BPF_PROG_LOAD 2227799e4d9SJoe Stringer * Description 2237799e4d9SJoe Stringer * Verify and load an eBPF program, returning a new file 2247799e4d9SJoe Stringer * descriptor associated with the program. 2257799e4d9SJoe Stringer * 2267799e4d9SJoe Stringer * Applying **close**\ (2) to the file descriptor returned by 2277799e4d9SJoe Stringer * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 2287799e4d9SJoe Stringer * 2297799e4d9SJoe Stringer * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 2307799e4d9SJoe Stringer * automatically enabled for the new file descriptor. 2317799e4d9SJoe Stringer * 2327799e4d9SJoe Stringer * Return 2337799e4d9SJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 2347799e4d9SJoe Stringer * error occurred (in which case, *errno* is set appropriately). 2357799e4d9SJoe Stringer * 236f67c9cbfSJoe Stringer * BPF_OBJ_PIN 237f67c9cbfSJoe Stringer * Description 238f67c9cbfSJoe Stringer * Pin an eBPF program or map referred by the specified *bpf_fd* 239f67c9cbfSJoe Stringer * to the provided *pathname* on the filesystem. 240f67c9cbfSJoe Stringer * 2418aacb3c8SJoe Stringer * The *pathname* argument must not contain a dot ("."). 2428aacb3c8SJoe Stringer * 2438aacb3c8SJoe Stringer * On success, *pathname* retains a reference to the eBPF object, 2448aacb3c8SJoe Stringer * preventing deallocation of the object when the original 2458aacb3c8SJoe Stringer * *bpf_fd* is closed. This allow the eBPF object to live beyond 2468aacb3c8SJoe Stringer * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 2478aacb3c8SJoe Stringer * process. 2488aacb3c8SJoe Stringer * 2498aacb3c8SJoe Stringer * Applying **unlink**\ (2) or similar calls to the *pathname* 2508aacb3c8SJoe Stringer * unpins the object from the filesystem, removing the reference. 2518aacb3c8SJoe Stringer * If no other file descriptors or filesystem nodes refer to the 2528aacb3c8SJoe Stringer * same object, it will be deallocated (see NOTES). 2538aacb3c8SJoe Stringer * 2548aacb3c8SJoe Stringer * The filesystem type for the parent directory of *pathname* must 2558aacb3c8SJoe Stringer * be **BPF_FS_MAGIC**. 2568aacb3c8SJoe Stringer * 257f67c9cbfSJoe Stringer * Return 258f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 259f67c9cbfSJoe Stringer * is set appropriately. 260f67c9cbfSJoe Stringer * 261f67c9cbfSJoe Stringer * BPF_OBJ_GET 262f67c9cbfSJoe Stringer * Description 263f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF object pinned to the 264f67c9cbfSJoe Stringer * specified *pathname*. 265f67c9cbfSJoe Stringer * 266f67c9cbfSJoe Stringer * Return 267f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 268f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 269f67c9cbfSJoe Stringer * 270f67c9cbfSJoe Stringer * BPF_PROG_ATTACH 271f67c9cbfSJoe Stringer * Description 272f67c9cbfSJoe Stringer * Attach an eBPF program to a *target_fd* at the specified 273f67c9cbfSJoe Stringer * *attach_type* hook. 274f67c9cbfSJoe Stringer * 27532e76b18SJoe Stringer * The *attach_type* specifies the eBPF attachment point to 27632e76b18SJoe Stringer * attach the program to, and must be one of *bpf_attach_type* 27732e76b18SJoe Stringer * (see below). 27832e76b18SJoe Stringer * 27932e76b18SJoe Stringer * The *attach_bpf_fd* must be a valid file descriptor for a 28032e76b18SJoe Stringer * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 28132e76b18SJoe Stringer * or sock_ops type corresponding to the specified *attach_type*. 28232e76b18SJoe Stringer * 28332e76b18SJoe Stringer * The *target_fd* must be a valid file descriptor for a kernel 28432e76b18SJoe Stringer * object which depends on the attach type of *attach_bpf_fd*: 28532e76b18SJoe Stringer * 28632e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_DEVICE**, 28732e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SKB**, 28832e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK**, 28932e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 29032e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 29132e76b18SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 29232e76b18SJoe Stringer * **BPF_PROG_TYPE_SOCK_OPS** 29332e76b18SJoe Stringer * 29432e76b18SJoe Stringer * Control Group v2 hierarchy with the eBPF controller 29532e76b18SJoe Stringer * enabled. Requires the kernel to be compiled with 29632e76b18SJoe Stringer * **CONFIG_CGROUP_BPF**. 29732e76b18SJoe Stringer * 29832e76b18SJoe Stringer * **BPF_PROG_TYPE_FLOW_DISSECTOR** 29932e76b18SJoe Stringer * 30032e76b18SJoe Stringer * Network namespace (eg /proc/self/ns/net). 30132e76b18SJoe Stringer * 30232e76b18SJoe Stringer * **BPF_PROG_TYPE_LIRC_MODE2** 30332e76b18SJoe Stringer * 30432e76b18SJoe Stringer * LIRC device path (eg /dev/lircN). Requires the kernel 30532e76b18SJoe Stringer * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 30632e76b18SJoe Stringer * 30732e76b18SJoe Stringer * **BPF_PROG_TYPE_SK_SKB**, 30832e76b18SJoe Stringer * **BPF_PROG_TYPE_SK_MSG** 30932e76b18SJoe Stringer * 31032e76b18SJoe Stringer * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 31132e76b18SJoe Stringer * 312f67c9cbfSJoe Stringer * Return 313f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 314f67c9cbfSJoe Stringer * is set appropriately. 315f67c9cbfSJoe Stringer * 316f67c9cbfSJoe Stringer * BPF_PROG_DETACH 317f67c9cbfSJoe Stringer * Description 318f67c9cbfSJoe Stringer * Detach the eBPF program associated with the *target_fd* at the 319f67c9cbfSJoe Stringer * hook specified by *attach_type*. The program must have been 320f67c9cbfSJoe Stringer * previously attached using **BPF_PROG_ATTACH**. 321f67c9cbfSJoe Stringer * 322f67c9cbfSJoe Stringer * Return 323f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 324f67c9cbfSJoe Stringer * is set appropriately. 325f67c9cbfSJoe Stringer * 326f67c9cbfSJoe Stringer * BPF_PROG_TEST_RUN 327f67c9cbfSJoe Stringer * Description 3282a3fdca4SJoe Stringer * Run the eBPF program associated with the *prog_fd* a *repeat* 3292a3fdca4SJoe Stringer * number of times against a provided program context *ctx_in* and 3302a3fdca4SJoe Stringer * data *data_in*, and return the modified program context 3312a3fdca4SJoe Stringer * *ctx_out*, *data_out* (for example, packet data), result of the 3322a3fdca4SJoe Stringer * execution *retval*, and *duration* of the test run. 333f67c9cbfSJoe Stringer * 334f3c45326SJoe Stringer * The sizes of the buffers provided as input and output 335f3c45326SJoe Stringer * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 336f3c45326SJoe Stringer * be provided in the corresponding variables *ctx_size_in*, 337f3c45326SJoe Stringer * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 338f3c45326SJoe Stringer * of these parameters are not provided (ie set to NULL), the 339f3c45326SJoe Stringer * corresponding size field must be zero. 340f3c45326SJoe Stringer * 341f3c45326SJoe Stringer * Some program types have particular requirements: 342f3c45326SJoe Stringer * 343f3c45326SJoe Stringer * **BPF_PROG_TYPE_SK_LOOKUP** 344f3c45326SJoe Stringer * *data_in* and *data_out* must be NULL. 345f3c45326SJoe Stringer * 346f3c45326SJoe Stringer * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 347f3c45326SJoe Stringer * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 348f3c45326SJoe Stringer * 349f3c45326SJoe Stringer * *ctx_out*, *data_in* and *data_out* must be NULL. 350f3c45326SJoe Stringer * *repeat* must be zero. 351f3c45326SJoe Stringer * 352e40fbbf0SUsama Arif * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 353e40fbbf0SUsama Arif * 354f67c9cbfSJoe Stringer * Return 355f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 356f67c9cbfSJoe Stringer * is set appropriately. 357f67c9cbfSJoe Stringer * 3582a3fdca4SJoe Stringer * **ENOSPC** 3592a3fdca4SJoe Stringer * Either *data_size_out* or *ctx_size_out* is too small. 3602a3fdca4SJoe Stringer * **ENOTSUPP** 3612a3fdca4SJoe Stringer * This command is not supported by the program type of 3622a3fdca4SJoe Stringer * the program referred to by *prog_fd*. 3632a3fdca4SJoe Stringer * 364f67c9cbfSJoe Stringer * BPF_PROG_GET_NEXT_ID 365f67c9cbfSJoe Stringer * Description 366f67c9cbfSJoe Stringer * Fetch the next eBPF program currently loaded into the kernel. 367f67c9cbfSJoe Stringer * 368f67c9cbfSJoe Stringer * Looks for the eBPF program with an id greater than *start_id* 369f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF programs 370f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 371f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 372f67c9cbfSJoe Stringer * 373f67c9cbfSJoe Stringer * Return 374f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 375f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 376f67c9cbfSJoe Stringer * 377f67c9cbfSJoe Stringer * BPF_MAP_GET_NEXT_ID 378f67c9cbfSJoe Stringer * Description 379f67c9cbfSJoe Stringer * Fetch the next eBPF map currently loaded into the kernel. 380f67c9cbfSJoe Stringer * 381f67c9cbfSJoe Stringer * Looks for the eBPF map with an id greater than *start_id* 382f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF maps 383f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 384f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 385f67c9cbfSJoe Stringer * 386f67c9cbfSJoe Stringer * Return 387f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 388f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 389f67c9cbfSJoe Stringer * 390f67c9cbfSJoe Stringer * BPF_PROG_GET_FD_BY_ID 391f67c9cbfSJoe Stringer * Description 392f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF program corresponding to 393f67c9cbfSJoe Stringer * *prog_id*. 394f67c9cbfSJoe Stringer * 395f67c9cbfSJoe Stringer * Return 396f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 397f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 398f67c9cbfSJoe Stringer * 399f67c9cbfSJoe Stringer * BPF_MAP_GET_FD_BY_ID 400f67c9cbfSJoe Stringer * Description 401f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF map corresponding to 402f67c9cbfSJoe Stringer * *map_id*. 403f67c9cbfSJoe Stringer * 404f67c9cbfSJoe Stringer * Return 405f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 406f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 407f67c9cbfSJoe Stringer * 408f67c9cbfSJoe Stringer * BPF_OBJ_GET_INFO_BY_FD 409f67c9cbfSJoe Stringer * Description 410f67c9cbfSJoe Stringer * Obtain information about the eBPF object corresponding to 411f67c9cbfSJoe Stringer * *bpf_fd*. 412f67c9cbfSJoe Stringer * 413f67c9cbfSJoe Stringer * Populates up to *info_len* bytes of *info*, which will be in 414f67c9cbfSJoe Stringer * one of the following formats depending on the eBPF object type 415f67c9cbfSJoe Stringer * of *bpf_fd*: 416f67c9cbfSJoe Stringer * 417f67c9cbfSJoe Stringer * * **struct bpf_prog_info** 418f67c9cbfSJoe Stringer * * **struct bpf_map_info** 419f67c9cbfSJoe Stringer * * **struct bpf_btf_info** 420f67c9cbfSJoe Stringer * * **struct bpf_link_info** 421f67c9cbfSJoe Stringer * 422f67c9cbfSJoe Stringer * Return 423f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 424f67c9cbfSJoe Stringer * is set appropriately. 425f67c9cbfSJoe Stringer * 426f67c9cbfSJoe Stringer * BPF_PROG_QUERY 427f67c9cbfSJoe Stringer * Description 428f67c9cbfSJoe Stringer * Obtain information about eBPF programs associated with the 429f67c9cbfSJoe Stringer * specified *attach_type* hook. 430f67c9cbfSJoe Stringer * 4315d999994SJoe Stringer * The *target_fd* must be a valid file descriptor for a kernel 4325d999994SJoe Stringer * object which depends on the attach type of *attach_bpf_fd*: 4335d999994SJoe Stringer * 4345d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_DEVICE**, 4355d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SKB**, 4365d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK**, 4375d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 4385d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 4395d999994SJoe Stringer * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 4405d999994SJoe Stringer * **BPF_PROG_TYPE_SOCK_OPS** 4415d999994SJoe Stringer * 4425d999994SJoe Stringer * Control Group v2 hierarchy with the eBPF controller 4435d999994SJoe Stringer * enabled. Requires the kernel to be compiled with 4445d999994SJoe Stringer * **CONFIG_CGROUP_BPF**. 4455d999994SJoe Stringer * 4465d999994SJoe Stringer * **BPF_PROG_TYPE_FLOW_DISSECTOR** 4475d999994SJoe Stringer * 4485d999994SJoe Stringer * Network namespace (eg /proc/self/ns/net). 4495d999994SJoe Stringer * 4505d999994SJoe Stringer * **BPF_PROG_TYPE_LIRC_MODE2** 4515d999994SJoe Stringer * 4525d999994SJoe Stringer * LIRC device path (eg /dev/lircN). Requires the kernel 4535d999994SJoe Stringer * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 4545d999994SJoe Stringer * 4555d999994SJoe Stringer * **BPF_PROG_QUERY** always fetches the number of programs 4565d999994SJoe Stringer * attached and the *attach_flags* which were used to attach those 4575d999994SJoe Stringer * programs. Additionally, if *prog_ids* is nonzero and the number 4585d999994SJoe Stringer * of attached programs is less than *prog_cnt*, populates 4595d999994SJoe Stringer * *prog_ids* with the eBPF program ids of the programs attached 4605d999994SJoe Stringer * at *target_fd*. 4615d999994SJoe Stringer * 4625d999994SJoe Stringer * The following flags may alter the result: 4635d999994SJoe Stringer * 4645d999994SJoe Stringer * **BPF_F_QUERY_EFFECTIVE** 4655d999994SJoe Stringer * Only return information regarding programs which are 4665d999994SJoe Stringer * currently effective at the specified *target_fd*. 4675d999994SJoe Stringer * 468f67c9cbfSJoe Stringer * Return 469f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 470f67c9cbfSJoe Stringer * is set appropriately. 471f67c9cbfSJoe Stringer * 472f67c9cbfSJoe Stringer * BPF_RAW_TRACEPOINT_OPEN 473f67c9cbfSJoe Stringer * Description 474f67c9cbfSJoe Stringer * Attach an eBPF program to a tracepoint *name* to access kernel 475f67c9cbfSJoe Stringer * internal arguments of the tracepoint in their raw form. 476f67c9cbfSJoe Stringer * 477f67c9cbfSJoe Stringer * The *prog_fd* must be a valid file descriptor associated with 478f67c9cbfSJoe Stringer * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 479f67c9cbfSJoe Stringer * 480f67c9cbfSJoe Stringer * No ABI guarantees are made about the content of tracepoint 481f67c9cbfSJoe Stringer * arguments exposed to the corresponding eBPF program. 482f67c9cbfSJoe Stringer * 483f67c9cbfSJoe Stringer * Applying **close**\ (2) to the file descriptor returned by 484f67c9cbfSJoe Stringer * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 485f67c9cbfSJoe Stringer * 486f67c9cbfSJoe Stringer * Return 487f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 488f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 489f67c9cbfSJoe Stringer * 490f67c9cbfSJoe Stringer * BPF_BTF_LOAD 491f67c9cbfSJoe Stringer * Description 492f67c9cbfSJoe Stringer * Verify and load BPF Type Format (BTF) metadata into the kernel, 493f67c9cbfSJoe Stringer * returning a new file descriptor associated with the metadata. 494f67c9cbfSJoe Stringer * BTF is described in more detail at 495f67c9cbfSJoe Stringer * https://www.kernel.org/doc/html/latest/bpf/btf.html. 496f67c9cbfSJoe Stringer * 497f67c9cbfSJoe Stringer * The *btf* parameter must point to valid memory providing 498f67c9cbfSJoe Stringer * *btf_size* bytes of BTF binary metadata. 499f67c9cbfSJoe Stringer * 500f67c9cbfSJoe Stringer * The returned file descriptor can be passed to other **bpf**\ () 501f67c9cbfSJoe Stringer * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 502f67c9cbfSJoe Stringer * associate the BTF with those objects. 503f67c9cbfSJoe Stringer * 504f67c9cbfSJoe Stringer * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 505f67c9cbfSJoe Stringer * parameters to specify a *btf_log_buf*, *btf_log_size* and 506f67c9cbfSJoe Stringer * *btf_log_level* which allow the kernel to return freeform log 507f67c9cbfSJoe Stringer * output regarding the BTF verification process. 508f67c9cbfSJoe Stringer * 509f67c9cbfSJoe Stringer * Return 510f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 511f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 512f67c9cbfSJoe Stringer * 513f67c9cbfSJoe Stringer * BPF_BTF_GET_FD_BY_ID 514f67c9cbfSJoe Stringer * Description 515f67c9cbfSJoe Stringer * Open a file descriptor for the BPF Type Format (BTF) 516f67c9cbfSJoe Stringer * corresponding to *btf_id*. 517f67c9cbfSJoe Stringer * 518f67c9cbfSJoe Stringer * Return 519f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 520f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 521f67c9cbfSJoe Stringer * 522f67c9cbfSJoe Stringer * BPF_TASK_FD_QUERY 523f67c9cbfSJoe Stringer * Description 524f67c9cbfSJoe Stringer * Obtain information about eBPF programs associated with the 525f67c9cbfSJoe Stringer * target process identified by *pid* and *fd*. 526f67c9cbfSJoe Stringer * 527f67c9cbfSJoe Stringer * If the *pid* and *fd* are associated with a tracepoint, kprobe 528f67c9cbfSJoe Stringer * or uprobe perf event, then the *prog_id* and *fd_type* will 529f67c9cbfSJoe Stringer * be populated with the eBPF program id and file descriptor type 530f67c9cbfSJoe Stringer * of type **bpf_task_fd_type**. If associated with a kprobe or 531f67c9cbfSJoe Stringer * uprobe, the *probe_offset* and *probe_addr* will also be 532f67c9cbfSJoe Stringer * populated. Optionally, if *buf* is provided, then up to 533f67c9cbfSJoe Stringer * *buf_len* bytes of *buf* will be populated with the name of 534f67c9cbfSJoe Stringer * the tracepoint, kprobe or uprobe. 535f67c9cbfSJoe Stringer * 536f67c9cbfSJoe Stringer * The resulting *prog_id* may be introspected in deeper detail 537f67c9cbfSJoe Stringer * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 538f67c9cbfSJoe Stringer * 539f67c9cbfSJoe Stringer * Return 540f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 541f67c9cbfSJoe Stringer * is set appropriately. 542f67c9cbfSJoe Stringer * 543f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_AND_DELETE_ELEM 544f67c9cbfSJoe Stringer * Description 545f67c9cbfSJoe Stringer * Look up an element with the given *key* in the map referred to 546f67c9cbfSJoe Stringer * by the file descriptor *fd*, and if found, delete the element. 547f67c9cbfSJoe Stringer * 5483e87f192SDenis Salopek * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 5493e87f192SDenis Salopek * types, the *flags* argument needs to be set to 0, but for other 5503e87f192SDenis Salopek * map types, it may be specified as: 5513e87f192SDenis Salopek * 5523e87f192SDenis Salopek * **BPF_F_LOCK** 5533e87f192SDenis Salopek * Look up and delete the value of a spin-locked map 5543e87f192SDenis Salopek * without returning the lock. This must be specified if 5553e87f192SDenis Salopek * the elements contain a spinlock. 5563e87f192SDenis Salopek * 557f67c9cbfSJoe Stringer * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 558f67c9cbfSJoe Stringer * implement this command as a "pop" operation, deleting the top 559f67c9cbfSJoe Stringer * element rather than one corresponding to *key*. 560f67c9cbfSJoe Stringer * The *key* and *key_len* parameters should be zeroed when 561f67c9cbfSJoe Stringer * issuing this operation for these map types. 562f67c9cbfSJoe Stringer * 563f67c9cbfSJoe Stringer * This command is only valid for the following map types: 564f67c9cbfSJoe Stringer * * **BPF_MAP_TYPE_QUEUE** 565f67c9cbfSJoe Stringer * * **BPF_MAP_TYPE_STACK** 5663e87f192SDenis Salopek * * **BPF_MAP_TYPE_HASH** 5673e87f192SDenis Salopek * * **BPF_MAP_TYPE_PERCPU_HASH** 5683e87f192SDenis Salopek * * **BPF_MAP_TYPE_LRU_HASH** 5693e87f192SDenis Salopek * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 570f67c9cbfSJoe Stringer * 571f67c9cbfSJoe Stringer * Return 572f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 573f67c9cbfSJoe Stringer * is set appropriately. 574f67c9cbfSJoe Stringer * 575f67c9cbfSJoe Stringer * BPF_MAP_FREEZE 576f67c9cbfSJoe Stringer * Description 577f67c9cbfSJoe Stringer * Freeze the permissions of the specified map. 578f67c9cbfSJoe Stringer * 579f67c9cbfSJoe Stringer * Write permissions may be frozen by passing zero *flags*. 580f67c9cbfSJoe Stringer * Upon success, no future syscall invocations may alter the 581f67c9cbfSJoe Stringer * map state of *map_fd*. Write operations from eBPF programs 582f67c9cbfSJoe Stringer * are still possible for a frozen map. 583f67c9cbfSJoe Stringer * 584f67c9cbfSJoe Stringer * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 585f67c9cbfSJoe Stringer * 586f67c9cbfSJoe Stringer * Return 587f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 588f67c9cbfSJoe Stringer * is set appropriately. 589f67c9cbfSJoe Stringer * 590f67c9cbfSJoe Stringer * BPF_BTF_GET_NEXT_ID 591f67c9cbfSJoe Stringer * Description 592f67c9cbfSJoe Stringer * Fetch the next BPF Type Format (BTF) object currently loaded 593f67c9cbfSJoe Stringer * into the kernel. 594f67c9cbfSJoe Stringer * 595f67c9cbfSJoe Stringer * Looks for the BTF object with an id greater than *start_id* 596f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other BTF objects 597f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 598f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 599f67c9cbfSJoe Stringer * 600f67c9cbfSJoe Stringer * Return 601f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 602f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 603f67c9cbfSJoe Stringer * 604f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_BATCH 605f67c9cbfSJoe Stringer * Description 606f67c9cbfSJoe Stringer * Iterate and fetch multiple elements in a map. 607f67c9cbfSJoe Stringer * 6080cb80454SJoe Stringer * Two opaque values are used to manage batch operations, 6090cb80454SJoe Stringer * *in_batch* and *out_batch*. Initially, *in_batch* must be set 6100cb80454SJoe Stringer * to NULL to begin the batched operation. After each subsequent 6110cb80454SJoe Stringer * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 6120cb80454SJoe Stringer * *out_batch* as the *in_batch* for the next operation to 6130cb80454SJoe Stringer * continue iteration from the current point. 6140cb80454SJoe Stringer * 6150cb80454SJoe Stringer * The *keys* and *values* are output parameters which must point 6160cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 6170cb80454SJoe Stringer * and value size of the map *map_fd*. The *keys* buffer must be 6180cb80454SJoe Stringer * of *key_size* * *count*. The *values* buffer must be of 6190cb80454SJoe Stringer * *value_size* * *count*. 6200cb80454SJoe Stringer * 6210cb80454SJoe Stringer * The *elem_flags* argument may be specified as one of the 6220cb80454SJoe Stringer * following: 6230cb80454SJoe Stringer * 6240cb80454SJoe Stringer * **BPF_F_LOCK** 6250cb80454SJoe Stringer * Look up the value of a spin-locked map without 6260cb80454SJoe Stringer * returning the lock. This must be specified if the 6270cb80454SJoe Stringer * elements contain a spinlock. 6280cb80454SJoe Stringer * 6290cb80454SJoe Stringer * On success, *count* elements from the map are copied into the 6300cb80454SJoe Stringer * user buffer, with the keys copied into *keys* and the values 6310cb80454SJoe Stringer * copied into the corresponding indices in *values*. 6320cb80454SJoe Stringer * 6330cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 6340cb80454SJoe Stringer * is set to the number of successfully processed elements. 6350cb80454SJoe Stringer * 636f67c9cbfSJoe Stringer * Return 637f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 638f67c9cbfSJoe Stringer * is set appropriately. 639f67c9cbfSJoe Stringer * 6400cb80454SJoe Stringer * May set *errno* to **ENOSPC** to indicate that *keys* or 6410cb80454SJoe Stringer * *values* is too small to dump an entire bucket during 6420cb80454SJoe Stringer * iteration of a hash-based map type. 6430cb80454SJoe Stringer * 644f67c9cbfSJoe Stringer * BPF_MAP_LOOKUP_AND_DELETE_BATCH 645f67c9cbfSJoe Stringer * Description 6460cb80454SJoe Stringer * Iterate and delete all elements in a map. 6470cb80454SJoe Stringer * 6480cb80454SJoe Stringer * This operation has the same behavior as 6490cb80454SJoe Stringer * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 6500cb80454SJoe Stringer * 6510cb80454SJoe Stringer * * Every element that is successfully returned is also deleted 6520cb80454SJoe Stringer * from the map. This is at least *count* elements. Note that 6530cb80454SJoe Stringer * *count* is both an input and an output parameter. 6540cb80454SJoe Stringer * * Upon returning with *errno* set to **EFAULT**, up to 6550cb80454SJoe Stringer * *count* elements may be deleted without returning the keys 6560cb80454SJoe Stringer * and values of the deleted elements. 657f67c9cbfSJoe Stringer * 658f67c9cbfSJoe Stringer * Return 659f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 660f67c9cbfSJoe Stringer * is set appropriately. 661f67c9cbfSJoe Stringer * 662f67c9cbfSJoe Stringer * BPF_MAP_UPDATE_BATCH 663f67c9cbfSJoe Stringer * Description 6640cb80454SJoe Stringer * Update multiple elements in a map by *key*. 6650cb80454SJoe Stringer * 6660cb80454SJoe Stringer * The *keys* and *values* are input parameters which must point 6670cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 6680cb80454SJoe Stringer * and value size of the map *map_fd*. The *keys* buffer must be 6690cb80454SJoe Stringer * of *key_size* * *count*. The *values* buffer must be of 6700cb80454SJoe Stringer * *value_size* * *count*. 6710cb80454SJoe Stringer * 6720cb80454SJoe Stringer * Each element specified in *keys* is sequentially updated to the 6730cb80454SJoe Stringer * value in the corresponding index in *values*. The *in_batch* 6740cb80454SJoe Stringer * and *out_batch* parameters are ignored and should be zeroed. 6750cb80454SJoe Stringer * 6760cb80454SJoe Stringer * The *elem_flags* argument should be specified as one of the 6770cb80454SJoe Stringer * following: 6780cb80454SJoe Stringer * 6790cb80454SJoe Stringer * **BPF_ANY** 6800cb80454SJoe Stringer * Create new elements or update a existing elements. 6810cb80454SJoe Stringer * **BPF_NOEXIST** 6820cb80454SJoe Stringer * Create new elements only if they do not exist. 6830cb80454SJoe Stringer * **BPF_EXIST** 6840cb80454SJoe Stringer * Update existing elements. 6850cb80454SJoe Stringer * **BPF_F_LOCK** 6860cb80454SJoe Stringer * Update spin_lock-ed map elements. This must be 6870cb80454SJoe Stringer * specified if the map value contains a spinlock. 6880cb80454SJoe Stringer * 6890cb80454SJoe Stringer * On success, *count* elements from the map are updated. 6900cb80454SJoe Stringer * 6910cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 6920cb80454SJoe Stringer * is set to the number of successfully processed elements. 693f67c9cbfSJoe Stringer * 694f67c9cbfSJoe Stringer * Return 695f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 696f67c9cbfSJoe Stringer * is set appropriately. 697f67c9cbfSJoe Stringer * 6980cb80454SJoe Stringer * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 6990cb80454SJoe Stringer * **E2BIG**. **E2BIG** indicates that the number of elements in 7000cb80454SJoe Stringer * the map reached the *max_entries* limit specified at map 7010cb80454SJoe Stringer * creation time. 7020cb80454SJoe Stringer * 7030cb80454SJoe Stringer * May set *errno* to one of the following error codes under 7040cb80454SJoe Stringer * specific circumstances: 7050cb80454SJoe Stringer * 7060cb80454SJoe Stringer * **EEXIST** 7070cb80454SJoe Stringer * If *flags* specifies **BPF_NOEXIST** and the element 7080cb80454SJoe Stringer * with *key* already exists in the map. 7090cb80454SJoe Stringer * **ENOENT** 7100cb80454SJoe Stringer * If *flags* specifies **BPF_EXIST** and the element with 7110cb80454SJoe Stringer * *key* does not exist in the map. 7120cb80454SJoe Stringer * 713f67c9cbfSJoe Stringer * BPF_MAP_DELETE_BATCH 714f67c9cbfSJoe Stringer * Description 7150cb80454SJoe Stringer * Delete multiple elements in a map by *key*. 7160cb80454SJoe Stringer * 7170cb80454SJoe Stringer * The *keys* parameter is an input parameter which must point 7180cb80454SJoe Stringer * to memory large enough to hold *count* items based on the key 7190cb80454SJoe Stringer * size of the map *map_fd*, that is, *key_size* * *count*. 7200cb80454SJoe Stringer * 7210cb80454SJoe Stringer * Each element specified in *keys* is sequentially deleted. The 7220cb80454SJoe Stringer * *in_batch*, *out_batch*, and *values* parameters are ignored 7230cb80454SJoe Stringer * and should be zeroed. 7240cb80454SJoe Stringer * 7250cb80454SJoe Stringer * The *elem_flags* argument may be specified as one of the 7260cb80454SJoe Stringer * following: 7270cb80454SJoe Stringer * 7280cb80454SJoe Stringer * **BPF_F_LOCK** 7290cb80454SJoe Stringer * Look up the value of a spin-locked map without 7300cb80454SJoe Stringer * returning the lock. This must be specified if the 7310cb80454SJoe Stringer * elements contain a spinlock. 7320cb80454SJoe Stringer * 7330cb80454SJoe Stringer * On success, *count* elements from the map are updated. 7340cb80454SJoe Stringer * 7350cb80454SJoe Stringer * If an error is returned and *errno* is not **EFAULT**, *count* 7360cb80454SJoe Stringer * is set to the number of successfully processed elements. If 7370cb80454SJoe Stringer * *errno* is **EFAULT**, up to *count* elements may be been 7380cb80454SJoe Stringer * deleted. 739f67c9cbfSJoe Stringer * 740f67c9cbfSJoe Stringer * Return 741f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 742f67c9cbfSJoe Stringer * is set appropriately. 743f67c9cbfSJoe Stringer * 744f67c9cbfSJoe Stringer * BPF_LINK_CREATE 745f67c9cbfSJoe Stringer * Description 746f67c9cbfSJoe Stringer * Attach an eBPF program to a *target_fd* at the specified 747f67c9cbfSJoe Stringer * *attach_type* hook and return a file descriptor handle for 748f67c9cbfSJoe Stringer * managing the link. 749f67c9cbfSJoe Stringer * 750f67c9cbfSJoe Stringer * Return 751f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 752f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 753f67c9cbfSJoe Stringer * 754f67c9cbfSJoe Stringer * BPF_LINK_UPDATE 755f67c9cbfSJoe Stringer * Description 756f67c9cbfSJoe Stringer * Update the eBPF program in the specified *link_fd* to 757f67c9cbfSJoe Stringer * *new_prog_fd*. 758f67c9cbfSJoe Stringer * 759f67c9cbfSJoe Stringer * Return 760f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 761f67c9cbfSJoe Stringer * is set appropriately. 762f67c9cbfSJoe Stringer * 763f67c9cbfSJoe Stringer * BPF_LINK_GET_FD_BY_ID 764f67c9cbfSJoe Stringer * Description 765f67c9cbfSJoe Stringer * Open a file descriptor for the eBPF Link corresponding to 766f67c9cbfSJoe Stringer * *link_id*. 767f67c9cbfSJoe Stringer * 768f67c9cbfSJoe Stringer * Return 769f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 770f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 771f67c9cbfSJoe Stringer * 772f67c9cbfSJoe Stringer * BPF_LINK_GET_NEXT_ID 773f67c9cbfSJoe Stringer * Description 774f67c9cbfSJoe Stringer * Fetch the next eBPF link currently loaded into the kernel. 775f67c9cbfSJoe Stringer * 776f67c9cbfSJoe Stringer * Looks for the eBPF link with an id greater than *start_id* 777f67c9cbfSJoe Stringer * and updates *next_id* on success. If no other eBPF links 778f67c9cbfSJoe Stringer * remain with ids higher than *start_id*, returns -1 and sets 779f67c9cbfSJoe Stringer * *errno* to **ENOENT**. 780f67c9cbfSJoe Stringer * 781f67c9cbfSJoe Stringer * Return 782f67c9cbfSJoe Stringer * Returns zero on success. On error, or when no id remains, -1 783f67c9cbfSJoe Stringer * is returned and *errno* is set appropriately. 784f67c9cbfSJoe Stringer * 785f67c9cbfSJoe Stringer * BPF_ENABLE_STATS 786f67c9cbfSJoe Stringer * Description 787f67c9cbfSJoe Stringer * Enable eBPF runtime statistics gathering. 788f67c9cbfSJoe Stringer * 789f67c9cbfSJoe Stringer * Runtime statistics gathering for the eBPF runtime is disabled 790f67c9cbfSJoe Stringer * by default to minimize the corresponding performance overhead. 791f67c9cbfSJoe Stringer * This command enables statistics globally. 792f67c9cbfSJoe Stringer * 793f67c9cbfSJoe Stringer * Multiple programs may independently enable statistics. 794f67c9cbfSJoe Stringer * After gathering the desired statistics, eBPF runtime statistics 795f67c9cbfSJoe Stringer * may be disabled again by calling **close**\ (2) for the file 796f67c9cbfSJoe Stringer * descriptor returned by this function. Statistics will only be 797f67c9cbfSJoe Stringer * disabled system-wide when all outstanding file descriptors 798f67c9cbfSJoe Stringer * returned by prior calls for this subcommand are closed. 799f67c9cbfSJoe Stringer * 800f67c9cbfSJoe Stringer * Return 801f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 802f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 803f67c9cbfSJoe Stringer * 804f67c9cbfSJoe Stringer * BPF_ITER_CREATE 805f67c9cbfSJoe Stringer * Description 806f67c9cbfSJoe Stringer * Create an iterator on top of the specified *link_fd* (as 807f67c9cbfSJoe Stringer * previously created using **BPF_LINK_CREATE**) and return a 808f67c9cbfSJoe Stringer * file descriptor that can be used to trigger the iteration. 809f67c9cbfSJoe Stringer * 810f67c9cbfSJoe Stringer * If the resulting file descriptor is pinned to the filesystem 811f67c9cbfSJoe Stringer * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 812f67c9cbfSJoe Stringer * for that path will trigger the iterator to read kernel state 813f67c9cbfSJoe Stringer * using the eBPF program attached to *link_fd*. 814f67c9cbfSJoe Stringer * 815f67c9cbfSJoe Stringer * Return 816f67c9cbfSJoe Stringer * A new file descriptor (a nonnegative integer), or -1 if an 817f67c9cbfSJoe Stringer * error occurred (in which case, *errno* is set appropriately). 818f67c9cbfSJoe Stringer * 819f67c9cbfSJoe Stringer * BPF_LINK_DETACH 820f67c9cbfSJoe Stringer * Description 821f67c9cbfSJoe Stringer * Forcefully detach the specified *link_fd* from its 822f67c9cbfSJoe Stringer * corresponding attachment point. 823f67c9cbfSJoe Stringer * 824f67c9cbfSJoe Stringer * Return 825f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 826f67c9cbfSJoe Stringer * is set appropriately. 827f67c9cbfSJoe Stringer * 828f67c9cbfSJoe Stringer * BPF_PROG_BIND_MAP 829f67c9cbfSJoe Stringer * Description 830f67c9cbfSJoe Stringer * Bind a map to the lifetime of an eBPF program. 831f67c9cbfSJoe Stringer * 832f67c9cbfSJoe Stringer * The map identified by *map_fd* is bound to the program 833f67c9cbfSJoe Stringer * identified by *prog_fd* and only released when *prog_fd* is 834f67c9cbfSJoe Stringer * released. This may be used in cases where metadata should be 835f67c9cbfSJoe Stringer * associated with a program which otherwise does not contain any 836f67c9cbfSJoe Stringer * references to the map (for example, embedded in the eBPF 837f67c9cbfSJoe Stringer * program instructions). 838f67c9cbfSJoe Stringer * 839f67c9cbfSJoe Stringer * Return 840f67c9cbfSJoe Stringer * Returns zero on success. On error, -1 is returned and *errno* 841f67c9cbfSJoe Stringer * is set appropriately. 842f67c9cbfSJoe Stringer * 8437799e4d9SJoe Stringer * NOTES 8447799e4d9SJoe Stringer * eBPF objects (maps and programs) can be shared between processes. 8458aacb3c8SJoe Stringer * 8468aacb3c8SJoe Stringer * * After **fork**\ (2), the child inherits file descriptors 8478aacb3c8SJoe Stringer * referring to the same eBPF objects. 8488aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be transferred over 8498aacb3c8SJoe Stringer * **unix**\ (7) domain sockets. 8508aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be duplicated in the 8518aacb3c8SJoe Stringer * usual way, using **dup**\ (2) and similar calls. 8528aacb3c8SJoe Stringer * * File descriptors referring to eBPF objects can be pinned to the 8538aacb3c8SJoe Stringer * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 8548aacb3c8SJoe Stringer * 8558aacb3c8SJoe Stringer * An eBPF object is deallocated only after all file descriptors referring 8568aacb3c8SJoe Stringer * to the object have been closed and no references remain pinned to the 8578aacb3c8SJoe Stringer * filesystem or attached (for example, bound to a program or device). 8587799e4d9SJoe Stringer */ 85999c55f7dSAlexei Starovoitov enum bpf_cmd { 86099c55f7dSAlexei Starovoitov BPF_MAP_CREATE, 861db20fd2bSAlexei Starovoitov BPF_MAP_LOOKUP_ELEM, 862db20fd2bSAlexei Starovoitov BPF_MAP_UPDATE_ELEM, 863db20fd2bSAlexei Starovoitov BPF_MAP_DELETE_ELEM, 864db20fd2bSAlexei Starovoitov BPF_MAP_GET_NEXT_KEY, 86509756af4SAlexei Starovoitov BPF_PROG_LOAD, 866b2197755SDaniel Borkmann BPF_OBJ_PIN, 867b2197755SDaniel Borkmann BPF_OBJ_GET, 868f4324551SDaniel Mack BPF_PROG_ATTACH, 869f4324551SDaniel Mack BPF_PROG_DETACH, 8701cf1cae9SAlexei Starovoitov BPF_PROG_TEST_RUN, 8715d67f349SAlexei Starovoitov BPF_PROG_RUN = BPF_PROG_TEST_RUN, 87234ad5580SMartin KaFai Lau BPF_PROG_GET_NEXT_ID, 87334ad5580SMartin KaFai Lau BPF_MAP_GET_NEXT_ID, 874b16d9aa4SMartin KaFai Lau BPF_PROG_GET_FD_BY_ID, 875bd5f5f4eSMartin KaFai Lau BPF_MAP_GET_FD_BY_ID, 8761e270976SMartin KaFai Lau BPF_OBJ_GET_INFO_BY_FD, 877468e2f64SAlexei Starovoitov BPF_PROG_QUERY, 878c4f6699dSAlexei Starovoitov BPF_RAW_TRACEPOINT_OPEN, 879f56a653cSMartin KaFai Lau BPF_BTF_LOAD, 88078958fcaSMartin KaFai Lau BPF_BTF_GET_FD_BY_ID, 88141bdc4b4SYonghong Song BPF_TASK_FD_QUERY, 882bd513cd0SMauricio Vasquez B BPF_MAP_LOOKUP_AND_DELETE_ELEM, 88387df15deSDaniel Borkmann BPF_MAP_FREEZE, 8841b9ed84eSQuentin Monnet BPF_BTF_GET_NEXT_ID, 885cb4d03abSBrian Vazquez BPF_MAP_LOOKUP_BATCH, 88605799638SYonghong Song BPF_MAP_LOOKUP_AND_DELETE_BATCH, 887aa2e93b8SBrian Vazquez BPF_MAP_UPDATE_BATCH, 888aa2e93b8SBrian Vazquez BPF_MAP_DELETE_BATCH, 889af6eea57SAndrii Nakryiko BPF_LINK_CREATE, 8900c991ebcSAndrii Nakryiko BPF_LINK_UPDATE, 8912d602c8cSAndrii Nakryiko BPF_LINK_GET_FD_BY_ID, 8922d602c8cSAndrii Nakryiko BPF_LINK_GET_NEXT_ID, 893d46edd67SSong Liu BPF_ENABLE_STATS, 894ac51d99bSYonghong Song BPF_ITER_CREATE, 89573b11c2aSAndrii Nakryiko BPF_LINK_DETACH, 896ef15314aSYiFei Zhu BPF_PROG_BIND_MAP, 89799c55f7dSAlexei Starovoitov }; 89899c55f7dSAlexei Starovoitov 89999c55f7dSAlexei Starovoitov enum bpf_map_type { 90099c55f7dSAlexei Starovoitov BPF_MAP_TYPE_UNSPEC, 9010f8e4bd8SAlexei Starovoitov BPF_MAP_TYPE_HASH, 90228fbcfa0SAlexei Starovoitov BPF_MAP_TYPE_ARRAY, 90304fd61abSAlexei Starovoitov BPF_MAP_TYPE_PROG_ARRAY, 904ea317b26SKaixu Xia BPF_MAP_TYPE_PERF_EVENT_ARRAY, 905824bd0ceSAlexei Starovoitov BPF_MAP_TYPE_PERCPU_HASH, 906a10423b8SAlexei Starovoitov BPF_MAP_TYPE_PERCPU_ARRAY, 907d5a3b1f6SAlexei Starovoitov BPF_MAP_TYPE_STACK_TRACE, 9084ed8ec52SMartin KaFai Lau BPF_MAP_TYPE_CGROUP_ARRAY, 90929ba732aSMartin KaFai Lau BPF_MAP_TYPE_LRU_HASH, 9108f844938SMartin KaFai Lau BPF_MAP_TYPE_LRU_PERCPU_HASH, 911b95a5c4dSDaniel Mack BPF_MAP_TYPE_LPM_TRIE, 91256f668dfSMartin KaFai Lau BPF_MAP_TYPE_ARRAY_OF_MAPS, 913bcc6b1b7SMartin KaFai Lau BPF_MAP_TYPE_HASH_OF_MAPS, 914546ac1ffSJohn Fastabend BPF_MAP_TYPE_DEVMAP, 915174a79ffSJohn Fastabend BPF_MAP_TYPE_SOCKMAP, 9166710e112SJesper Dangaard Brouer BPF_MAP_TYPE_CPUMAP, 917fbfc504aSBjörn Töpel BPF_MAP_TYPE_XSKMAP, 91881110384SJohn Fastabend BPF_MAP_TYPE_SOCKHASH, 919de9cbbaaSRoman Gushchin BPF_MAP_TYPE_CGROUP_STORAGE, 9205dc4c4b7SMartin KaFai Lau BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 921b741f163SRoman Gushchin BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 922f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_QUEUE, 923f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_STACK, 9246ac99e8fSMartin KaFai Lau BPF_MAP_TYPE_SK_STORAGE, 9256f9d451aSToke Høiland-Jørgensen BPF_MAP_TYPE_DEVMAP_HASH, 92685d33df3SMartin KaFai Lau BPF_MAP_TYPE_STRUCT_OPS, 927457f4436SAndrii Nakryiko BPF_MAP_TYPE_RINGBUF, 9288ea63684SKP Singh BPF_MAP_TYPE_INODE_STORAGE, 9294cf1bc1fSKP Singh BPF_MAP_TYPE_TASK_STORAGE, 9309330986cSJoanne Koong BPF_MAP_TYPE_BLOOM_FILTER, 931*583c1f42SDavid Vernet BPF_MAP_TYPE_USER_RINGBUF, 93299c55f7dSAlexei Starovoitov }; 93399c55f7dSAlexei Starovoitov 9346c4fc209SDaniel Borkmann /* Note that tracing related programs such as 9356c4fc209SDaniel Borkmann * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 9366c4fc209SDaniel Borkmann * are not subject to a stable API since kernel internal data 9376c4fc209SDaniel Borkmann * structures can change from release to release and may 9386c4fc209SDaniel Borkmann * therefore break existing tracing BPF programs. Tracing BPF 9396c4fc209SDaniel Borkmann * programs correspond to /a/ specific kernel which is to be 9406c4fc209SDaniel Borkmann * analyzed, and not /a/ specific kernel /and/ all future ones. 9416c4fc209SDaniel Borkmann */ 94209756af4SAlexei Starovoitov enum bpf_prog_type { 94309756af4SAlexei Starovoitov BPF_PROG_TYPE_UNSPEC, 944ddd872bcSAlexei Starovoitov BPF_PROG_TYPE_SOCKET_FILTER, 9452541517cSAlexei Starovoitov BPF_PROG_TYPE_KPROBE, 94696be4325SDaniel Borkmann BPF_PROG_TYPE_SCHED_CLS, 94794caee8cSDaniel Borkmann BPF_PROG_TYPE_SCHED_ACT, 94898b5c2c6SAlexei Starovoitov BPF_PROG_TYPE_TRACEPOINT, 9496a773a15SBrenden Blanco BPF_PROG_TYPE_XDP, 9500515e599SAlexei Starovoitov BPF_PROG_TYPE_PERF_EVENT, 9510e33661dSDaniel Mack BPF_PROG_TYPE_CGROUP_SKB, 95261023658SDavid Ahern BPF_PROG_TYPE_CGROUP_SOCK, 9533a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_IN, 9543a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_OUT, 9553a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_XMIT, 95640304b2aSLawrence Brakmo BPF_PROG_TYPE_SOCK_OPS, 957b005fd18SJohn Fastabend BPF_PROG_TYPE_SK_SKB, 958ebc614f6SRoman Gushchin BPF_PROG_TYPE_CGROUP_DEVICE, 9594f738adbSJohn Fastabend BPF_PROG_TYPE_SK_MSG, 960c4f6699dSAlexei Starovoitov BPF_PROG_TYPE_RAW_TRACEPOINT, 9614fbac77dSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 962004d4b27SMathieu Xhonneux BPF_PROG_TYPE_LWT_SEG6LOCAL, 963f4364dcfSSean Young BPF_PROG_TYPE_LIRC_MODE2, 9642dbb9b9eSMartin KaFai Lau BPF_PROG_TYPE_SK_REUSEPORT, 965d58e468bSPetar Penkov BPF_PROG_TYPE_FLOW_DISSECTOR, 9667b146cebSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SYSCTL, 9679df1c28bSMatt Mullins BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 9680d01da6aSStanislav Fomichev BPF_PROG_TYPE_CGROUP_SOCKOPT, 969f1b9509cSAlexei Starovoitov BPF_PROG_TYPE_TRACING, 97027ae7997SMartin KaFai Lau BPF_PROG_TYPE_STRUCT_OPS, 971be8704ffSAlexei Starovoitov BPF_PROG_TYPE_EXT, 972fc611f47SKP Singh BPF_PROG_TYPE_LSM, 973e9ddbb77SJakub Sitnicki BPF_PROG_TYPE_SK_LOOKUP, 97479a7f8bdSAlexei Starovoitov BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 97509756af4SAlexei Starovoitov }; 97609756af4SAlexei Starovoitov 9770e33661dSDaniel Mack enum bpf_attach_type { 9780e33661dSDaniel Mack BPF_CGROUP_INET_INGRESS, 9790e33661dSDaniel Mack BPF_CGROUP_INET_EGRESS, 98061023658SDavid Ahern BPF_CGROUP_INET_SOCK_CREATE, 98140304b2aSLawrence Brakmo BPF_CGROUP_SOCK_OPS, 982464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_PARSER, 983464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_VERDICT, 984ebc614f6SRoman Gushchin BPF_CGROUP_DEVICE, 9854f738adbSJohn Fastabend BPF_SK_MSG_VERDICT, 9864fbac77dSAndrey Ignatov BPF_CGROUP_INET4_BIND, 9874fbac77dSAndrey Ignatov BPF_CGROUP_INET6_BIND, 988d74bad4eSAndrey Ignatov BPF_CGROUP_INET4_CONNECT, 989d74bad4eSAndrey Ignatov BPF_CGROUP_INET6_CONNECT, 990aac3fc32SAndrey Ignatov BPF_CGROUP_INET4_POST_BIND, 991aac3fc32SAndrey Ignatov BPF_CGROUP_INET6_POST_BIND, 9921cedee13SAndrey Ignatov BPF_CGROUP_UDP4_SENDMSG, 9931cedee13SAndrey Ignatov BPF_CGROUP_UDP6_SENDMSG, 994f4364dcfSSean Young BPF_LIRC_MODE2, 995d58e468bSPetar Penkov BPF_FLOW_DISSECTOR, 9967b146cebSAndrey Ignatov BPF_CGROUP_SYSCTL, 997983695faSDaniel Borkmann BPF_CGROUP_UDP4_RECVMSG, 998983695faSDaniel Borkmann BPF_CGROUP_UDP6_RECVMSG, 9990d01da6aSStanislav Fomichev BPF_CGROUP_GETSOCKOPT, 10000d01da6aSStanislav Fomichev BPF_CGROUP_SETSOCKOPT, 1001f1b9509cSAlexei Starovoitov BPF_TRACE_RAW_TP, 1002fec56f58SAlexei Starovoitov BPF_TRACE_FENTRY, 1003fec56f58SAlexei Starovoitov BPF_TRACE_FEXIT, 1004ae240823SKP Singh BPF_MODIFY_RETURN, 1005fc611f47SKP Singh BPF_LSM_MAC, 100615d83c4dSYonghong Song BPF_TRACE_ITER, 10071b66d253SDaniel Borkmann BPF_CGROUP_INET4_GETPEERNAME, 10081b66d253SDaniel Borkmann BPF_CGROUP_INET6_GETPEERNAME, 10091b66d253SDaniel Borkmann BPF_CGROUP_INET4_GETSOCKNAME, 10101b66d253SDaniel Borkmann BPF_CGROUP_INET6_GETSOCKNAME, 1011fbee97feSDavid Ahern BPF_XDP_DEVMAP, 1012f5836749SStanislav Fomichev BPF_CGROUP_INET_SOCK_RELEASE, 101392164774SLorenzo Bianconi BPF_XDP_CPUMAP, 1014e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP, 1015aa8d3a71SAndrii Nakryiko BPF_XDP, 1016a7ba4558SCong Wang BPF_SK_SKB_VERDICT, 1017d5e4ddaeSKuniyuki Iwashima BPF_SK_REUSEPORT_SELECT, 1018d5e4ddaeSKuniyuki Iwashima BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 1019b89fbfbbSAndrii Nakryiko BPF_PERF_EVENT, 10200dcac272SJiri Olsa BPF_TRACE_KPROBE_MULTI, 102169fd337aSStanislav Fomichev BPF_LSM_CGROUP, 10220e33661dSDaniel Mack __MAX_BPF_ATTACH_TYPE 10230e33661dSDaniel Mack }; 10240e33661dSDaniel Mack 10250e33661dSDaniel Mack #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 10260e33661dSDaniel Mack 1027f2e10bffSAndrii Nakryiko enum bpf_link_type { 1028f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_UNSPEC = 0, 1029f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1030f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_TRACING = 2, 1031f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_CGROUP = 3, 1032de4e05caSYonghong Song BPF_LINK_TYPE_ITER = 4, 10337f045a49SJakub Sitnicki BPF_LINK_TYPE_NETNS = 5, 1034aa8d3a71SAndrii Nakryiko BPF_LINK_TYPE_XDP = 6, 1035b89fbfbbSAndrii Nakryiko BPF_LINK_TYPE_PERF_EVENT = 7, 10360dcac272SJiri Olsa BPF_LINK_TYPE_KPROBE_MULTI = 8, 1037f7e0beafSKui-Feng Lee BPF_LINK_TYPE_STRUCT_OPS = 9, 1038f2e10bffSAndrii Nakryiko 1039f2e10bffSAndrii Nakryiko MAX_BPF_LINK_TYPE, 1040f2e10bffSAndrii Nakryiko }; 1041f2e10bffSAndrii Nakryiko 1042324bda9eSAlexei Starovoitov /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1043324bda9eSAlexei Starovoitov * 1044324bda9eSAlexei Starovoitov * NONE(default): No further bpf programs allowed in the subtree. 1045324bda9eSAlexei Starovoitov * 1046324bda9eSAlexei Starovoitov * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1047324bda9eSAlexei Starovoitov * the program in this cgroup yields to sub-cgroup program. 1048324bda9eSAlexei Starovoitov * 1049324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1050324bda9eSAlexei Starovoitov * that cgroup program gets run in addition to the program in this cgroup. 1051324bda9eSAlexei Starovoitov * 1052324bda9eSAlexei Starovoitov * Only one program is allowed to be attached to a cgroup with 1053324bda9eSAlexei Starovoitov * NONE or BPF_F_ALLOW_OVERRIDE flag. 1054324bda9eSAlexei Starovoitov * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1055324bda9eSAlexei Starovoitov * release old program and attach the new one. Attach flags has to match. 1056324bda9eSAlexei Starovoitov * 1057324bda9eSAlexei Starovoitov * Multiple programs are allowed to be attached to a cgroup with 1058324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1059324bda9eSAlexei Starovoitov * (those that were attached first, run first) 1060324bda9eSAlexei Starovoitov * The programs of sub-cgroup are executed first, then programs of 1061324bda9eSAlexei Starovoitov * this cgroup and then programs of parent cgroup. 1062324bda9eSAlexei Starovoitov * When children program makes decision (like picking TCP CA or sock bind) 1063324bda9eSAlexei Starovoitov * parent program has a chance to override it. 1064324bda9eSAlexei Starovoitov * 10657dd68b32SAndrey Ignatov * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 10667dd68b32SAndrey Ignatov * programs for a cgroup. Though it's possible to replace an old program at 10677dd68b32SAndrey Ignatov * any position by also specifying BPF_F_REPLACE flag and position itself in 10687dd68b32SAndrey Ignatov * replace_bpf_fd attribute. Old program at this position will be released. 10697dd68b32SAndrey Ignatov * 1070324bda9eSAlexei Starovoitov * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1071324bda9eSAlexei Starovoitov * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1072324bda9eSAlexei Starovoitov * Ex1: 1073324bda9eSAlexei Starovoitov * cgrp1 (MULTI progs A, B) -> 1074324bda9eSAlexei Starovoitov * cgrp2 (OVERRIDE prog C) -> 1075324bda9eSAlexei Starovoitov * cgrp3 (MULTI prog D) -> 1076324bda9eSAlexei Starovoitov * cgrp4 (OVERRIDE prog E) -> 1077324bda9eSAlexei Starovoitov * cgrp5 (NONE prog F) 1078324bda9eSAlexei Starovoitov * the event in cgrp5 triggers execution of F,D,A,B in that order. 1079324bda9eSAlexei Starovoitov * if prog F is detached, the execution is E,D,A,B 1080324bda9eSAlexei Starovoitov * if prog F and D are detached, the execution is E,A,B 1081324bda9eSAlexei Starovoitov * if prog F, E and D are detached, the execution is C,A,B 1082324bda9eSAlexei Starovoitov * 1083324bda9eSAlexei Starovoitov * All eligible programs are executed regardless of return code from 1084324bda9eSAlexei Starovoitov * earlier programs. 10857f677633SAlexei Starovoitov */ 10867f677633SAlexei Starovoitov #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1087324bda9eSAlexei Starovoitov #define BPF_F_ALLOW_MULTI (1U << 1) 10887dd68b32SAndrey Ignatov #define BPF_F_REPLACE (1U << 2) 10897f677633SAlexei Starovoitov 1090e07b98d9SDavid S. Miller /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1091e07b98d9SDavid S. Miller * verifier will perform strict alignment checking as if the kernel 1092e07b98d9SDavid S. Miller * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1093e07b98d9SDavid S. Miller * and NET_IP_ALIGN defined to 2. 1094e07b98d9SDavid S. Miller */ 1095e07b98d9SDavid S. Miller #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1096e07b98d9SDavid S. Miller 1097e9ee9efcSDavid Miller /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 1098e9ee9efcSDavid Miller * verifier will allow any alignment whatsoever. On platforms 1099e9ee9efcSDavid Miller * with strict alignment requirements for loads ands stores (such 1100e9ee9efcSDavid Miller * as sparc and mips) the verifier validates that all loads and 1101e9ee9efcSDavid Miller * stores provably follow this requirement. This flag turns that 1102e9ee9efcSDavid Miller * checking and enforcement off. 1103e9ee9efcSDavid Miller * 1104e9ee9efcSDavid Miller * It is mostly used for testing when we want to validate the 1105e9ee9efcSDavid Miller * context and memory access aspects of the verifier, but because 1106e9ee9efcSDavid Miller * of an unaligned access the alignment check would trigger before 1107e9ee9efcSDavid Miller * the one we are interested in. 1108e9ee9efcSDavid Miller */ 1109e9ee9efcSDavid Miller #define BPF_F_ANY_ALIGNMENT (1U << 1) 1110e9ee9efcSDavid Miller 1111c240eff6SJiong Wang /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1112c240eff6SJiong Wang * Verifier does sub-register def/use analysis and identifies instructions whose 1113c240eff6SJiong Wang * def only matters for low 32-bit, high 32-bit is never referenced later 1114c240eff6SJiong Wang * through implicit zero extension. Therefore verifier notifies JIT back-ends 1115c240eff6SJiong Wang * that it is safe to ignore clearing high 32-bit for these instructions. This 1116c240eff6SJiong Wang * saves some back-ends a lot of code-gen. However such optimization is not 1117c240eff6SJiong Wang * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1118c240eff6SJiong Wang * hence hasn't used verifier's analysis result. But, we really want to have a 1119c240eff6SJiong Wang * way to be able to verify the correctness of the described optimization on 1120c240eff6SJiong Wang * x86_64 on which testsuites are frequently exercised. 1121c240eff6SJiong Wang * 1122c240eff6SJiong Wang * So, this flag is introduced. Once it is set, verifier will randomize high 1123c240eff6SJiong Wang * 32-bit for those instructions who has been identified as safe to ignore them. 1124c240eff6SJiong Wang * Then, if verifier is not doing correct analysis, such randomization will 1125c240eff6SJiong Wang * regress tests to expose bugs. 1126c240eff6SJiong Wang */ 1127c240eff6SJiong Wang #define BPF_F_TEST_RND_HI32 (1U << 2) 1128c240eff6SJiong Wang 112910d274e8SAlexei Starovoitov /* The verifier internal test flag. Behavior is undefined */ 113010d274e8SAlexei Starovoitov #define BPF_F_TEST_STATE_FREQ (1U << 3) 113110d274e8SAlexei Starovoitov 11321e6c62a8SAlexei Starovoitov /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 11331e6c62a8SAlexei Starovoitov * restrict map and helper usage for such programs. Sleepable BPF programs can 11341e6c62a8SAlexei Starovoitov * only be attached to hooks where kernel execution context allows sleeping. 11351e6c62a8SAlexei Starovoitov * Such programs are allowed to use helpers that may sleep like 11361e6c62a8SAlexei Starovoitov * bpf_copy_from_user(). 11371e6c62a8SAlexei Starovoitov */ 11381e6c62a8SAlexei Starovoitov #define BPF_F_SLEEPABLE (1U << 4) 11391e6c62a8SAlexei Starovoitov 1140c2f2cdbeSLorenzo Bianconi /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1141c2f2cdbeSLorenzo Bianconi * fully support xdp frags. 1142c2f2cdbeSLorenzo Bianconi */ 1143c2f2cdbeSLorenzo Bianconi #define BPF_F_XDP_HAS_FRAGS (1U << 5) 1144c2f2cdbeSLorenzo Bianconi 11450dcac272SJiri Olsa /* link_create.kprobe_multi.flags used in LINK_CREATE command for 11460dcac272SJiri Olsa * BPF_TRACE_KPROBE_MULTI attach type to create return probe. 11470dcac272SJiri Olsa */ 11480dcac272SJiri Olsa #define BPF_F_KPROBE_MULTI_RETURN (1U << 0) 11490dcac272SJiri Olsa 1150d8eca5bbSDaniel Borkmann /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 11514976b718SHao Luo * the following extensions: 1152d8eca5bbSDaniel Borkmann * 1153387544bfSAlexei Starovoitov * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1154387544bfSAlexei Starovoitov * insn[0].imm: map fd or fd_idx 11554976b718SHao Luo * insn[1].imm: 0 11564976b718SHao Luo * insn[0].off: 0 11574976b718SHao Luo * insn[1].off: 0 11584976b718SHao Luo * ldimm64 rewrite: address of map 11594976b718SHao Luo * verifier type: CONST_PTR_TO_MAP 1160d8eca5bbSDaniel Borkmann */ 1161f1a66f85SDaniel Borkmann #define BPF_PSEUDO_MAP_FD 1 1162387544bfSAlexei Starovoitov #define BPF_PSEUDO_MAP_IDX 5 1163387544bfSAlexei Starovoitov 1164387544bfSAlexei Starovoitov /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1165387544bfSAlexei Starovoitov * insn[0].imm: map fd or fd_idx 11664976b718SHao Luo * insn[1].imm: offset into value 11674976b718SHao Luo * insn[0].off: 0 11684976b718SHao Luo * insn[1].off: 0 11694976b718SHao Luo * ldimm64 rewrite: address of map[0]+offset 11704976b718SHao Luo * verifier type: PTR_TO_MAP_VALUE 11714976b718SHao Luo */ 1172d8eca5bbSDaniel Borkmann #define BPF_PSEUDO_MAP_VALUE 2 1173387544bfSAlexei Starovoitov #define BPF_PSEUDO_MAP_IDX_VALUE 6 1174387544bfSAlexei Starovoitov 11754976b718SHao Luo /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 11764976b718SHao Luo * insn[0].imm: kernel btd id of VAR 11774976b718SHao Luo * insn[1].imm: 0 11784976b718SHao Luo * insn[0].off: 0 11794976b718SHao Luo * insn[1].off: 0 11804976b718SHao Luo * ldimm64 rewrite: address of the kernel variable 11814976b718SHao Luo * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 11824976b718SHao Luo * is struct/union. 11834976b718SHao Luo */ 11844976b718SHao Luo #define BPF_PSEUDO_BTF_ID 3 118569c087baSYonghong Song /* insn[0].src_reg: BPF_PSEUDO_FUNC 118669c087baSYonghong Song * insn[0].imm: insn offset to the func 118769c087baSYonghong Song * insn[1].imm: 0 118869c087baSYonghong Song * insn[0].off: 0 118969c087baSYonghong Song * insn[1].off: 0 119069c087baSYonghong Song * ldimm64 rewrite: address of the function 119169c087baSYonghong Song * verifier type: PTR_TO_FUNC. 119269c087baSYonghong Song */ 119369c087baSYonghong Song #define BPF_PSEUDO_FUNC 4 1194f1a66f85SDaniel Borkmann 1195cc8b0b92SAlexei Starovoitov /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1196cc8b0b92SAlexei Starovoitov * offset to another bpf function 1197cc8b0b92SAlexei Starovoitov */ 1198cc8b0b92SAlexei Starovoitov #define BPF_PSEUDO_CALL 1 1199e6ac2450SMartin KaFai Lau /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1200e6ac2450SMartin KaFai Lau * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1201e6ac2450SMartin KaFai Lau */ 1202e6ac2450SMartin KaFai Lau #define BPF_PSEUDO_KFUNC_CALL 2 1203cc8b0b92SAlexei Starovoitov 12043274f520SAlexei Starovoitov /* flags for BPF_MAP_UPDATE_ELEM command */ 12051aae4bddSAndrii Nakryiko enum { 12061aae4bddSAndrii Nakryiko BPF_ANY = 0, /* create new element or update existing */ 12071aae4bddSAndrii Nakryiko BPF_NOEXIST = 1, /* create new element if it didn't exist */ 12081aae4bddSAndrii Nakryiko BPF_EXIST = 2, /* update existing element */ 12091aae4bddSAndrii Nakryiko BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 12101aae4bddSAndrii Nakryiko }; 12113274f520SAlexei Starovoitov 121296eabe7aSMartin KaFai Lau /* flags for BPF_MAP_CREATE command */ 12131aae4bddSAndrii Nakryiko enum { 12141aae4bddSAndrii Nakryiko BPF_F_NO_PREALLOC = (1U << 0), 121529ba732aSMartin KaFai Lau /* Instead of having one common LRU list in the 12168f844938SMartin KaFai Lau * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 121729ba732aSMartin KaFai Lau * which can scale and perform better. 121829ba732aSMartin KaFai Lau * Note, the LRU nodes (including free nodes) cannot be moved 121929ba732aSMartin KaFai Lau * across different LRU lists. 122029ba732aSMartin KaFai Lau */ 12211aae4bddSAndrii Nakryiko BPF_F_NO_COMMON_LRU = (1U << 1), 122296eabe7aSMartin KaFai Lau /* Specify numa node during map creation */ 12231aae4bddSAndrii Nakryiko BPF_F_NUMA_NODE = (1U << 2), 1224cb4d2b3fSMartin KaFai Lau 1225591fe988SDaniel Borkmann /* Flags for accessing BPF object from syscall side. */ 12261aae4bddSAndrii Nakryiko BPF_F_RDONLY = (1U << 3), 12271aae4bddSAndrii Nakryiko BPF_F_WRONLY = (1U << 4), 12286e71b04aSChenbo Feng 1229615755a7SSong Liu /* Flag for stack_map, store build_id+offset instead of pointer */ 12301aae4bddSAndrii Nakryiko BPF_F_STACK_BUILD_ID = (1U << 5), 1231615755a7SSong Liu 123296b3b6c9SLorenz Bauer /* Zero-initialize hash function seed. This should only be used for testing. */ 12331aae4bddSAndrii Nakryiko BPF_F_ZERO_SEED = (1U << 6), 123496b3b6c9SLorenz Bauer 1235591fe988SDaniel Borkmann /* Flags for accessing BPF object from program side. */ 12361aae4bddSAndrii Nakryiko BPF_F_RDONLY_PROG = (1U << 7), 12371aae4bddSAndrii Nakryiko BPF_F_WRONLY_PROG = (1U << 8), 1238591fe988SDaniel Borkmann 12398f51dfc7SStanislav Fomichev /* Clone map from listener for newly accepted socket */ 12401aae4bddSAndrii Nakryiko BPF_F_CLONE = (1U << 9), 12418f51dfc7SStanislav Fomichev 1242fc970227SAndrii Nakryiko /* Enable memory-mapping BPF map */ 12431aae4bddSAndrii Nakryiko BPF_F_MMAPABLE = (1U << 10), 1244792cacccSSong Liu 1245792cacccSSong Liu /* Share perf_event among processes */ 1246792cacccSSong Liu BPF_F_PRESERVE_ELEMS = (1U << 11), 12474a8f87e6SDaniel Borkmann 12484a8f87e6SDaniel Borkmann /* Create a map that is suitable to be an inner map with dynamic max entries */ 12494a8f87e6SDaniel Borkmann BPF_F_INNER_MAP = (1U << 12), 12501aae4bddSAndrii Nakryiko }; 1251fc970227SAndrii Nakryiko 1252f5bfcd95SAndrey Ignatov /* Flags for BPF_PROG_QUERY. */ 1253f5bfcd95SAndrey Ignatov 1254f5bfcd95SAndrey Ignatov /* Query effective (directly attached + inherited from ancestor cgroups) 1255f5bfcd95SAndrey Ignatov * programs that will be executed for events within a cgroup. 1256f5bfcd95SAndrey Ignatov * attach_flags with this flag are returned only for directly attached programs. 1257f5bfcd95SAndrey Ignatov */ 12582f183360SLorenz Bauer #define BPF_F_QUERY_EFFECTIVE (1U << 0) 12592f183360SLorenz Bauer 12601b4d60ecSSong Liu /* Flags for BPF_PROG_TEST_RUN */ 12611b4d60ecSSong Liu 12621b4d60ecSSong Liu /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 12631b4d60ecSSong Liu #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1264b530e9e1SToke Høiland-Jørgensen /* If set, XDP frames will be transmitted after processing */ 1265b530e9e1SToke Høiland-Jørgensen #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 12661b4d60ecSSong Liu 1267d46edd67SSong Liu /* type for BPF_ENABLE_STATS */ 1268d46edd67SSong Liu enum bpf_stats_type { 1269d46edd67SSong Liu /* enabled run_time_ns and run_cnt */ 1270d46edd67SSong Liu BPF_STATS_RUN_TIME = 0, 1271d46edd67SSong Liu }; 1272d46edd67SSong Liu 1273615755a7SSong Liu enum bpf_stack_build_id_status { 1274615755a7SSong Liu /* user space need an empty entry to identify end of a trace */ 1275615755a7SSong Liu BPF_STACK_BUILD_ID_EMPTY = 0, 1276615755a7SSong Liu /* with valid build_id and offset */ 1277615755a7SSong Liu BPF_STACK_BUILD_ID_VALID = 1, 1278615755a7SSong Liu /* couldn't get build_id, fallback to ip */ 1279615755a7SSong Liu BPF_STACK_BUILD_ID_IP = 2, 1280615755a7SSong Liu }; 1281615755a7SSong Liu 1282615755a7SSong Liu #define BPF_BUILD_ID_SIZE 20 1283615755a7SSong Liu struct bpf_stack_build_id { 1284615755a7SSong Liu __s32 status; 1285615755a7SSong Liu unsigned char build_id[BPF_BUILD_ID_SIZE]; 1286615755a7SSong Liu union { 1287615755a7SSong Liu __u64 offset; 1288615755a7SSong Liu __u64 ip; 1289615755a7SSong Liu }; 1290615755a7SSong Liu }; 1291615755a7SSong Liu 12921aae4bddSAndrii Nakryiko #define BPF_OBJ_NAME_LEN 16U 12931aae4bddSAndrii Nakryiko 129499c55f7dSAlexei Starovoitov union bpf_attr { 129599c55f7dSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_CREATE command */ 129699c55f7dSAlexei Starovoitov __u32 map_type; /* one of enum bpf_map_type */ 129799c55f7dSAlexei Starovoitov __u32 key_size; /* size of key in bytes */ 129899c55f7dSAlexei Starovoitov __u32 value_size; /* size of value in bytes */ 129999c55f7dSAlexei Starovoitov __u32 max_entries; /* max number of entries in a map */ 130096eabe7aSMartin KaFai Lau __u32 map_flags; /* BPF_MAP_CREATE related 130196eabe7aSMartin KaFai Lau * flags defined above. 130296eabe7aSMartin KaFai Lau */ 130356f668dfSMartin KaFai Lau __u32 inner_map_fd; /* fd pointing to the inner map */ 130496eabe7aSMartin KaFai Lau __u32 numa_node; /* numa node (effective only if 130596eabe7aSMartin KaFai Lau * BPF_F_NUMA_NODE is set). 130696eabe7aSMartin KaFai Lau */ 1307067cae47SMartin KaFai Lau char map_name[BPF_OBJ_NAME_LEN]; 1308a3884572SJakub Kicinski __u32 map_ifindex; /* ifindex of netdev to create on */ 1309a26ca7c9SMartin KaFai Lau __u32 btf_fd; /* fd pointing to a BTF type data */ 13109b2cf328SMartin KaFai Lau __u32 btf_key_type_id; /* BTF type_id of the key */ 13119b2cf328SMartin KaFai Lau __u32 btf_value_type_id; /* BTF type_id of the value */ 131285d33df3SMartin KaFai Lau __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 131385d33df3SMartin KaFai Lau * struct stored as the 131485d33df3SMartin KaFai Lau * map value 131585d33df3SMartin KaFai Lau */ 13169330986cSJoanne Koong /* Any per-map-type extra fields 13179330986cSJoanne Koong * 13189330986cSJoanne Koong * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 13199330986cSJoanne Koong * number of hash functions (if 0, the bloom filter will default 13209330986cSJoanne Koong * to using 5 hash functions). 13219330986cSJoanne Koong */ 13229330986cSJoanne Koong __u64 map_extra; 132399c55f7dSAlexei Starovoitov }; 1324db20fd2bSAlexei Starovoitov 1325db20fd2bSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1326db20fd2bSAlexei Starovoitov __u32 map_fd; 1327db20fd2bSAlexei Starovoitov __aligned_u64 key; 1328db20fd2bSAlexei Starovoitov union { 1329db20fd2bSAlexei Starovoitov __aligned_u64 value; 1330db20fd2bSAlexei Starovoitov __aligned_u64 next_key; 1331db20fd2bSAlexei Starovoitov }; 13323274f520SAlexei Starovoitov __u64 flags; 1333db20fd2bSAlexei Starovoitov }; 133409756af4SAlexei Starovoitov 1335cb4d03abSBrian Vazquez struct { /* struct used by BPF_MAP_*_BATCH commands */ 1336cb4d03abSBrian Vazquez __aligned_u64 in_batch; /* start batch, 1337cb4d03abSBrian Vazquez * NULL to start from beginning 1338cb4d03abSBrian Vazquez */ 1339cb4d03abSBrian Vazquez __aligned_u64 out_batch; /* output: next start batch */ 1340cb4d03abSBrian Vazquez __aligned_u64 keys; 1341cb4d03abSBrian Vazquez __aligned_u64 values; 1342cb4d03abSBrian Vazquez __u32 count; /* input/output: 1343cb4d03abSBrian Vazquez * input: # of key/value 1344cb4d03abSBrian Vazquez * elements 1345cb4d03abSBrian Vazquez * output: # of filled elements 1346cb4d03abSBrian Vazquez */ 1347cb4d03abSBrian Vazquez __u32 map_fd; 1348cb4d03abSBrian Vazquez __u64 elem_flags; 1349cb4d03abSBrian Vazquez __u64 flags; 1350cb4d03abSBrian Vazquez } batch; 1351cb4d03abSBrian Vazquez 135209756af4SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_LOAD command */ 135309756af4SAlexei Starovoitov __u32 prog_type; /* one of enum bpf_prog_type */ 135409756af4SAlexei Starovoitov __u32 insn_cnt; 135509756af4SAlexei Starovoitov __aligned_u64 insns; 135609756af4SAlexei Starovoitov __aligned_u64 license; 1357cbd35700SAlexei Starovoitov __u32 log_level; /* verbosity level of verifier */ 1358cbd35700SAlexei Starovoitov __u32 log_size; /* size of user buffer */ 1359cbd35700SAlexei Starovoitov __aligned_u64 log_buf; /* user supplied buffer */ 13606c4fc209SDaniel Borkmann __u32 kern_version; /* not used */ 1361e07b98d9SDavid S. Miller __u32 prog_flags; 1362067cae47SMartin KaFai Lau char prog_name[BPF_OBJ_NAME_LEN]; 13631f6f4cb7SJakub Kicinski __u32 prog_ifindex; /* ifindex of netdev to prep for */ 13645e43f899SAndrey Ignatov /* For some prog types expected attach type must be known at 13655e43f899SAndrey Ignatov * load time to verify attach type specific parts of prog 13665e43f899SAndrey Ignatov * (context accesses, allowed helpers, etc). 13675e43f899SAndrey Ignatov */ 13685e43f899SAndrey Ignatov __u32 expected_attach_type; 1369838e9690SYonghong Song __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1370838e9690SYonghong Song __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1371838e9690SYonghong Song __aligned_u64 func_info; /* func info */ 1372838e9690SYonghong Song __u32 func_info_cnt; /* number of bpf_func_info records */ 1373c454a46bSMartin KaFai Lau __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1374c454a46bSMartin KaFai Lau __aligned_u64 line_info; /* line info */ 1375c454a46bSMartin KaFai Lau __u32 line_info_cnt; /* number of bpf_line_info records */ 1376ccfe29ebSAlexei Starovoitov __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1377290248a5SAndrii Nakryiko union { 1378290248a5SAndrii Nakryiko /* valid prog_fd to attach to bpf prog */ 1379290248a5SAndrii Nakryiko __u32 attach_prog_fd; 1380290248a5SAndrii Nakryiko /* or valid module BTF object fd or 0 to attach to vmlinux */ 1381290248a5SAndrii Nakryiko __u32 attach_btf_obj_fd; 1382290248a5SAndrii Nakryiko }; 1383fbd94c7aSAlexei Starovoitov __u32 core_relo_cnt; /* number of bpf_core_relo */ 1384387544bfSAlexei Starovoitov __aligned_u64 fd_array; /* array of FDs */ 1385fbd94c7aSAlexei Starovoitov __aligned_u64 core_relos; 1386fbd94c7aSAlexei Starovoitov __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 138709756af4SAlexei Starovoitov }; 1388b2197755SDaniel Borkmann 1389b2197755SDaniel Borkmann struct { /* anonymous struct used by BPF_OBJ_* commands */ 1390b2197755SDaniel Borkmann __aligned_u64 pathname; 1391b2197755SDaniel Borkmann __u32 bpf_fd; 13926e71b04aSChenbo Feng __u32 file_flags; 1393b2197755SDaniel Borkmann }; 1394f4324551SDaniel Mack 1395f4324551SDaniel Mack struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1396f4324551SDaniel Mack __u32 target_fd; /* container object to attach to */ 1397f4324551SDaniel Mack __u32 attach_bpf_fd; /* eBPF program to attach */ 1398f4324551SDaniel Mack __u32 attach_type; 13997f677633SAlexei Starovoitov __u32 attach_flags; 14007dd68b32SAndrey Ignatov __u32 replace_bpf_fd; /* previously attached eBPF 14017dd68b32SAndrey Ignatov * program to replace if 14027dd68b32SAndrey Ignatov * BPF_F_REPLACE is used 14037dd68b32SAndrey Ignatov */ 1404f4324551SDaniel Mack }; 14051cf1cae9SAlexei Starovoitov 14061cf1cae9SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 14071cf1cae9SAlexei Starovoitov __u32 prog_fd; 14081cf1cae9SAlexei Starovoitov __u32 retval; 1409b5a36b1eSLorenz Bauer __u32 data_size_in; /* input: len of data_in */ 1410b5a36b1eSLorenz Bauer __u32 data_size_out; /* input/output: len of data_out 1411b5a36b1eSLorenz Bauer * returns ENOSPC if data_out 1412b5a36b1eSLorenz Bauer * is too small. 1413b5a36b1eSLorenz Bauer */ 14141cf1cae9SAlexei Starovoitov __aligned_u64 data_in; 14151cf1cae9SAlexei Starovoitov __aligned_u64 data_out; 14161cf1cae9SAlexei Starovoitov __u32 repeat; 14171cf1cae9SAlexei Starovoitov __u32 duration; 1418b0b9395dSStanislav Fomichev __u32 ctx_size_in; /* input: len of ctx_in */ 1419b0b9395dSStanislav Fomichev __u32 ctx_size_out; /* input/output: len of ctx_out 1420b0b9395dSStanislav Fomichev * returns ENOSPC if ctx_out 1421b0b9395dSStanislav Fomichev * is too small. 1422b0b9395dSStanislav Fomichev */ 1423b0b9395dSStanislav Fomichev __aligned_u64 ctx_in; 1424b0b9395dSStanislav Fomichev __aligned_u64 ctx_out; 14251b4d60ecSSong Liu __u32 flags; 14261b4d60ecSSong Liu __u32 cpu; 1427b530e9e1SToke Høiland-Jørgensen __u32 batch_size; 14281cf1cae9SAlexei Starovoitov } test; 142934ad5580SMartin KaFai Lau 1430b16d9aa4SMartin KaFai Lau struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1431b16d9aa4SMartin KaFai Lau union { 143234ad5580SMartin KaFai Lau __u32 start_id; 1433b16d9aa4SMartin KaFai Lau __u32 prog_id; 1434bd5f5f4eSMartin KaFai Lau __u32 map_id; 143578958fcaSMartin KaFai Lau __u32 btf_id; 1436a3b80e10SAndrii Nakryiko __u32 link_id; 1437b16d9aa4SMartin KaFai Lau }; 143834ad5580SMartin KaFai Lau __u32 next_id; 14396e71b04aSChenbo Feng __u32 open_flags; 144034ad5580SMartin KaFai Lau }; 14411e270976SMartin KaFai Lau 14421e270976SMartin KaFai Lau struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 14431e270976SMartin KaFai Lau __u32 bpf_fd; 14441e270976SMartin KaFai Lau __u32 info_len; 14451e270976SMartin KaFai Lau __aligned_u64 info; 14461e270976SMartin KaFai Lau } info; 1447468e2f64SAlexei Starovoitov 1448468e2f64SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1449468e2f64SAlexei Starovoitov __u32 target_fd; /* container object to query */ 1450468e2f64SAlexei Starovoitov __u32 attach_type; 1451468e2f64SAlexei Starovoitov __u32 query_flags; 1452468e2f64SAlexei Starovoitov __u32 attach_flags; 1453468e2f64SAlexei Starovoitov __aligned_u64 prog_ids; 1454468e2f64SAlexei Starovoitov __u32 prog_cnt; 1455b79c9fc9SStanislav Fomichev __aligned_u64 prog_attach_flags; /* output: per-program attach_flags */ 1456468e2f64SAlexei Starovoitov } query; 1457c4f6699dSAlexei Starovoitov 1458af6eea57SAndrii Nakryiko struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1459c4f6699dSAlexei Starovoitov __u64 name; 1460c4f6699dSAlexei Starovoitov __u32 prog_fd; 1461c4f6699dSAlexei Starovoitov } raw_tracepoint; 1462f56a653cSMartin KaFai Lau 1463f56a653cSMartin KaFai Lau struct { /* anonymous struct for BPF_BTF_LOAD */ 1464f56a653cSMartin KaFai Lau __aligned_u64 btf; 1465f56a653cSMartin KaFai Lau __aligned_u64 btf_log_buf; 1466f56a653cSMartin KaFai Lau __u32 btf_size; 1467f56a653cSMartin KaFai Lau __u32 btf_log_size; 1468f56a653cSMartin KaFai Lau __u32 btf_log_level; 1469f56a653cSMartin KaFai Lau }; 147041bdc4b4SYonghong Song 147141bdc4b4SYonghong Song struct { 147241bdc4b4SYonghong Song __u32 pid; /* input: pid */ 147341bdc4b4SYonghong Song __u32 fd; /* input: fd */ 147441bdc4b4SYonghong Song __u32 flags; /* input: flags */ 147541bdc4b4SYonghong Song __u32 buf_len; /* input/output: buf len */ 147641bdc4b4SYonghong Song __aligned_u64 buf; /* input/output: 147741bdc4b4SYonghong Song * tp_name for tracepoint 147841bdc4b4SYonghong Song * symbol for kprobe 147941bdc4b4SYonghong Song * filename for uprobe 148041bdc4b4SYonghong Song */ 148141bdc4b4SYonghong Song __u32 prog_id; /* output: prod_id */ 148241bdc4b4SYonghong Song __u32 fd_type; /* output: BPF_FD_TYPE_* */ 148341bdc4b4SYonghong Song __u64 probe_offset; /* output: probe_offset */ 148441bdc4b4SYonghong Song __u64 probe_addr; /* output: probe_addr */ 148541bdc4b4SYonghong Song } task_fd_query; 1486af6eea57SAndrii Nakryiko 1487af6eea57SAndrii Nakryiko struct { /* struct used by BPF_LINK_CREATE command */ 1488af6eea57SAndrii Nakryiko __u32 prog_fd; /* eBPF program to attach */ 1489aa8d3a71SAndrii Nakryiko union { 1490af6eea57SAndrii Nakryiko __u32 target_fd; /* object to attach to */ 1491aa8d3a71SAndrii Nakryiko __u32 target_ifindex; /* target ifindex */ 1492aa8d3a71SAndrii Nakryiko }; 1493af6eea57SAndrii Nakryiko __u32 attach_type; /* attach type */ 1494af6eea57SAndrii Nakryiko __u32 flags; /* extra flags */ 14954a1e7c0cSToke Høiland-Jørgensen union { 14964a1e7c0cSToke Høiland-Jørgensen __u32 target_btf_id; /* btf_id of target to attach to */ 14974a1e7c0cSToke Høiland-Jørgensen struct { 14985e7b3020SYonghong Song __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 14995e7b3020SYonghong Song __u32 iter_info_len; /* iter_info length */ 15004a1e7c0cSToke Høiland-Jørgensen }; 150182e6b1eeSAndrii Nakryiko struct { 150282e6b1eeSAndrii Nakryiko /* black box user-provided value passed through 150382e6b1eeSAndrii Nakryiko * to BPF program at the execution time and 150482e6b1eeSAndrii Nakryiko * accessible through bpf_get_attach_cookie() BPF helper 150582e6b1eeSAndrii Nakryiko */ 150682e6b1eeSAndrii Nakryiko __u64 bpf_cookie; 150782e6b1eeSAndrii Nakryiko } perf_event; 15080dcac272SJiri Olsa struct { 15090dcac272SJiri Olsa __u32 flags; 15100dcac272SJiri Olsa __u32 cnt; 15110dcac272SJiri Olsa __aligned_u64 syms; 15120dcac272SJiri Olsa __aligned_u64 addrs; 1513ca74823cSJiri Olsa __aligned_u64 cookies; 15140dcac272SJiri Olsa } kprobe_multi; 15152fcc8241SKui-Feng Lee struct { 15162fcc8241SKui-Feng Lee /* this is overlaid with the target_btf_id above. */ 15172fcc8241SKui-Feng Lee __u32 target_btf_id; 15182fcc8241SKui-Feng Lee /* black box user-provided value passed through 15192fcc8241SKui-Feng Lee * to BPF program at the execution time and 15202fcc8241SKui-Feng Lee * accessible through bpf_get_attach_cookie() BPF helper 15212fcc8241SKui-Feng Lee */ 15222fcc8241SKui-Feng Lee __u64 cookie; 15232fcc8241SKui-Feng Lee } tracing; 15244a1e7c0cSToke Høiland-Jørgensen }; 1525af6eea57SAndrii Nakryiko } link_create; 15260c991ebcSAndrii Nakryiko 15270c991ebcSAndrii Nakryiko struct { /* struct used by BPF_LINK_UPDATE command */ 15280c991ebcSAndrii Nakryiko __u32 link_fd; /* link fd */ 15290c991ebcSAndrii Nakryiko /* new program fd to update link with */ 15300c991ebcSAndrii Nakryiko __u32 new_prog_fd; 15310c991ebcSAndrii Nakryiko __u32 flags; /* extra flags */ 15320c991ebcSAndrii Nakryiko /* expected link's program fd; is specified only if 15330c991ebcSAndrii Nakryiko * BPF_F_REPLACE flag is set in flags */ 15340c991ebcSAndrii Nakryiko __u32 old_prog_fd; 15350c991ebcSAndrii Nakryiko } link_update; 15360c991ebcSAndrii Nakryiko 153773b11c2aSAndrii Nakryiko struct { 153873b11c2aSAndrii Nakryiko __u32 link_fd; 153973b11c2aSAndrii Nakryiko } link_detach; 154073b11c2aSAndrii Nakryiko 1541d46edd67SSong Liu struct { /* struct used by BPF_ENABLE_STATS command */ 1542d46edd67SSong Liu __u32 type; 1543d46edd67SSong Liu } enable_stats; 1544d46edd67SSong Liu 1545ac51d99bSYonghong Song struct { /* struct used by BPF_ITER_CREATE command */ 1546ac51d99bSYonghong Song __u32 link_fd; 1547ac51d99bSYonghong Song __u32 flags; 1548ac51d99bSYonghong Song } iter_create; 1549ac51d99bSYonghong Song 1550ef15314aSYiFei Zhu struct { /* struct used by BPF_PROG_BIND_MAP command */ 1551ef15314aSYiFei Zhu __u32 prog_fd; 1552ef15314aSYiFei Zhu __u32 map_fd; 1553ef15314aSYiFei Zhu __u32 flags; /* extra flags */ 1554ef15314aSYiFei Zhu } prog_bind_map; 1555ef15314aSYiFei Zhu 155699c55f7dSAlexei Starovoitov } __attribute__((aligned(8))); 155799c55f7dSAlexei Starovoitov 155856a092c8SQuentin Monnet /* The description below is an attempt at providing documentation to eBPF 155956a092c8SQuentin Monnet * developers about the multiple available eBPF helper functions. It can be 156056a092c8SQuentin Monnet * parsed and used to produce a manual page. The workflow is the following, 156156a092c8SQuentin Monnet * and requires the rst2man utility: 1562ebb676daSThomas Graf * 1563923a932cSJoe Stringer * $ ./scripts/bpf_doc.py \ 156456a092c8SQuentin Monnet * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 156556a092c8SQuentin Monnet * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 156656a092c8SQuentin Monnet * $ man /tmp/bpf-helpers.7 1567ebb676daSThomas Graf * 156856a092c8SQuentin Monnet * Note that in order to produce this external documentation, some RST 156956a092c8SQuentin Monnet * formatting is used in the descriptions to get "bold" and "italics" in 157056a092c8SQuentin Monnet * manual pages. Also note that the few trailing white spaces are 157156a092c8SQuentin Monnet * intentional, removing them would break paragraphs for rst2man. 1572ebb676daSThomas Graf * 157356a092c8SQuentin Monnet * Start of BPF helper function descriptions: 1574ad4a5223SQuentin Monnet * 1575ad4a5223SQuentin Monnet * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1576ad4a5223SQuentin Monnet * Description 1577ad4a5223SQuentin Monnet * Perform a lookup in *map* for an entry associated to *key*. 1578ad4a5223SQuentin Monnet * Return 1579ad4a5223SQuentin Monnet * Map value associated to *key*, or **NULL** if no entry was 1580ad4a5223SQuentin Monnet * found. 1581ad4a5223SQuentin Monnet * 1582bdb7b79bSAndrii Nakryiko * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1583ad4a5223SQuentin Monnet * Description 1584ad4a5223SQuentin Monnet * Add or update the value of the entry associated to *key* in 1585ad4a5223SQuentin Monnet * *map* with *value*. *flags* is one of: 1586ad4a5223SQuentin Monnet * 1587ad4a5223SQuentin Monnet * **BPF_NOEXIST** 1588ad4a5223SQuentin Monnet * The entry for *key* must not exist in the map. 1589ad4a5223SQuentin Monnet * **BPF_EXIST** 1590ad4a5223SQuentin Monnet * The entry for *key* must already exist in the map. 1591ad4a5223SQuentin Monnet * **BPF_ANY** 1592ad4a5223SQuentin Monnet * No condition on the existence of the entry for *key*. 1593ad4a5223SQuentin Monnet * 1594ad4a5223SQuentin Monnet * Flag value **BPF_NOEXIST** cannot be used for maps of types 1595ad4a5223SQuentin Monnet * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1596ad4a5223SQuentin Monnet * elements always exist), the helper would return an error. 1597ad4a5223SQuentin Monnet * Return 1598ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1599ad4a5223SQuentin Monnet * 1600bdb7b79bSAndrii Nakryiko * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1601ad4a5223SQuentin Monnet * Description 1602ad4a5223SQuentin Monnet * Delete entry with *key* from *map*. 1603ad4a5223SQuentin Monnet * Return 1604ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1605ad4a5223SQuentin Monnet * 1606bdb7b79bSAndrii Nakryiko * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1607ad4a5223SQuentin Monnet * Description 1608ad4a5223SQuentin Monnet * For tracing programs, safely attempt to read *size* bytes from 16096ae08ae3SDaniel Borkmann * kernel space address *unsafe_ptr* and store the data in *dst*. 16106ae08ae3SDaniel Borkmann * 1611ab8d7809SQuentin Monnet * Generally, use **bpf_probe_read_user**\ () or 1612ab8d7809SQuentin Monnet * **bpf_probe_read_kernel**\ () instead. 1613ad4a5223SQuentin Monnet * Return 1614ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1615ad4a5223SQuentin Monnet * 1616ad4a5223SQuentin Monnet * u64 bpf_ktime_get_ns(void) 1617ad4a5223SQuentin Monnet * Description 1618ad4a5223SQuentin Monnet * Return the time elapsed since system boot, in nanoseconds. 161971d19214SMaciej Żenczykowski * Does not include time the system was suspended. 1620ab8d7809SQuentin Monnet * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1621ad4a5223SQuentin Monnet * Return 1622ad4a5223SQuentin Monnet * Current *ktime*. 1623ad4a5223SQuentin Monnet * 1624bdb7b79bSAndrii Nakryiko * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1625ad4a5223SQuentin Monnet * Description 1626ad4a5223SQuentin Monnet * This helper is a "printk()-like" facility for debugging. It 1627ad4a5223SQuentin Monnet * prints a message defined by format *fmt* (of size *fmt_size*) 1628ad4a5223SQuentin Monnet * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 1629ad4a5223SQuentin Monnet * available. It can take up to three additional **u64** 1630ad4a5223SQuentin Monnet * arguments (as an eBPF helpers, the total number of arguments is 1631ad4a5223SQuentin Monnet * limited to five). 1632ad4a5223SQuentin Monnet * 1633ad4a5223SQuentin Monnet * Each time the helper is called, it appends a line to the trace. 163455c33dfbSPeter Wu * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 163555c33dfbSPeter Wu * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 1636ad4a5223SQuentin Monnet * The format of the trace is customizable, and the exact output 1637ad4a5223SQuentin Monnet * one will get depends on the options set in 1638ad4a5223SQuentin Monnet * *\/sys/kernel/debug/tracing/trace_options* (see also the 1639ad4a5223SQuentin Monnet * *README* file under the same directory). However, it usually 1640ad4a5223SQuentin Monnet * defaults to something like: 1641ad4a5223SQuentin Monnet * 1642ad4a5223SQuentin Monnet * :: 1643ad4a5223SQuentin Monnet * 1644ad4a5223SQuentin Monnet * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1645ad4a5223SQuentin Monnet * 1646ad4a5223SQuentin Monnet * In the above: 1647ad4a5223SQuentin Monnet * 1648ad4a5223SQuentin Monnet * * ``telnet`` is the name of the current task. 1649ad4a5223SQuentin Monnet * * ``470`` is the PID of the current task. 1650ad4a5223SQuentin Monnet * * ``001`` is the CPU number on which the task is 1651ad4a5223SQuentin Monnet * running. 1652ad4a5223SQuentin Monnet * * In ``.N..``, each character refers to a set of 1653ad4a5223SQuentin Monnet * options (whether irqs are enabled, scheduling 1654ad4a5223SQuentin Monnet * options, whether hard/softirqs are running, level of 1655ad4a5223SQuentin Monnet * preempt_disabled respectively). **N** means that 1656ad4a5223SQuentin Monnet * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1657ad4a5223SQuentin Monnet * are set. 1658ad4a5223SQuentin Monnet * * ``419421.045894`` is a timestamp. 1659ad4a5223SQuentin Monnet * * ``0x00000001`` is a fake value used by BPF for the 1660ad4a5223SQuentin Monnet * instruction pointer register. 1661ad4a5223SQuentin Monnet * * ``<formatted msg>`` is the message formatted with 1662ad4a5223SQuentin Monnet * *fmt*. 1663ad4a5223SQuentin Monnet * 1664ad4a5223SQuentin Monnet * The conversion specifiers supported by *fmt* are similar, but 1665ad4a5223SQuentin Monnet * more limited than for printk(). They are **%d**, **%i**, 1666ad4a5223SQuentin Monnet * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1667ad4a5223SQuentin Monnet * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1668ad4a5223SQuentin Monnet * of field, padding with zeroes, etc.) is available, and the 1669ad4a5223SQuentin Monnet * helper will return **-EINVAL** (but print nothing) if it 1670ad4a5223SQuentin Monnet * encounters an unknown specifier. 1671ad4a5223SQuentin Monnet * 1672ad4a5223SQuentin Monnet * Also, note that **bpf_trace_printk**\ () is slow, and should 1673ad4a5223SQuentin Monnet * only be used for debugging purposes. For this reason, a notice 1674b16fc097STobias Klauser * block (spanning several lines) is printed to kernel logs and 1675ad4a5223SQuentin Monnet * states that the helper should not be used "for production use" 1676ad4a5223SQuentin Monnet * the first time this helper is used (or more precisely, when 1677ad4a5223SQuentin Monnet * **trace_printk**\ () buffers are allocated). For passing values 1678ad4a5223SQuentin Monnet * to user space, perf events should be preferred. 1679ad4a5223SQuentin Monnet * Return 1680ad4a5223SQuentin Monnet * The number of bytes written to the buffer, or a negative error 1681ad4a5223SQuentin Monnet * in case of failure. 1682ad4a5223SQuentin Monnet * 16831fdd08beSQuentin Monnet * u32 bpf_get_prandom_u32(void) 16841fdd08beSQuentin Monnet * Description 16851fdd08beSQuentin Monnet * Get a pseudo-random number. 16861fdd08beSQuentin Monnet * 16871fdd08beSQuentin Monnet * From a security point of view, this helper uses its own 16881fdd08beSQuentin Monnet * pseudo-random internal state, and cannot be used to infer the 16891fdd08beSQuentin Monnet * seed of other random functions in the kernel. However, it is 16901fdd08beSQuentin Monnet * essential to note that the generator used by the helper is not 16911fdd08beSQuentin Monnet * cryptographically secure. 16921fdd08beSQuentin Monnet * Return 16931fdd08beSQuentin Monnet * A random 32-bit unsigned value. 16941fdd08beSQuentin Monnet * 16951fdd08beSQuentin Monnet * u32 bpf_get_smp_processor_id(void) 16961fdd08beSQuentin Monnet * Description 16971fdd08beSQuentin Monnet * Get the SMP (symmetric multiprocessing) processor id. Note that 169833656275SMatteo Croce * all programs run with migration disabled, which means that the 16991fdd08beSQuentin Monnet * SMP processor id is stable during all the execution of the 17001fdd08beSQuentin Monnet * program. 17011fdd08beSQuentin Monnet * Return 17021fdd08beSQuentin Monnet * The SMP id of the processor running the program. 17031fdd08beSQuentin Monnet * 1704bdb7b79bSAndrii Nakryiko * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1705ad4a5223SQuentin Monnet * Description 1706ad4a5223SQuentin Monnet * Store *len* bytes from address *from* into the packet 1707ad4a5223SQuentin Monnet * associated to *skb*, at *offset*. *flags* are a combination of 1708ad4a5223SQuentin Monnet * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1709ad4a5223SQuentin Monnet * checksum for the packet after storing the bytes) and 1710ad4a5223SQuentin Monnet * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1711ad4a5223SQuentin Monnet * **->swhash** and *skb*\ **->l4hash** to 0). 1712ad4a5223SQuentin Monnet * 171332e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1714ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1715ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1716ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1717ad4a5223SQuentin Monnet * direct packet access. 1718ad4a5223SQuentin Monnet * Return 1719ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1720ad4a5223SQuentin Monnet * 1721bdb7b79bSAndrii Nakryiko * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1722ad4a5223SQuentin Monnet * Description 1723ad4a5223SQuentin Monnet * Recompute the layer 3 (e.g. IP) checksum for the packet 1724ad4a5223SQuentin Monnet * associated to *skb*. Computation is incremental, so the helper 1725ad4a5223SQuentin Monnet * must know the former value of the header field that was 1726ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 1727ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored in *size*. 1728ad4a5223SQuentin Monnet * Alternatively, it is possible to store the difference between 1729ad4a5223SQuentin Monnet * the previous and the new values of the header field in *to*, by 1730ad4a5223SQuentin Monnet * setting *from* and *size* to 0. For both methods, *offset* 1731ad4a5223SQuentin Monnet * indicates the location of the IP checksum within the packet. 1732ad4a5223SQuentin Monnet * 1733ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 1734ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 1735ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 1736ad4a5223SQuentin Monnet * checksum to update. 1737ad4a5223SQuentin Monnet * 173832e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1739ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1740ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1741ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1742ad4a5223SQuentin Monnet * direct packet access. 1743ad4a5223SQuentin Monnet * Return 1744ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1745ad4a5223SQuentin Monnet * 1746bdb7b79bSAndrii Nakryiko * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1747ad4a5223SQuentin Monnet * Description 1748ad4a5223SQuentin Monnet * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1749ad4a5223SQuentin Monnet * packet associated to *skb*. Computation is incremental, so the 1750ad4a5223SQuentin Monnet * helper must know the former value of the header field that was 1751ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 1752ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored on the lowest 1753ad4a5223SQuentin Monnet * four bits of *flags*. Alternatively, it is possible to store 1754ad4a5223SQuentin Monnet * the difference between the previous and the new values of the 1755ad4a5223SQuentin Monnet * header field in *to*, by setting *from* and the four lowest 1756ad4a5223SQuentin Monnet * bits of *flags* to 0. For both methods, *offset* indicates the 1757ad4a5223SQuentin Monnet * location of the IP checksum within the packet. In addition to 1758ad4a5223SQuentin Monnet * the size of the field, *flags* can be added (bitwise OR) actual 1759ad4a5223SQuentin Monnet * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1760ad4a5223SQuentin Monnet * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1761ad4a5223SQuentin Monnet * for updates resulting in a null checksum the value is set to 1762ad4a5223SQuentin Monnet * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1763ad4a5223SQuentin Monnet * the checksum is to be computed against a pseudo-header. 1764ad4a5223SQuentin Monnet * 1765ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 1766ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 1767ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 1768ad4a5223SQuentin Monnet * checksum to update. 1769ad4a5223SQuentin Monnet * 177032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1771ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1772ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1773ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1774ad4a5223SQuentin Monnet * direct packet access. 1775ad4a5223SQuentin Monnet * Return 1776ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1777ad4a5223SQuentin Monnet * 1778bdb7b79bSAndrii Nakryiko * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 1779ad4a5223SQuentin Monnet * Description 1780ad4a5223SQuentin Monnet * This special helper is used to trigger a "tail call", or in 1781ad4a5223SQuentin Monnet * other words, to jump into another eBPF program. The same stack 1782ad4a5223SQuentin Monnet * frame is used (but values on stack and in registers for the 1783ad4a5223SQuentin Monnet * caller are not accessible to the callee). This mechanism allows 1784ad4a5223SQuentin Monnet * for program chaining, either for raising the maximum number of 1785ad4a5223SQuentin Monnet * available eBPF instructions, or to execute given programs in 1786ad4a5223SQuentin Monnet * conditional blocks. For security reasons, there is an upper 1787ad4a5223SQuentin Monnet * limit to the number of successive tail calls that can be 1788ad4a5223SQuentin Monnet * performed. 1789ad4a5223SQuentin Monnet * 1790ad4a5223SQuentin Monnet * Upon call of this helper, the program attempts to jump into a 1791ad4a5223SQuentin Monnet * program referenced at index *index* in *prog_array_map*, a 1792ad4a5223SQuentin Monnet * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 1793ad4a5223SQuentin Monnet * *ctx*, a pointer to the context. 1794ad4a5223SQuentin Monnet * 1795ad4a5223SQuentin Monnet * If the call succeeds, the kernel immediately runs the first 1796ad4a5223SQuentin Monnet * instruction of the new program. This is not a function call, 1797ad4a5223SQuentin Monnet * and it never returns to the previous program. If the call 1798ad4a5223SQuentin Monnet * fails, then the helper has no effect, and the caller continues 1799ad4a5223SQuentin Monnet * to run its subsequent instructions. A call can fail if the 1800ad4a5223SQuentin Monnet * destination program for the jump does not exist (i.e. *index* 1801ad4a5223SQuentin Monnet * is superior to the number of entries in *prog_array_map*), or 1802ad4a5223SQuentin Monnet * if the maximum number of tail calls has been reached for this 1803ad4a5223SQuentin Monnet * chain of programs. This limit is defined in the kernel by the 1804ad4a5223SQuentin Monnet * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 1805ebf7f6f0STiezhu Yang * which is currently set to 33. 1806ad4a5223SQuentin Monnet * Return 1807ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1808ad4a5223SQuentin Monnet * 1809bdb7b79bSAndrii Nakryiko * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 1810ad4a5223SQuentin Monnet * Description 1811ad4a5223SQuentin Monnet * Clone and redirect the packet associated to *skb* to another 1812ad4a5223SQuentin Monnet * net device of index *ifindex*. Both ingress and egress 1813ad4a5223SQuentin Monnet * interfaces can be used for redirection. The **BPF_F_INGRESS** 1814ad4a5223SQuentin Monnet * value in *flags* is used to make the distinction (ingress path 1815ad4a5223SQuentin Monnet * is selected if the flag is present, egress path otherwise). 1816ad4a5223SQuentin Monnet * This is the only flag supported for now. 1817ad4a5223SQuentin Monnet * 1818ad4a5223SQuentin Monnet * In comparison with **bpf_redirect**\ () helper, 1819ad4a5223SQuentin Monnet * **bpf_clone_redirect**\ () has the associated cost of 1820ad4a5223SQuentin Monnet * duplicating the packet buffer, but this can be executed out of 1821ad4a5223SQuentin Monnet * the eBPF program. Conversely, **bpf_redirect**\ () is more 1822ad4a5223SQuentin Monnet * efficient, but it is handled through an action code where the 1823ad4a5223SQuentin Monnet * redirection happens only after the eBPF program has returned. 1824ad4a5223SQuentin Monnet * 182532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1826ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1827ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 1828ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 1829ad4a5223SQuentin Monnet * direct packet access. 1830ad4a5223SQuentin Monnet * Return 1831ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 1832c456dec4SQuentin Monnet * 1833c456dec4SQuentin Monnet * u64 bpf_get_current_pid_tgid(void) 1834e40fbbf0SUsama Arif * Description 1835e40fbbf0SUsama Arif * Get the current pid and tgid. 1836c456dec4SQuentin Monnet * Return 1837c456dec4SQuentin Monnet * A 64-bit integer containing the current tgid and pid, and 1838c456dec4SQuentin Monnet * created as such: 1839c456dec4SQuentin Monnet * *current_task*\ **->tgid << 32 \|** 1840c456dec4SQuentin Monnet * *current_task*\ **->pid**. 1841c456dec4SQuentin Monnet * 1842c456dec4SQuentin Monnet * u64 bpf_get_current_uid_gid(void) 1843e40fbbf0SUsama Arif * Description 1844e40fbbf0SUsama Arif * Get the current uid and gid. 1845c456dec4SQuentin Monnet * Return 1846c456dec4SQuentin Monnet * A 64-bit integer containing the current GID and UID, and 1847c456dec4SQuentin Monnet * created as such: *current_gid* **<< 32 \|** *current_uid*. 1848c456dec4SQuentin Monnet * 1849bdb7b79bSAndrii Nakryiko * long bpf_get_current_comm(void *buf, u32 size_of_buf) 1850c456dec4SQuentin Monnet * Description 1851c456dec4SQuentin Monnet * Copy the **comm** attribute of the current task into *buf* of 1852c456dec4SQuentin Monnet * *size_of_buf*. The **comm** attribute contains the name of 1853c456dec4SQuentin Monnet * the executable (excluding the path) for the current task. The 1854c456dec4SQuentin Monnet * *size_of_buf* must be strictly positive. On success, the 1855c456dec4SQuentin Monnet * helper makes sure that the *buf* is NUL-terminated. On failure, 1856c456dec4SQuentin Monnet * it is filled with zeroes. 1857c456dec4SQuentin Monnet * Return 1858c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1859c456dec4SQuentin Monnet * 18601fdd08beSQuentin Monnet * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 18611fdd08beSQuentin Monnet * Description 18621fdd08beSQuentin Monnet * Retrieve the classid for the current task, i.e. for the net_cls 18631fdd08beSQuentin Monnet * cgroup to which *skb* belongs. 18641fdd08beSQuentin Monnet * 18651fdd08beSQuentin Monnet * This helper can be used on TC egress path, but not on ingress. 18661fdd08beSQuentin Monnet * 18671fdd08beSQuentin Monnet * The net_cls cgroup provides an interface to tag network packets 18681fdd08beSQuentin Monnet * based on a user-provided identifier for all traffic coming from 18691fdd08beSQuentin Monnet * the tasks belonging to the related cgroup. See also the related 18701fdd08beSQuentin Monnet * kernel documentation, available from the Linux sources in file 1871da82c92fSMauro Carvalho Chehab * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 18721fdd08beSQuentin Monnet * 18731fdd08beSQuentin Monnet * The Linux kernel has two versions for cgroups: there are 18741fdd08beSQuentin Monnet * cgroups v1 and cgroups v2. Both are available to users, who can 18751fdd08beSQuentin Monnet * use a mixture of them, but note that the net_cls cgroup is for 18761fdd08beSQuentin Monnet * cgroup v1 only. This makes it incompatible with BPF programs 18771fdd08beSQuentin Monnet * run on cgroups, which is a cgroup-v2-only feature (a socket can 18781fdd08beSQuentin Monnet * only hold data for one version of cgroups at a time). 18791fdd08beSQuentin Monnet * 18801fdd08beSQuentin Monnet * This helper is only available is the kernel was compiled with 18811fdd08beSQuentin Monnet * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 18821fdd08beSQuentin Monnet * "**y**" or to "**m**". 18831fdd08beSQuentin Monnet * Return 18841fdd08beSQuentin Monnet * The classid, or 0 for the default unconfigured classid. 18851fdd08beSQuentin Monnet * 1886bdb7b79bSAndrii Nakryiko * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1887c456dec4SQuentin Monnet * Description 1888c456dec4SQuentin Monnet * Push a *vlan_tci* (VLAN tag control information) of protocol 1889c456dec4SQuentin Monnet * *vlan_proto* to the packet associated to *skb*, then update 1890c456dec4SQuentin Monnet * the checksum. Note that if *vlan_proto* is different from 1891c456dec4SQuentin Monnet * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1892c456dec4SQuentin Monnet * be **ETH_P_8021Q**. 1893c456dec4SQuentin Monnet * 189432e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1895c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1896c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 1897c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 1898c456dec4SQuentin Monnet * direct packet access. 1899c456dec4SQuentin Monnet * Return 1900c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1901c456dec4SQuentin Monnet * 1902bdb7b79bSAndrii Nakryiko * long bpf_skb_vlan_pop(struct sk_buff *skb) 1903c456dec4SQuentin Monnet * Description 1904c456dec4SQuentin Monnet * Pop a VLAN header from the packet associated to *skb*. 1905c456dec4SQuentin Monnet * 190632e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1907c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1908c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 1909c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 1910c456dec4SQuentin Monnet * direct packet access. 1911c456dec4SQuentin Monnet * Return 1912c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1913c456dec4SQuentin Monnet * 1914bdb7b79bSAndrii Nakryiko * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1915c456dec4SQuentin Monnet * Description 1916c456dec4SQuentin Monnet * Get tunnel metadata. This helper takes a pointer *key* to an 1917c456dec4SQuentin Monnet * empty **struct bpf_tunnel_key** of **size**, that will be 1918c456dec4SQuentin Monnet * filled with tunnel metadata for the packet associated to *skb*. 1919c456dec4SQuentin Monnet * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1920c456dec4SQuentin Monnet * indicates that the tunnel is based on IPv6 protocol instead of 1921c456dec4SQuentin Monnet * IPv4. 1922c456dec4SQuentin Monnet * 1923c456dec4SQuentin Monnet * The **struct bpf_tunnel_key** is an object that generalizes the 1924c456dec4SQuentin Monnet * principal parameters used by various tunneling protocols into a 1925c456dec4SQuentin Monnet * single struct. This way, it can be used to easily make a 1926c456dec4SQuentin Monnet * decision based on the contents of the encapsulation header, 1927c456dec4SQuentin Monnet * "summarized" in this struct. In particular, it holds the IP 1928c456dec4SQuentin Monnet * address of the remote end (IPv4 or IPv6, depending on the case) 1929c456dec4SQuentin Monnet * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1930c456dec4SQuentin Monnet * this struct exposes the *key*\ **->tunnel_id**, which is 1931c456dec4SQuentin Monnet * generally mapped to a VNI (Virtual Network Identifier), making 1932c456dec4SQuentin Monnet * it programmable together with the **bpf_skb_set_tunnel_key**\ 1933c456dec4SQuentin Monnet * () helper. 1934c456dec4SQuentin Monnet * 1935c456dec4SQuentin Monnet * Let's imagine that the following code is part of a program 1936c456dec4SQuentin Monnet * attached to the TC ingress interface, on one end of a GRE 1937c456dec4SQuentin Monnet * tunnel, and is supposed to filter out all messages coming from 1938c456dec4SQuentin Monnet * remote ends with IPv4 address other than 10.0.0.1: 1939c456dec4SQuentin Monnet * 1940c456dec4SQuentin Monnet * :: 1941c456dec4SQuentin Monnet * 1942c456dec4SQuentin Monnet * int ret; 1943c456dec4SQuentin Monnet * struct bpf_tunnel_key key = {}; 1944c456dec4SQuentin Monnet * 1945c456dec4SQuentin Monnet * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1946c456dec4SQuentin Monnet * if (ret < 0) 1947c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 1948c456dec4SQuentin Monnet * 1949c456dec4SQuentin Monnet * if (key.remote_ipv4 != 0x0a000001) 1950c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 1951c456dec4SQuentin Monnet * 1952c456dec4SQuentin Monnet * return TC_ACT_OK; // accept packet 1953c456dec4SQuentin Monnet * 1954c456dec4SQuentin Monnet * This interface can also be used with all encapsulation devices 1955c456dec4SQuentin Monnet * that can operate in "collect metadata" mode: instead of having 1956c456dec4SQuentin Monnet * one network device per specific configuration, the "collect 1957c456dec4SQuentin Monnet * metadata" mode only requires a single device where the 1958c456dec4SQuentin Monnet * configuration can be extracted from this helper. 1959c456dec4SQuentin Monnet * 1960c456dec4SQuentin Monnet * This can be used together with various tunnels such as VXLan, 1961c456dec4SQuentin Monnet * Geneve, GRE or IP in IP (IPIP). 1962c456dec4SQuentin Monnet * Return 1963c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1964c456dec4SQuentin Monnet * 1965bdb7b79bSAndrii Nakryiko * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1966c456dec4SQuentin Monnet * Description 1967c456dec4SQuentin Monnet * Populate tunnel metadata for packet associated to *skb.* The 1968c456dec4SQuentin Monnet * tunnel metadata is set to the contents of *key*, of *size*. The 1969c456dec4SQuentin Monnet * *flags* can be set to a combination of the following values: 1970c456dec4SQuentin Monnet * 1971c456dec4SQuentin Monnet * **BPF_F_TUNINFO_IPV6** 1972c456dec4SQuentin Monnet * Indicate that the tunnel is based on IPv6 protocol 1973c456dec4SQuentin Monnet * instead of IPv4. 1974c456dec4SQuentin Monnet * **BPF_F_ZERO_CSUM_TX** 1975c456dec4SQuentin Monnet * For IPv4 packets, add a flag to tunnel metadata 1976c456dec4SQuentin Monnet * indicating that checksum computation should be skipped 1977c456dec4SQuentin Monnet * and checksum set to zeroes. 1978c456dec4SQuentin Monnet * **BPF_F_DONT_FRAGMENT** 1979c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that the 1980c456dec4SQuentin Monnet * packet should not be fragmented. 1981c456dec4SQuentin Monnet * **BPF_F_SEQ_NUMBER** 1982c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that a 1983c456dec4SQuentin Monnet * sequence number should be added to tunnel header before 1984c456dec4SQuentin Monnet * sending the packet. This flag was added for GRE 1985c456dec4SQuentin Monnet * encapsulation, but might be used with other protocols 1986c456dec4SQuentin Monnet * as well in the future. 1987c456dec4SQuentin Monnet * 1988c456dec4SQuentin Monnet * Here is a typical usage on the transmit path: 1989c456dec4SQuentin Monnet * 1990c456dec4SQuentin Monnet * :: 1991c456dec4SQuentin Monnet * 1992c456dec4SQuentin Monnet * struct bpf_tunnel_key key; 1993c456dec4SQuentin Monnet * populate key ... 1994c456dec4SQuentin Monnet * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1995c456dec4SQuentin Monnet * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1996c456dec4SQuentin Monnet * 1997c456dec4SQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_key**\ () 1998c456dec4SQuentin Monnet * helper for additional information. 1999c456dec4SQuentin Monnet * Return 2000c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 2001c456dec4SQuentin Monnet * 2002c6b5fb86SQuentin Monnet * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 2003c6b5fb86SQuentin Monnet * Description 2004c6b5fb86SQuentin Monnet * Read the value of a perf event counter. This helper relies on a 2005c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 2006c6b5fb86SQuentin Monnet * the perf event counter is selected when *map* is updated with 2007c6b5fb86SQuentin Monnet * perf event file descriptors. The *map* is an array whose size 2008c6b5fb86SQuentin Monnet * is the number of available CPUs, and each cell contains a value 2009c6b5fb86SQuentin Monnet * relative to one CPU. The value to retrieve is indicated by 2010c6b5fb86SQuentin Monnet * *flags*, that contains the index of the CPU to look up, masked 2011c6b5fb86SQuentin Monnet * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2012c6b5fb86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 2013c6b5fb86SQuentin Monnet * current CPU should be retrieved. 2014c6b5fb86SQuentin Monnet * 2015c6b5fb86SQuentin Monnet * Note that before Linux 4.13, only hardware perf event can be 2016c6b5fb86SQuentin Monnet * retrieved. 2017c6b5fb86SQuentin Monnet * 2018c6b5fb86SQuentin Monnet * Also, be aware that the newer helper 2019c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () is recommended over 20203bd5a09bSQuentin Monnet * **bpf_perf_event_read**\ () in general. The latter has some ABI 2021c6b5fb86SQuentin Monnet * quirks where error and counter value are used as a return code 2022c6b5fb86SQuentin Monnet * (which is wrong to do since ranges may overlap). This issue is 20233bd5a09bSQuentin Monnet * fixed with **bpf_perf_event_read_value**\ (), which at the same 20243bd5a09bSQuentin Monnet * time provides more features over the **bpf_perf_event_read**\ 20253bd5a09bSQuentin Monnet * () interface. Please refer to the description of 2026c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () for details. 2027c6b5fb86SQuentin Monnet * Return 2028c6b5fb86SQuentin Monnet * The value of the perf event counter read from the map, or a 2029c6b5fb86SQuentin Monnet * negative error code in case of failure. 2030c6b5fb86SQuentin Monnet * 2031bdb7b79bSAndrii Nakryiko * long bpf_redirect(u32 ifindex, u64 flags) 2032c456dec4SQuentin Monnet * Description 2033c456dec4SQuentin Monnet * Redirect the packet to another net device of index *ifindex*. 2034c456dec4SQuentin Monnet * This helper is somewhat similar to **bpf_clone_redirect**\ 2035c456dec4SQuentin Monnet * (), except that the packet is not cloned, which provides 2036c456dec4SQuentin Monnet * increased performance. 2037c456dec4SQuentin Monnet * 2038c456dec4SQuentin Monnet * Except for XDP, both ingress and egress interfaces can be used 2039c456dec4SQuentin Monnet * for redirection. The **BPF_F_INGRESS** value in *flags* is used 2040c456dec4SQuentin Monnet * to make the distinction (ingress path is selected if the flag 2041c456dec4SQuentin Monnet * is present, egress path otherwise). Currently, XDP only 2042c456dec4SQuentin Monnet * supports redirection to the egress interface, and accepts no 2043c456dec4SQuentin Monnet * flag at all. 2044c456dec4SQuentin Monnet * 2045f25975f4SToke Høiland-Jørgensen * The same effect can also be attained with the more generic 2046f25975f4SToke Høiland-Jørgensen * **bpf_redirect_map**\ (), which uses a BPF map to store the 2047f25975f4SToke Høiland-Jørgensen * redirect target instead of providing it directly to the helper. 2048c456dec4SQuentin Monnet * Return 2049c456dec4SQuentin Monnet * For XDP, the helper returns **XDP_REDIRECT** on success or 2050c456dec4SQuentin Monnet * **XDP_ABORTED** on error. For other program types, the values 2051c456dec4SQuentin Monnet * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2052c456dec4SQuentin Monnet * error. 2053c456dec4SQuentin Monnet * 20541fdd08beSQuentin Monnet * u32 bpf_get_route_realm(struct sk_buff *skb) 20551fdd08beSQuentin Monnet * Description 20561fdd08beSQuentin Monnet * Retrieve the realm or the route, that is to say the 20571fdd08beSQuentin Monnet * **tclassid** field of the destination for the *skb*. The 2058b16fc097STobias Klauser * identifier retrieved is a user-provided tag, similar to the 20591fdd08beSQuentin Monnet * one used with the net_cls cgroup (see description for 20601fdd08beSQuentin Monnet * **bpf_get_cgroup_classid**\ () helper), but here this tag is 20611fdd08beSQuentin Monnet * held by a route (a destination entry), not by a task. 20621fdd08beSQuentin Monnet * 20631fdd08beSQuentin Monnet * Retrieving this identifier works with the clsact TC egress hook 20641fdd08beSQuentin Monnet * (see also **tc-bpf(8)**), or alternatively on conventional 20651fdd08beSQuentin Monnet * classful egress qdiscs, but not on TC ingress path. In case of 20661fdd08beSQuentin Monnet * clsact TC egress hook, this has the advantage that, internally, 20671fdd08beSQuentin Monnet * the destination entry has not been dropped yet in the transmit 20681fdd08beSQuentin Monnet * path. Therefore, the destination entry does not need to be 20691fdd08beSQuentin Monnet * artificially held via **netif_keep_dst**\ () for a classful 20701fdd08beSQuentin Monnet * qdisc until the *skb* is freed. 20711fdd08beSQuentin Monnet * 20721fdd08beSQuentin Monnet * This helper is available only if the kernel was compiled with 20731fdd08beSQuentin Monnet * **CONFIG_IP_ROUTE_CLASSID** configuration option. 20741fdd08beSQuentin Monnet * Return 20751fdd08beSQuentin Monnet * The realm of the route for the packet associated to *skb*, or 0 20761fdd08beSQuentin Monnet * if none was found. 20771fdd08beSQuentin Monnet * 2078bdb7b79bSAndrii Nakryiko * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2079c456dec4SQuentin Monnet * Description 2080c456dec4SQuentin Monnet * Write raw *data* blob into a special BPF perf event held by 2081c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2082c456dec4SQuentin Monnet * event must have the following attributes: **PERF_SAMPLE_RAW** 2083c456dec4SQuentin Monnet * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2084c456dec4SQuentin Monnet * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2085c456dec4SQuentin Monnet * 2086c456dec4SQuentin Monnet * The *flags* are used to indicate the index in *map* for which 2087c456dec4SQuentin Monnet * the value must be put, masked with **BPF_F_INDEX_MASK**. 2088c456dec4SQuentin Monnet * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2089c456dec4SQuentin Monnet * to indicate that the index of the current CPU core should be 2090c456dec4SQuentin Monnet * used. 2091c456dec4SQuentin Monnet * 2092c456dec4SQuentin Monnet * The value to write, of *size*, is passed through eBPF stack and 2093c456dec4SQuentin Monnet * pointed by *data*. 2094c456dec4SQuentin Monnet * 2095c456dec4SQuentin Monnet * The context of the program *ctx* needs also be passed to the 2096c456dec4SQuentin Monnet * helper. 2097c456dec4SQuentin Monnet * 2098c456dec4SQuentin Monnet * On user space, a program willing to read the values needs to 2099c456dec4SQuentin Monnet * call **perf_event_open**\ () on the perf event (either for 2100c456dec4SQuentin Monnet * one or for all CPUs) and to store the file descriptor into the 2101c456dec4SQuentin Monnet * *map*. This must be done before the eBPF program can send data 2102c456dec4SQuentin Monnet * into it. An example is available in file 2103c456dec4SQuentin Monnet * *samples/bpf/trace_output_user.c* in the Linux kernel source 2104c456dec4SQuentin Monnet * tree (the eBPF program counterpart is in 2105c456dec4SQuentin Monnet * *samples/bpf/trace_output_kern.c*). 2106c456dec4SQuentin Monnet * 2107c456dec4SQuentin Monnet * **bpf_perf_event_output**\ () achieves better performance 2108c456dec4SQuentin Monnet * than **bpf_trace_printk**\ () for sharing data with user 2109c456dec4SQuentin Monnet * space, and is much better suitable for streaming data from eBPF 2110c456dec4SQuentin Monnet * programs. 2111c456dec4SQuentin Monnet * 2112c456dec4SQuentin Monnet * Note that this helper is not restricted to tracing use cases 2113c456dec4SQuentin Monnet * and can be used with programs attached to TC or XDP as well, 2114c456dec4SQuentin Monnet * where it allows for passing data to user space listeners. Data 2115c456dec4SQuentin Monnet * can be: 2116c456dec4SQuentin Monnet * 2117c456dec4SQuentin Monnet * * Only custom structs, 2118c456dec4SQuentin Monnet * * Only the packet payload, or 2119c456dec4SQuentin Monnet * * A combination of both. 2120c456dec4SQuentin Monnet * Return 2121c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 2122c456dec4SQuentin Monnet * 2123bdb7b79bSAndrii Nakryiko * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 21241fdd08beSQuentin Monnet * Description 21251fdd08beSQuentin Monnet * This helper was provided as an easy way to load data from a 21261fdd08beSQuentin Monnet * packet. It can be used to load *len* bytes from *offset* from 21271fdd08beSQuentin Monnet * the packet associated to *skb*, into the buffer pointed by 21281fdd08beSQuentin Monnet * *to*. 21291fdd08beSQuentin Monnet * 21301fdd08beSQuentin Monnet * Since Linux 4.7, usage of this helper has mostly been replaced 21311fdd08beSQuentin Monnet * by "direct packet access", enabling packet data to be 21321fdd08beSQuentin Monnet * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 21331fdd08beSQuentin Monnet * pointing respectively to the first byte of packet data and to 21341fdd08beSQuentin Monnet * the byte after the last byte of packet data. However, it 21351fdd08beSQuentin Monnet * remains useful if one wishes to read large quantities of data 21361fdd08beSQuentin Monnet * at once from a packet into the eBPF stack. 21371fdd08beSQuentin Monnet * Return 21381fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 21391fdd08beSQuentin Monnet * 2140bdb7b79bSAndrii Nakryiko * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2141c456dec4SQuentin Monnet * Description 2142c456dec4SQuentin Monnet * Walk a user or a kernel stack and return its id. To achieve 2143c456dec4SQuentin Monnet * this, the helper needs *ctx*, which is a pointer to the context 2144c456dec4SQuentin Monnet * on which the tracing program is executed, and a pointer to a 2145c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2146c456dec4SQuentin Monnet * 2147c456dec4SQuentin Monnet * The last argument, *flags*, holds the number of stack frames to 2148c456dec4SQuentin Monnet * skip (from 0 to 255), masked with 2149c456dec4SQuentin Monnet * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2150c456dec4SQuentin Monnet * a combination of the following flags: 2151c456dec4SQuentin Monnet * 2152c456dec4SQuentin Monnet * **BPF_F_USER_STACK** 2153c456dec4SQuentin Monnet * Collect a user space stack instead of a kernel stack. 2154c456dec4SQuentin Monnet * **BPF_F_FAST_STACK_CMP** 2155c456dec4SQuentin Monnet * Compare stacks by hash only. 2156c456dec4SQuentin Monnet * **BPF_F_REUSE_STACKID** 2157c456dec4SQuentin Monnet * If two different stacks hash into the same *stackid*, 2158c456dec4SQuentin Monnet * discard the old one. 2159c456dec4SQuentin Monnet * 2160c456dec4SQuentin Monnet * The stack id retrieved is a 32 bit long integer handle which 2161c456dec4SQuentin Monnet * can be further combined with other data (including other stack 2162c456dec4SQuentin Monnet * ids) and used as a key into maps. This can be useful for 2163c456dec4SQuentin Monnet * generating a variety of graphs (such as flame graphs or off-cpu 2164c456dec4SQuentin Monnet * graphs). 2165c456dec4SQuentin Monnet * 2166c456dec4SQuentin Monnet * For walking a stack, this helper is an improvement over 2167c456dec4SQuentin Monnet * **bpf_probe_read**\ (), which can be used with unrolled loops 2168c456dec4SQuentin Monnet * but is not efficient and consumes a lot of eBPF instructions. 2169c456dec4SQuentin Monnet * Instead, **bpf_get_stackid**\ () can collect up to 2170c456dec4SQuentin Monnet * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2171c456dec4SQuentin Monnet * this limit can be controlled with the **sysctl** program, and 2172c456dec4SQuentin Monnet * that it should be manually increased in order to profile long 2173c456dec4SQuentin Monnet * user stacks (such as stacks for Java programs). To do so, use: 2174c456dec4SQuentin Monnet * 2175c456dec4SQuentin Monnet * :: 2176c456dec4SQuentin Monnet * 2177c456dec4SQuentin Monnet * # sysctl kernel.perf_event_max_stack=<new value> 2178c456dec4SQuentin Monnet * Return 2179c456dec4SQuentin Monnet * The positive or null stack id on success, or a negative error 2180c456dec4SQuentin Monnet * in case of failure. 2181c456dec4SQuentin Monnet * 21821fdd08beSQuentin Monnet * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 21831fdd08beSQuentin Monnet * Description 21841fdd08beSQuentin Monnet * Compute a checksum difference, from the raw buffer pointed by 21851fdd08beSQuentin Monnet * *from*, of length *from_size* (that must be a multiple of 4), 21861fdd08beSQuentin Monnet * towards the raw buffer pointed by *to*, of size *to_size* 21871fdd08beSQuentin Monnet * (same remark). An optional *seed* can be added to the value 21881fdd08beSQuentin Monnet * (this can be cascaded, the seed may come from a previous call 21891fdd08beSQuentin Monnet * to the helper). 21901fdd08beSQuentin Monnet * 21911fdd08beSQuentin Monnet * This is flexible enough to be used in several ways: 21921fdd08beSQuentin Monnet * 21931fdd08beSQuentin Monnet * * With *from_size* == 0, *to_size* > 0 and *seed* set to 21941fdd08beSQuentin Monnet * checksum, it can be used when pushing new data. 21951fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* == 0 and *seed* set to 21961fdd08beSQuentin Monnet * checksum, it can be used when removing data from a packet. 21971fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 21981fdd08beSQuentin Monnet * can be used to compute a diff. Note that *from_size* and 21991fdd08beSQuentin Monnet * *to_size* do not need to be equal. 22001fdd08beSQuentin Monnet * 22011fdd08beSQuentin Monnet * This helper can be used in combination with 22021fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 22031fdd08beSQuentin Monnet * which one can feed in the difference computed with 22041fdd08beSQuentin Monnet * **bpf_csum_diff**\ (). 22051fdd08beSQuentin Monnet * Return 22061fdd08beSQuentin Monnet * The checksum result, or a negative error code in case of 22071fdd08beSQuentin Monnet * failure. 22081fdd08beSQuentin Monnet * 2209bdb7b79bSAndrii Nakryiko * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 22101fdd08beSQuentin Monnet * Description 22111fdd08beSQuentin Monnet * Retrieve tunnel options metadata for the packet associated to 22121fdd08beSQuentin Monnet * *skb*, and store the raw tunnel option data to the buffer *opt* 22131fdd08beSQuentin Monnet * of *size*. 22141fdd08beSQuentin Monnet * 22151fdd08beSQuentin Monnet * This helper can be used with encapsulation devices that can 22161fdd08beSQuentin Monnet * operate in "collect metadata" mode (please refer to the related 22171fdd08beSQuentin Monnet * note in the description of **bpf_skb_get_tunnel_key**\ () for 22181fdd08beSQuentin Monnet * more details). A particular example where this can be used is 22191fdd08beSQuentin Monnet * in combination with the Geneve encapsulation protocol, where it 22201fdd08beSQuentin Monnet * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 22211fdd08beSQuentin Monnet * and retrieving arbitrary TLVs (Type-Length-Value headers) from 22221fdd08beSQuentin Monnet * the eBPF program. This allows for full customization of these 22231fdd08beSQuentin Monnet * headers. 22241fdd08beSQuentin Monnet * Return 22251fdd08beSQuentin Monnet * The size of the option data retrieved. 22261fdd08beSQuentin Monnet * 2227bdb7b79bSAndrii Nakryiko * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 22281fdd08beSQuentin Monnet * Description 22291fdd08beSQuentin Monnet * Set tunnel options metadata for the packet associated to *skb* 22301fdd08beSQuentin Monnet * to the option data contained in the raw buffer *opt* of *size*. 22311fdd08beSQuentin Monnet * 22321fdd08beSQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_opt**\ () 22331fdd08beSQuentin Monnet * helper for additional information. 22341fdd08beSQuentin Monnet * Return 22351fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 22361fdd08beSQuentin Monnet * 2237bdb7b79bSAndrii Nakryiko * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 22381fdd08beSQuentin Monnet * Description 22391fdd08beSQuentin Monnet * Change the protocol of the *skb* to *proto*. Currently 22401fdd08beSQuentin Monnet * supported are transition from IPv4 to IPv6, and from IPv6 to 22411fdd08beSQuentin Monnet * IPv4. The helper takes care of the groundwork for the 22421fdd08beSQuentin Monnet * transition, including resizing the socket buffer. The eBPF 22431fdd08beSQuentin Monnet * program is expected to fill the new headers, if any, via 22441fdd08beSQuentin Monnet * **skb_store_bytes**\ () and to recompute the checksums with 22451fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 22461fdd08beSQuentin Monnet * (). The main case for this helper is to perform NAT64 22471fdd08beSQuentin Monnet * operations out of an eBPF program. 22481fdd08beSQuentin Monnet * 22491fdd08beSQuentin Monnet * Internally, the GSO type is marked as dodgy so that headers are 22501fdd08beSQuentin Monnet * checked and segments are recalculated by the GSO/GRO engine. 22511fdd08beSQuentin Monnet * The size for GSO target is adapted as well. 22521fdd08beSQuentin Monnet * 22531fdd08beSQuentin Monnet * All values for *flags* are reserved for future usage, and must 22541fdd08beSQuentin Monnet * be left at zero. 22551fdd08beSQuentin Monnet * 225632e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 22571fdd08beSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 22581fdd08beSQuentin Monnet * previously done by the verifier are invalidated and must be 22591fdd08beSQuentin Monnet * performed again, if the helper is used in combination with 22601fdd08beSQuentin Monnet * direct packet access. 22611fdd08beSQuentin Monnet * Return 22621fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 22631fdd08beSQuentin Monnet * 2264bdb7b79bSAndrii Nakryiko * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 22651fdd08beSQuentin Monnet * Description 22661fdd08beSQuentin Monnet * Change the packet type for the packet associated to *skb*. This 22671fdd08beSQuentin Monnet * comes down to setting *skb*\ **->pkt_type** to *type*, except 22681fdd08beSQuentin Monnet * the eBPF program does not have a write access to *skb*\ 22691fdd08beSQuentin Monnet * **->pkt_type** beside this helper. Using a helper here allows 22701fdd08beSQuentin Monnet * for graceful handling of errors. 22711fdd08beSQuentin Monnet * 22721fdd08beSQuentin Monnet * The major use case is to change incoming *skb*s to 22731fdd08beSQuentin Monnet * **PACKET_HOST** in a programmatic way instead of having to 22741fdd08beSQuentin Monnet * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 22751fdd08beSQuentin Monnet * example. 22761fdd08beSQuentin Monnet * 22771fdd08beSQuentin Monnet * Note that *type* only allows certain values. At this time, they 22781fdd08beSQuentin Monnet * are: 22791fdd08beSQuentin Monnet * 22801fdd08beSQuentin Monnet * **PACKET_HOST** 22811fdd08beSQuentin Monnet * Packet is for us. 22821fdd08beSQuentin Monnet * **PACKET_BROADCAST** 22831fdd08beSQuentin Monnet * Send packet to all. 22841fdd08beSQuentin Monnet * **PACKET_MULTICAST** 22851fdd08beSQuentin Monnet * Send packet to group. 22861fdd08beSQuentin Monnet * **PACKET_OTHERHOST** 22871fdd08beSQuentin Monnet * Send packet to someone else. 22881fdd08beSQuentin Monnet * Return 22891fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 22901fdd08beSQuentin Monnet * 2291bdb7b79bSAndrii Nakryiko * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2292c6b5fb86SQuentin Monnet * Description 2293c6b5fb86SQuentin Monnet * Check whether *skb* is a descendant of the cgroup2 held by 2294c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2295c6b5fb86SQuentin Monnet * Return 2296c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 2297c6b5fb86SQuentin Monnet * 2298c6b5fb86SQuentin Monnet * * 0, if the *skb* failed the cgroup2 descendant test. 2299c6b5fb86SQuentin Monnet * * 1, if the *skb* succeeded the cgroup2 descendant test. 2300c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 2301c6b5fb86SQuentin Monnet * 2302fa15601aSQuentin Monnet * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2303fa15601aSQuentin Monnet * Description 2304fa15601aSQuentin Monnet * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2305fa15601aSQuentin Monnet * not set, in particular if the hash was cleared due to mangling, 2306fa15601aSQuentin Monnet * recompute this hash. Later accesses to the hash can be done 2307fa15601aSQuentin Monnet * directly with *skb*\ **->hash**. 2308fa15601aSQuentin Monnet * 2309fa15601aSQuentin Monnet * Calling **bpf_set_hash_invalid**\ (), changing a packet 2310fa15601aSQuentin Monnet * prototype with **bpf_skb_change_proto**\ (), or calling 2311fa15601aSQuentin Monnet * **bpf_skb_store_bytes**\ () with the 2312fa15601aSQuentin Monnet * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2313fa15601aSQuentin Monnet * the hash and to trigger a new computation for the next call to 2314fa15601aSQuentin Monnet * **bpf_get_hash_recalc**\ (). 2315fa15601aSQuentin Monnet * Return 2316fa15601aSQuentin Monnet * The 32-bit hash. 2317fa15601aSQuentin Monnet * 2318c456dec4SQuentin Monnet * u64 bpf_get_current_task(void) 2319e40fbbf0SUsama Arif * Description 2320e40fbbf0SUsama Arif * Get the current task. 2321c456dec4SQuentin Monnet * Return 2322c456dec4SQuentin Monnet * A pointer to the current task struct. 2323fa15601aSQuentin Monnet * 2324bdb7b79bSAndrii Nakryiko * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2325c6b5fb86SQuentin Monnet * Description 2326c6b5fb86SQuentin Monnet * Attempt in a safe way to write *len* bytes from the buffer 2327c6b5fb86SQuentin Monnet * *src* to *dst* in memory. It only works for threads that are in 2328c6b5fb86SQuentin Monnet * user context, and *dst* must be a valid user space address. 2329c6b5fb86SQuentin Monnet * 2330c6b5fb86SQuentin Monnet * This helper should not be used to implement any kind of 2331c6b5fb86SQuentin Monnet * security mechanism because of TOC-TOU attacks, but rather to 2332c6b5fb86SQuentin Monnet * debug, divert, and manipulate execution of semi-cooperative 2333c6b5fb86SQuentin Monnet * processes. 2334c6b5fb86SQuentin Monnet * 2335c6b5fb86SQuentin Monnet * Keep in mind that this feature is meant for experiments, and it 2336c6b5fb86SQuentin Monnet * has a risk of crashing the system and running programs. 2337c6b5fb86SQuentin Monnet * Therefore, when an eBPF program using this helper is attached, 2338c6b5fb86SQuentin Monnet * a warning including PID and process name is printed to kernel 2339c6b5fb86SQuentin Monnet * logs. 2340c6b5fb86SQuentin Monnet * Return 2341c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2342c6b5fb86SQuentin Monnet * 2343bdb7b79bSAndrii Nakryiko * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2344c6b5fb86SQuentin Monnet * Description 2345c6b5fb86SQuentin Monnet * Check whether the probe is being run is the context of a given 2346c6b5fb86SQuentin Monnet * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2347c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2348c6b5fb86SQuentin Monnet * Return 2349c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 2350c6b5fb86SQuentin Monnet * 235158617014SHengqi Chen * * 1, if current task belongs to the cgroup2. 235258617014SHengqi Chen * * 0, if current task does not belong to the cgroup2. 2353c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 2354c6b5fb86SQuentin Monnet * 2355bdb7b79bSAndrii Nakryiko * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2356fa15601aSQuentin Monnet * Description 2357fa15601aSQuentin Monnet * Resize (trim or grow) the packet associated to *skb* to the 2358fa15601aSQuentin Monnet * new *len*. The *flags* are reserved for future usage, and must 2359fa15601aSQuentin Monnet * be left at zero. 2360fa15601aSQuentin Monnet * 2361fa15601aSQuentin Monnet * The basic idea is that the helper performs the needed work to 2362fa15601aSQuentin Monnet * change the size of the packet, then the eBPF program rewrites 2363fa15601aSQuentin Monnet * the rest via helpers like **bpf_skb_store_bytes**\ (), 2364fa15601aSQuentin Monnet * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2365fa15601aSQuentin Monnet * and others. This helper is a slow path utility intended for 2366fa15601aSQuentin Monnet * replies with control messages. And because it is targeted for 2367fa15601aSQuentin Monnet * slow path, the helper itself can afford to be slow: it 2368fa15601aSQuentin Monnet * implicitly linearizes, unclones and drops offloads from the 2369fa15601aSQuentin Monnet * *skb*. 2370fa15601aSQuentin Monnet * 237132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2372fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2373fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2374fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2375fa15601aSQuentin Monnet * direct packet access. 2376fa15601aSQuentin Monnet * Return 2377fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2378fa15601aSQuentin Monnet * 2379bdb7b79bSAndrii Nakryiko * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2380fa15601aSQuentin Monnet * Description 2381fa15601aSQuentin Monnet * Pull in non-linear data in case the *skb* is non-linear and not 2382fa15601aSQuentin Monnet * all of *len* are part of the linear section. Make *len* bytes 2383fa15601aSQuentin Monnet * from *skb* readable and writable. If a zero value is passed for 2384bdb2bc75SJoanne Koong * *len*, then all bytes in the linear part of *skb* will be made 2385bdb2bc75SJoanne Koong * readable and writable. 2386fa15601aSQuentin Monnet * 2387fa15601aSQuentin Monnet * This helper is only needed for reading and writing with direct 2388fa15601aSQuentin Monnet * packet access. 2389fa15601aSQuentin Monnet * 2390fa15601aSQuentin Monnet * For direct packet access, testing that offsets to access 2391fa15601aSQuentin Monnet * are within packet boundaries (test on *skb*\ **->data_end**) is 2392fa15601aSQuentin Monnet * susceptible to fail if offsets are invalid, or if the requested 2393fa15601aSQuentin Monnet * data is in non-linear parts of the *skb*. On failure the 2394fa15601aSQuentin Monnet * program can just bail out, or in the case of a non-linear 2395fa15601aSQuentin Monnet * buffer, use a helper to make the data available. The 2396fa15601aSQuentin Monnet * **bpf_skb_load_bytes**\ () helper is a first solution to access 2397fa15601aSQuentin Monnet * the data. Another one consists in using **bpf_skb_pull_data** 2398fa15601aSQuentin Monnet * to pull in once the non-linear parts, then retesting and 2399fa15601aSQuentin Monnet * eventually access the data. 2400fa15601aSQuentin Monnet * 2401fa15601aSQuentin Monnet * At the same time, this also makes sure the *skb* is uncloned, 2402fa15601aSQuentin Monnet * which is a necessary condition for direct write. As this needs 2403fa15601aSQuentin Monnet * to be an invariant for the write part only, the verifier 2404fa15601aSQuentin Monnet * detects writes and adds a prologue that is calling 2405fa15601aSQuentin Monnet * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2406fa15601aSQuentin Monnet * the very beginning in case it is indeed cloned. 2407fa15601aSQuentin Monnet * 240832e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2409fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2410fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2411fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2412fa15601aSQuentin Monnet * direct packet access. 2413fa15601aSQuentin Monnet * Return 2414fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2415fa15601aSQuentin Monnet * 2416fa15601aSQuentin Monnet * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2417fa15601aSQuentin Monnet * Description 2418fa15601aSQuentin Monnet * Add the checksum *csum* into *skb*\ **->csum** in case the 2419fa15601aSQuentin Monnet * driver has supplied a checksum for the entire packet into that 2420fa15601aSQuentin Monnet * field. Return an error otherwise. This helper is intended to be 2421fa15601aSQuentin Monnet * used in combination with **bpf_csum_diff**\ (), in particular 2422fa15601aSQuentin Monnet * when the checksum needs to be updated after data has been 2423fa15601aSQuentin Monnet * written into the packet through direct packet access. 2424fa15601aSQuentin Monnet * Return 2425fa15601aSQuentin Monnet * The checksum on success, or a negative error code in case of 2426fa15601aSQuentin Monnet * failure. 2427fa15601aSQuentin Monnet * 2428fa15601aSQuentin Monnet * void bpf_set_hash_invalid(struct sk_buff *skb) 2429fa15601aSQuentin Monnet * Description 2430fa15601aSQuentin Monnet * Invalidate the current *skb*\ **->hash**. It can be used after 2431fa15601aSQuentin Monnet * mangling on headers through direct packet access, in order to 2432fa15601aSQuentin Monnet * indicate that the hash is outdated and to trigger a 2433fa15601aSQuentin Monnet * recalculation the next time the kernel tries to access this 2434fa15601aSQuentin Monnet * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2435e40fbbf0SUsama Arif * Return 2436e40fbbf0SUsama Arif * void. 2437fa15601aSQuentin Monnet * 2438bdb7b79bSAndrii Nakryiko * long bpf_get_numa_node_id(void) 2439fa15601aSQuentin Monnet * Description 2440fa15601aSQuentin Monnet * Return the id of the current NUMA node. The primary use case 2441fa15601aSQuentin Monnet * for this helper is the selection of sockets for the local NUMA 2442fa15601aSQuentin Monnet * node, when the program is attached to sockets using the 2443fa15601aSQuentin Monnet * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2444fa15601aSQuentin Monnet * but the helper is also available to other eBPF program types, 2445fa15601aSQuentin Monnet * similarly to **bpf_get_smp_processor_id**\ (). 2446fa15601aSQuentin Monnet * Return 2447fa15601aSQuentin Monnet * The id of current NUMA node. 2448fa15601aSQuentin Monnet * 2449bdb7b79bSAndrii Nakryiko * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2450c6b5fb86SQuentin Monnet * Description 2451c6b5fb86SQuentin Monnet * Grows headroom of packet associated to *skb* and adjusts the 2452c6b5fb86SQuentin Monnet * offset of the MAC header accordingly, adding *len* bytes of 2453c6b5fb86SQuentin Monnet * space. It automatically extends and reallocates memory as 2454c6b5fb86SQuentin Monnet * required. 2455c6b5fb86SQuentin Monnet * 2456c6b5fb86SQuentin Monnet * This helper can be used on a layer 3 *skb* to push a MAC header 2457c6b5fb86SQuentin Monnet * for redirection into a layer 2 device. 2458c6b5fb86SQuentin Monnet * 2459c6b5fb86SQuentin Monnet * All values for *flags* are reserved for future usage, and must 2460c6b5fb86SQuentin Monnet * be left at zero. 2461c6b5fb86SQuentin Monnet * 246232e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2463c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2464c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 2465c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 2466c6b5fb86SQuentin Monnet * direct packet access. 2467c6b5fb86SQuentin Monnet * Return 2468c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2469c6b5fb86SQuentin Monnet * 2470bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2471c6b5fb86SQuentin Monnet * Description 2472c6b5fb86SQuentin Monnet * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2473c6b5fb86SQuentin Monnet * it is possible to use a negative value for *delta*. This helper 2474c6b5fb86SQuentin Monnet * can be used to prepare the packet for pushing or popping 2475c6b5fb86SQuentin Monnet * headers. 2476c6b5fb86SQuentin Monnet * 247732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2478c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2479c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 2480c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 2481c6b5fb86SQuentin Monnet * direct packet access. 2482c6b5fb86SQuentin Monnet * Return 2483c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 2484c6b5fb86SQuentin Monnet * 2485bdb7b79bSAndrii Nakryiko * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2486c6b5fb86SQuentin Monnet * Description 24876ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe kernel address 2488ab8d7809SQuentin Monnet * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 24896ae08ae3SDaniel Borkmann * more details. 2490c6b5fb86SQuentin Monnet * 2491ab8d7809SQuentin Monnet * Generally, use **bpf_probe_read_user_str**\ () or 2492ab8d7809SQuentin Monnet * **bpf_probe_read_kernel_str**\ () instead. 2493c6b5fb86SQuentin Monnet * Return 2494c6b5fb86SQuentin Monnet * On success, the strictly positive length of the string, 2495c6b5fb86SQuentin Monnet * including the trailing NUL character. On error, a negative 2496c6b5fb86SQuentin Monnet * value. 2497c6b5fb86SQuentin Monnet * 2498c6b5fb86SQuentin Monnet * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2499c6b5fb86SQuentin Monnet * Description 2500c6b5fb86SQuentin Monnet * If the **struct sk_buff** pointed by *skb* has a known socket, 2501c6b5fb86SQuentin Monnet * retrieve the cookie (generated by the kernel) of this socket. 2502c6b5fb86SQuentin Monnet * If no cookie has been set yet, generate a new cookie. Once 2503c6b5fb86SQuentin Monnet * generated, the socket cookie remains stable for the life of the 2504c6b5fb86SQuentin Monnet * socket. This helper can be useful for monitoring per socket 2505cd48bddaSDaniel Borkmann * networking traffic statistics as it provides a global socket 2506cd48bddaSDaniel Borkmann * identifier that can be assumed unique. 2507c6b5fb86SQuentin Monnet * Return 250807881ccbSFlorent Revest * A 8-byte long unique number on success, or 0 if the socket 250907881ccbSFlorent Revest * field is missing inside *skb*. 2510c6b5fb86SQuentin Monnet * 2511d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2512d692f113SAndrey Ignatov * Description 2513d692f113SAndrey Ignatov * Equivalent to bpf_get_socket_cookie() helper that accepts 251462369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_addr** context. 2515d692f113SAndrey Ignatov * Return 251607881ccbSFlorent Revest * A 8-byte long unique number. 2517d692f113SAndrey Ignatov * 2518d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2519d692f113SAndrey Ignatov * Description 2520ab8d7809SQuentin Monnet * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 252162369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_ops** context. 2522d692f113SAndrey Ignatov * Return 252307881ccbSFlorent Revest * A 8-byte long unique number. 2524d692f113SAndrey Ignatov * 2525c5dbb89fSFlorent Revest * u64 bpf_get_socket_cookie(struct sock *sk) 2526c5dbb89fSFlorent Revest * Description 2527c5dbb89fSFlorent Revest * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2528c5dbb89fSFlorent Revest * *sk*, but gets socket from a BTF **struct sock**. This helper 2529c5dbb89fSFlorent Revest * also works for sleepable programs. 2530c5dbb89fSFlorent Revest * Return 2531c5dbb89fSFlorent Revest * A 8-byte long unique number or 0 if *sk* is NULL. 2532c5dbb89fSFlorent Revest * 2533c6b5fb86SQuentin Monnet * u32 bpf_get_socket_uid(struct sk_buff *skb) 2534e40fbbf0SUsama Arif * Description 2535e40fbbf0SUsama Arif * Get the owner UID of the socked associated to *skb*. 2536c6b5fb86SQuentin Monnet * Return 2537c6b5fb86SQuentin Monnet * The owner UID of the socket associated to *skb*. If the socket 2538c6b5fb86SQuentin Monnet * is **NULL**, or if it is not a full socket (i.e. if it is a 2539c6b5fb86SQuentin Monnet * time-wait or a request socket instead), **overflowuid** value 2540c6b5fb86SQuentin Monnet * is returned (note that **overflowuid** might also be the actual 2541c6b5fb86SQuentin Monnet * UID value for the socket). 2542c6b5fb86SQuentin Monnet * 2543bdb7b79bSAndrii Nakryiko * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2544fa15601aSQuentin Monnet * Description 2545fa15601aSQuentin Monnet * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2546fa15601aSQuentin Monnet * to value *hash*. 2547fa15601aSQuentin Monnet * Return 2548fa15601aSQuentin Monnet * 0 2549fa15601aSQuentin Monnet * 2550bdb7b79bSAndrii Nakryiko * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 25517aa79a86SQuentin Monnet * Description 25527aa79a86SQuentin Monnet * Emulate a call to **setsockopt()** on the socket associated to 25537aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 25547aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 25557aa79a86SQuentin Monnet * must be specified, see **setsockopt(2)** for more information. 25567aa79a86SQuentin Monnet * The option value of length *optlen* is pointed by *optval*. 25577aa79a86SQuentin Monnet * 2558beecf11bSStanislav Fomichev * *bpf_socket* should be one of the following: 2559ab8d7809SQuentin Monnet * 2560beecf11bSStanislav Fomichev * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2561beecf11bSStanislav Fomichev * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2562beecf11bSStanislav Fomichev * and **BPF_CGROUP_INET6_CONNECT**. 2563beecf11bSStanislav Fomichev * 25647aa79a86SQuentin Monnet * This helper actually implements a subset of **setsockopt()**. 25657aa79a86SQuentin Monnet * It supports the following *level*\ s: 25667aa79a86SQuentin Monnet * 25677aa79a86SQuentin Monnet * * **SOL_SOCKET**, which supports the following *optname*\ s: 25687aa79a86SQuentin Monnet * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2569f9bcf968SDmitry Yakunin * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2570f9bcf968SDmitry Yakunin * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 25717aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports the following *optname*\ s: 25727aa79a86SQuentin Monnet * **TCP_CONGESTION**, **TCP_BPF_IW**, 2573f9bcf968SDmitry Yakunin * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2574f9bcf968SDmitry Yakunin * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2575eca43ee6SNikita V. Shirokov * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**. 25767aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 25777aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 25787aa79a86SQuentin Monnet * Return 25797aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 25807aa79a86SQuentin Monnet * 2581bdb7b79bSAndrii Nakryiko * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2582fa15601aSQuentin Monnet * Description 2583fa15601aSQuentin Monnet * Grow or shrink the room for data in the packet associated to 2584fa15601aSQuentin Monnet * *skb* by *len_diff*, and according to the selected *mode*. 2585fa15601aSQuentin Monnet * 2586836e66c2SDaniel Borkmann * By default, the helper will reset any offloaded checksum 2587836e66c2SDaniel Borkmann * indicator of the skb to CHECKSUM_NONE. This can be avoided 2588836e66c2SDaniel Borkmann * by the following flag: 2589836e66c2SDaniel Borkmann * 2590836e66c2SDaniel Borkmann * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2591836e66c2SDaniel Borkmann * checksum data of the skb to CHECKSUM_NONE. 2592836e66c2SDaniel Borkmann * 259314aa3192SWillem de Bruijn * There are two supported modes at this time: 259414aa3192SWillem de Bruijn * 259514aa3192SWillem de Bruijn * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 25964961d077SQuentin Monnet * (room space is added or removed between the layer 2 and 25974961d077SQuentin Monnet * layer 3 headers). 2598fa15601aSQuentin Monnet * 2599fa15601aSQuentin Monnet * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 26004961d077SQuentin Monnet * (room space is added or removed between the layer 3 and 26014961d077SQuentin Monnet * layer 4 headers). 2602fa15601aSQuentin Monnet * 2603868d5235SWillem de Bruijn * The following flags are supported at this time: 26042278f6ccSWillem de Bruijn * 26052278f6ccSWillem de Bruijn * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 26062278f6ccSWillem de Bruijn * Adjusting mss in this way is not allowed for datagrams. 2607fa15601aSQuentin Monnet * 260880867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 260980867c5eSQuentin Monnet * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2610868d5235SWillem de Bruijn * Any new space is reserved to hold a tunnel header. 2611868d5235SWillem de Bruijn * Configure skb offsets and other fields accordingly. 2612868d5235SWillem de Bruijn * 261380867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 261480867c5eSQuentin Monnet * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2615868d5235SWillem de Bruijn * Use with ENCAP_L3 flags to further specify the tunnel type. 2616868d5235SWillem de Bruijn * 261780867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 261858dfc900SAlan Maguire * Use with ENCAP_L3/L4 flags to further specify the tunnel 261980867c5eSQuentin Monnet * type; *len* is the length of the inner MAC header. 262058dfc900SAlan Maguire * 2621d01b59c9SXuesen Huang * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2622d01b59c9SXuesen Huang * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2623d01b59c9SXuesen Huang * L2 type as Ethernet. 2624d01b59c9SXuesen Huang * 262532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2626fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2627fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2628fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2629fa15601aSQuentin Monnet * direct packet access. 2630fa15601aSQuentin Monnet * Return 2631fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 2632fa15601aSQuentin Monnet * 2633bdb7b79bSAndrii Nakryiko * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 2634ab127040SQuentin Monnet * Description 2635ab127040SQuentin Monnet * Redirect the packet to the endpoint referenced by *map* at 2636ab127040SQuentin Monnet * index *key*. Depending on its type, this *map* can contain 2637ab127040SQuentin Monnet * references to net devices (for forwarding packets through other 2638ab127040SQuentin Monnet * ports), or to CPUs (for redirecting XDP frames to another CPU; 2639ab127040SQuentin Monnet * but this is only implemented for native XDP (with driver 2640ab127040SQuentin Monnet * support) as of this writing). 2641ab127040SQuentin Monnet * 264243e74c02SToke Høiland-Jørgensen * The lower two bits of *flags* are used as the return code if 264343e74c02SToke Høiland-Jørgensen * the map lookup fails. This is so that the return value can be 2644ab8d7809SQuentin Monnet * one of the XDP program return codes up to **XDP_TX**, as chosen 2645e624d4edSHangbin Liu * by the caller. The higher bits of *flags* can be set to 2646e624d4edSHangbin Liu * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2647e624d4edSHangbin Liu * 2648e624d4edSHangbin Liu * With BPF_F_BROADCAST the packet will be broadcasted to all the 2649e624d4edSHangbin Liu * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2650e624d4edSHangbin Liu * interface will be excluded when do broadcasting. 2651ab127040SQuentin Monnet * 2652ab8d7809SQuentin Monnet * See also **bpf_redirect**\ (), which only supports redirecting 2653ab8d7809SQuentin Monnet * to an ifindex, but doesn't require a map to do so. 2654ab127040SQuentin Monnet * Return 2655f25975f4SToke Høiland-Jørgensen * **XDP_REDIRECT** on success, or the value of the two lower bits 2656a33d3147SJakub Wilk * of the *flags* argument on error. 2657ab127040SQuentin Monnet * 2658bdb7b79bSAndrii Nakryiko * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2659ab127040SQuentin Monnet * Description 2660ab127040SQuentin Monnet * Redirect the packet to the socket referenced by *map* (of type 2661ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2662ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 2663ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 2664ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 2665ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 2666ab127040SQuentin Monnet * Return 2667ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 2668ab127040SQuentin Monnet * 2669bdb7b79bSAndrii Nakryiko * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2670ab127040SQuentin Monnet * Description 2671ab127040SQuentin Monnet * Add an entry to, or update a *map* referencing sockets. The 2672ab127040SQuentin Monnet * *skops* is used as a new value for the entry associated to 2673ab127040SQuentin Monnet * *key*. *flags* is one of: 2674ab127040SQuentin Monnet * 2675ab127040SQuentin Monnet * **BPF_NOEXIST** 2676ab127040SQuentin Monnet * The entry for *key* must not exist in the map. 2677ab127040SQuentin Monnet * **BPF_EXIST** 2678ab127040SQuentin Monnet * The entry for *key* must already exist in the map. 2679ab127040SQuentin Monnet * **BPF_ANY** 2680ab127040SQuentin Monnet * No condition on the existence of the entry for *key*. 2681ab127040SQuentin Monnet * 2682ab127040SQuentin Monnet * If the *map* has eBPF programs (parser and verdict), those will 2683ab127040SQuentin Monnet * be inherited by the socket being added. If the socket is 2684ab127040SQuentin Monnet * already attached to eBPF programs, this results in an error. 2685ab127040SQuentin Monnet * Return 2686ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 2687ab127040SQuentin Monnet * 2688bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2689fa15601aSQuentin Monnet * Description 2690fa15601aSQuentin Monnet * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2691fa15601aSQuentin Monnet * *delta* (which can be positive or negative). Note that this 2692fa15601aSQuentin Monnet * operation modifies the address stored in *xdp_md*\ **->data**, 2693fa15601aSQuentin Monnet * so the latter must be loaded only after the helper has been 2694fa15601aSQuentin Monnet * called. 2695fa15601aSQuentin Monnet * 2696fa15601aSQuentin Monnet * The use of *xdp_md*\ **->data_meta** is optional and programs 2697fa15601aSQuentin Monnet * are not required to use it. The rationale is that when the 2698fa15601aSQuentin Monnet * packet is processed with XDP (e.g. as DoS filter), it is 2699fa15601aSQuentin Monnet * possible to push further meta data along with it before passing 2700fa15601aSQuentin Monnet * to the stack, and to give the guarantee that an ingress eBPF 2701fa15601aSQuentin Monnet * program attached as a TC classifier on the same device can pick 2702fa15601aSQuentin Monnet * this up for further post-processing. Since TC works with socket 2703fa15601aSQuentin Monnet * buffers, it remains possible to set from XDP the **mark** or 2704fa15601aSQuentin Monnet * **priority** pointers, or other pointers for the socket buffer. 2705fa15601aSQuentin Monnet * Having this scratch space generic and programmable allows for 2706fa15601aSQuentin Monnet * more flexibility as the user is free to store whatever meta 2707fa15601aSQuentin Monnet * data they need. 2708fa15601aSQuentin Monnet * 270932e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2710fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2711fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 2712fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 2713fa15601aSQuentin Monnet * direct packet access. 2714fa15601aSQuentin Monnet * Return 2715fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 27167aa79a86SQuentin Monnet * 2717bdb7b79bSAndrii Nakryiko * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 27187aa79a86SQuentin Monnet * Description 27197aa79a86SQuentin Monnet * Read the value of a perf event counter, and store it into *buf* 27207aa79a86SQuentin Monnet * of size *buf_size*. This helper relies on a *map* of type 27217aa79a86SQuentin Monnet * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 27227aa79a86SQuentin Monnet * counter is selected when *map* is updated with perf event file 27237aa79a86SQuentin Monnet * descriptors. The *map* is an array whose size is the number of 27247aa79a86SQuentin Monnet * available CPUs, and each cell contains a value relative to one 27257aa79a86SQuentin Monnet * CPU. The value to retrieve is indicated by *flags*, that 27267aa79a86SQuentin Monnet * contains the index of the CPU to look up, masked with 27277aa79a86SQuentin Monnet * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 27287aa79a86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 27297aa79a86SQuentin Monnet * current CPU should be retrieved. 27307aa79a86SQuentin Monnet * 27317aa79a86SQuentin Monnet * This helper behaves in a way close to 27327aa79a86SQuentin Monnet * **bpf_perf_event_read**\ () helper, save that instead of 27337aa79a86SQuentin Monnet * just returning the value observed, it fills the *buf* 27347aa79a86SQuentin Monnet * structure. This allows for additional data to be retrieved: in 27357aa79a86SQuentin Monnet * particular, the enabled and running times (in *buf*\ 27367aa79a86SQuentin Monnet * **->enabled** and *buf*\ **->running**, respectively) are 27377aa79a86SQuentin Monnet * copied. In general, **bpf_perf_event_read_value**\ () is 27387aa79a86SQuentin Monnet * recommended over **bpf_perf_event_read**\ (), which has some 27397aa79a86SQuentin Monnet * ABI issues and provides fewer functionalities. 27407aa79a86SQuentin Monnet * 27417aa79a86SQuentin Monnet * These values are interesting, because hardware PMU (Performance 27427aa79a86SQuentin Monnet * Monitoring Unit) counters are limited resources. When there are 27437aa79a86SQuentin Monnet * more PMU based perf events opened than available counters, 27447aa79a86SQuentin Monnet * kernel will multiplex these events so each event gets certain 27457aa79a86SQuentin Monnet * percentage (but not all) of the PMU time. In case that 27467aa79a86SQuentin Monnet * multiplexing happens, the number of samples or counter value 27477aa79a86SQuentin Monnet * will not reflect the case compared to when no multiplexing 27487aa79a86SQuentin Monnet * occurs. This makes comparison between different runs difficult. 27497aa79a86SQuentin Monnet * Typically, the counter value should be normalized before 27507aa79a86SQuentin Monnet * comparing to other experiments. The usual normalization is done 27517aa79a86SQuentin Monnet * as follows. 27527aa79a86SQuentin Monnet * 27537aa79a86SQuentin Monnet * :: 27547aa79a86SQuentin Monnet * 27557aa79a86SQuentin Monnet * normalized_counter = counter * t_enabled / t_running 27567aa79a86SQuentin Monnet * 27577aa79a86SQuentin Monnet * Where t_enabled is the time enabled for event and t_running is 27587aa79a86SQuentin Monnet * the time running for event since last normalization. The 27597aa79a86SQuentin Monnet * enabled and running times are accumulated since the perf event 27607aa79a86SQuentin Monnet * open. To achieve scaling factor between two invocations of an 2761ab8d7809SQuentin Monnet * eBPF program, users can use CPU id as the key (which is 27627aa79a86SQuentin Monnet * typical for perf array usage model) to remember the previous 27637aa79a86SQuentin Monnet * value and do the calculation inside the eBPF program. 27647aa79a86SQuentin Monnet * Return 27657aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 27667aa79a86SQuentin Monnet * 2767bdb7b79bSAndrii Nakryiko * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 27687aa79a86SQuentin Monnet * Description 27697aa79a86SQuentin Monnet * For en eBPF program attached to a perf event, retrieve the 27707aa79a86SQuentin Monnet * value of the event counter associated to *ctx* and store it in 27717aa79a86SQuentin Monnet * the structure pointed by *buf* and of size *buf_size*. Enabled 27727aa79a86SQuentin Monnet * and running times are also stored in the structure (see 27737aa79a86SQuentin Monnet * description of helper **bpf_perf_event_read_value**\ () for 27747aa79a86SQuentin Monnet * more details). 27757aa79a86SQuentin Monnet * Return 27767aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 27777aa79a86SQuentin Monnet * 2778bdb7b79bSAndrii Nakryiko * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 27797aa79a86SQuentin Monnet * Description 27807aa79a86SQuentin Monnet * Emulate a call to **getsockopt()** on the socket associated to 27817aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 27827aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 27837aa79a86SQuentin Monnet * must be specified, see **getsockopt(2)** for more information. 27847aa79a86SQuentin Monnet * The retrieved value is stored in the structure pointed by 27857aa79a86SQuentin Monnet * *opval* and of length *optlen*. 27867aa79a86SQuentin Monnet * 2787beecf11bSStanislav Fomichev * *bpf_socket* should be one of the following: 2788ab8d7809SQuentin Monnet * 2789beecf11bSStanislav Fomichev * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2790beecf11bSStanislav Fomichev * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2791beecf11bSStanislav Fomichev * and **BPF_CGROUP_INET6_CONNECT**. 2792beecf11bSStanislav Fomichev * 27937aa79a86SQuentin Monnet * This helper actually implements a subset of **getsockopt()**. 27947aa79a86SQuentin Monnet * It supports the following *level*\ s: 27957aa79a86SQuentin Monnet * 27967aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports *optname* 27977aa79a86SQuentin Monnet * **TCP_CONGESTION**. 27987aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 27997aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 28007aa79a86SQuentin Monnet * Return 28017aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 28027aa79a86SQuentin Monnet * 2803bdb7b79bSAndrii Nakryiko * long bpf_override_return(struct pt_regs *regs, u64 rc) 28047aa79a86SQuentin Monnet * Description 28057aa79a86SQuentin Monnet * Used for error injection, this helper uses kprobes to override 28067aa79a86SQuentin Monnet * the return value of the probed function, and to set it to *rc*. 28077aa79a86SQuentin Monnet * The first argument is the context *regs* on which the kprobe 28087aa79a86SQuentin Monnet * works. 28097aa79a86SQuentin Monnet * 2810ab8d7809SQuentin Monnet * This helper works by setting the PC (program counter) 28117aa79a86SQuentin Monnet * to an override function which is run in place of the original 28127aa79a86SQuentin Monnet * probed function. This means the probed function is not run at 28137aa79a86SQuentin Monnet * all. The replacement function just returns with the required 28147aa79a86SQuentin Monnet * value. 28157aa79a86SQuentin Monnet * 28167aa79a86SQuentin Monnet * This helper has security implications, and thus is subject to 28177aa79a86SQuentin Monnet * restrictions. It is only available if the kernel was compiled 28187aa79a86SQuentin Monnet * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 28197aa79a86SQuentin Monnet * option, and in this case it only works on functions tagged with 28207aa79a86SQuentin Monnet * **ALLOW_ERROR_INJECTION** in the kernel code. 28217aa79a86SQuentin Monnet * 28227aa79a86SQuentin Monnet * Also, the helper is only available for the architectures having 28237aa79a86SQuentin Monnet * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 28247aa79a86SQuentin Monnet * x86 architecture is the only one to support this feature. 28257aa79a86SQuentin Monnet * Return 28267aa79a86SQuentin Monnet * 0 28277aa79a86SQuentin Monnet * 2828bdb7b79bSAndrii Nakryiko * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 28297aa79a86SQuentin Monnet * Description 28307aa79a86SQuentin Monnet * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 28317aa79a86SQuentin Monnet * for the full TCP socket associated to *bpf_sock_ops* to 28327aa79a86SQuentin Monnet * *argval*. 28337aa79a86SQuentin Monnet * 28347aa79a86SQuentin Monnet * The primary use of this field is to determine if there should 28357aa79a86SQuentin Monnet * be calls to eBPF programs of type 28367aa79a86SQuentin Monnet * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 28377aa79a86SQuentin Monnet * code. A program of the same type can change its value, per 28387aa79a86SQuentin Monnet * connection and as necessary, when the connection is 28397aa79a86SQuentin Monnet * established. This field is directly accessible for reading, but 28407aa79a86SQuentin Monnet * this helper must be used for updates in order to return an 28417aa79a86SQuentin Monnet * error if an eBPF program tries to set a callback that is not 28427aa79a86SQuentin Monnet * supported in the current kernel. 28437aa79a86SQuentin Monnet * 2844725721a6SViet Hoang Tran * *argval* is a flag array which can combine these flags: 28457aa79a86SQuentin Monnet * 28467aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 28477aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 28487aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 284923729ff2SStanislav Fomichev * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 28507aa79a86SQuentin Monnet * 2851725721a6SViet Hoang Tran * Therefore, this function can be used to clear a callback flag by 2852725721a6SViet Hoang Tran * setting the appropriate bit to zero. e.g. to disable the RTO 2853725721a6SViet Hoang Tran * callback: 2854725721a6SViet Hoang Tran * 2855725721a6SViet Hoang Tran * **bpf_sock_ops_cb_flags_set(bpf_sock,** 2856725721a6SViet Hoang Tran * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 2857725721a6SViet Hoang Tran * 28587aa79a86SQuentin Monnet * Here are some examples of where one could call such eBPF 28597aa79a86SQuentin Monnet * program: 28607aa79a86SQuentin Monnet * 28617aa79a86SQuentin Monnet * * When RTO fires. 28627aa79a86SQuentin Monnet * * When a packet is retransmitted. 28637aa79a86SQuentin Monnet * * When the connection terminates. 28647aa79a86SQuentin Monnet * * When a packet is sent. 28657aa79a86SQuentin Monnet * * When a packet is received. 28667aa79a86SQuentin Monnet * Return 28677aa79a86SQuentin Monnet * Code **-EINVAL** if the socket is not a full TCP socket; 28687aa79a86SQuentin Monnet * otherwise, a positive number containing the bits that could not 28697aa79a86SQuentin Monnet * be set is returned (which comes down to 0 if all bits were set 28707aa79a86SQuentin Monnet * as required). 28717aa79a86SQuentin Monnet * 2872bdb7b79bSAndrii Nakryiko * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 2873ab127040SQuentin Monnet * Description 2874ab127040SQuentin Monnet * This helper is used in programs implementing policies at the 2875ab127040SQuentin Monnet * socket level. If the message *msg* is allowed to pass (i.e. if 2876ab127040SQuentin Monnet * the verdict eBPF program returns **SK_PASS**), redirect it to 2877ab127040SQuentin Monnet * the socket referenced by *map* (of type 2878ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2879ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 2880ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 2881ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 2882ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 2883ab127040SQuentin Monnet * Return 2884ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 2885ab127040SQuentin Monnet * 2886bdb7b79bSAndrii Nakryiko * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 2887ab127040SQuentin Monnet * Description 2888ab127040SQuentin Monnet * For socket policies, apply the verdict of the eBPF program to 2889ab127040SQuentin Monnet * the next *bytes* (number of bytes) of message *msg*. 2890ab127040SQuentin Monnet * 2891ab127040SQuentin Monnet * For example, this helper can be used in the following cases: 2892ab127040SQuentin Monnet * 2893ab127040SQuentin Monnet * * A single **sendmsg**\ () or **sendfile**\ () system call 2894ab127040SQuentin Monnet * contains multiple logical messages that the eBPF program is 2895ab127040SQuentin Monnet * supposed to read and for which it should apply a verdict. 2896ab127040SQuentin Monnet * * An eBPF program only cares to read the first *bytes* of a 2897ab127040SQuentin Monnet * *msg*. If the message has a large payload, then setting up 2898ab127040SQuentin Monnet * and calling the eBPF program repeatedly for all bytes, even 2899ab127040SQuentin Monnet * though the verdict is already known, would create unnecessary 2900ab127040SQuentin Monnet * overhead. 2901ab127040SQuentin Monnet * 2902ab127040SQuentin Monnet * When called from within an eBPF program, the helper sets a 2903ab127040SQuentin Monnet * counter internal to the BPF infrastructure, that is used to 2904ab127040SQuentin Monnet * apply the last verdict to the next *bytes*. If *bytes* is 2905ab127040SQuentin Monnet * smaller than the current data being processed from a 2906ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () system call, the first 2907ab127040SQuentin Monnet * *bytes* will be sent and the eBPF program will be re-run with 2908ab127040SQuentin Monnet * the pointer for start of data pointing to byte number *bytes* 2909ab127040SQuentin Monnet * **+ 1**. If *bytes* is larger than the current data being 2910ab127040SQuentin Monnet * processed, then the eBPF verdict will be applied to multiple 2911ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2912ab127040SQuentin Monnet * consumed. 2913ab127040SQuentin Monnet * 2914ab127040SQuentin Monnet * Note that if a socket closes with the internal counter holding 2915ab127040SQuentin Monnet * a non-zero value, this is not a problem because data is not 2916ab127040SQuentin Monnet * being buffered for *bytes* and is sent as it is received. 2917ab127040SQuentin Monnet * Return 2918ab127040SQuentin Monnet * 0 2919ab127040SQuentin Monnet * 2920bdb7b79bSAndrii Nakryiko * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 2921ab127040SQuentin Monnet * Description 2922ab127040SQuentin Monnet * For socket policies, prevent the execution of the verdict eBPF 2923ab127040SQuentin Monnet * program for message *msg* until *bytes* (byte number) have been 2924ab127040SQuentin Monnet * accumulated. 2925ab127040SQuentin Monnet * 2926ab127040SQuentin Monnet * This can be used when one needs a specific number of bytes 2927ab127040SQuentin Monnet * before a verdict can be assigned, even if the data spans 2928ab127040SQuentin Monnet * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 2929ab127040SQuentin Monnet * case would be a user calling **sendmsg**\ () repeatedly with 2930ab127040SQuentin Monnet * 1-byte long message segments. Obviously, this is bad for 2931ab127040SQuentin Monnet * performance, but it is still valid. If the eBPF program needs 2932ab127040SQuentin Monnet * *bytes* bytes to validate a header, this helper can be used to 2933ab127040SQuentin Monnet * prevent the eBPF program to be called again until *bytes* have 2934ab127040SQuentin Monnet * been accumulated. 2935ab127040SQuentin Monnet * Return 2936ab127040SQuentin Monnet * 0 2937ab127040SQuentin Monnet * 2938bdb7b79bSAndrii Nakryiko * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2939ab127040SQuentin Monnet * Description 2940ab127040SQuentin Monnet * For socket policies, pull in non-linear data from user space 2941ab127040SQuentin Monnet * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2942ab127040SQuentin Monnet * **->data_end** to *start* and *end* bytes offsets into *msg*, 2943ab127040SQuentin Monnet * respectively. 2944ab127040SQuentin Monnet * 2945ab127040SQuentin Monnet * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2946ab127040SQuentin Monnet * *msg* it can only parse data that the (**data**, **data_end**) 2947ab127040SQuentin Monnet * pointers have already consumed. For **sendmsg**\ () hooks this 2948ab127040SQuentin Monnet * is likely the first scatterlist element. But for calls relying 2949ab127040SQuentin Monnet * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2950ab127040SQuentin Monnet * be the range (**0**, **0**) because the data is shared with 2951ab127040SQuentin Monnet * user space and by default the objective is to avoid allowing 2952ab127040SQuentin Monnet * user space to modify data while (or after) eBPF verdict is 2953ab127040SQuentin Monnet * being decided. This helper can be used to pull in data and to 2954ab127040SQuentin Monnet * set the start and end pointer to given values. Data will be 2955ab127040SQuentin Monnet * copied if necessary (i.e. if data was not linear and if start 2956ab127040SQuentin Monnet * and end pointers do not point to the same chunk). 2957ab127040SQuentin Monnet * 295832e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2959ab127040SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2960ab127040SQuentin Monnet * previously done by the verifier are invalidated and must be 2961ab127040SQuentin Monnet * performed again, if the helper is used in combination with 2962ab127040SQuentin Monnet * direct packet access. 2963ab127040SQuentin Monnet * 2964ab127040SQuentin Monnet * All values for *flags* are reserved for future usage, and must 2965ab127040SQuentin Monnet * be left at zero. 2966ab127040SQuentin Monnet * Return 2967ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 2968ab127040SQuentin Monnet * 2969bdb7b79bSAndrii Nakryiko * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 29707aa79a86SQuentin Monnet * Description 29717aa79a86SQuentin Monnet * Bind the socket associated to *ctx* to the address pointed by 29727aa79a86SQuentin Monnet * *addr*, of length *addr_len*. This allows for making outgoing 29737aa79a86SQuentin Monnet * connection from the desired IP address, which can be useful for 29747aa79a86SQuentin Monnet * example when all processes inside a cgroup should use one 29757aa79a86SQuentin Monnet * single IP address on a host that has multiple IP configured. 29767aa79a86SQuentin Monnet * 29777aa79a86SQuentin Monnet * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 29787aa79a86SQuentin Monnet * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 29798086fbafSStanislav Fomichev * **AF_INET6**). It's advised to pass zero port (**sin_port** 29808086fbafSStanislav Fomichev * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 29818086fbafSStanislav Fomichev * behavior and lets the kernel efficiently pick up an unused 29828086fbafSStanislav Fomichev * port as long as 4-tuple is unique. Passing non-zero port might 29838086fbafSStanislav Fomichev * lead to degraded performance. 29847aa79a86SQuentin Monnet * Return 29857aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 29862d020dd7SQuentin Monnet * 2987bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 29882d020dd7SQuentin Monnet * Description 29892d020dd7SQuentin Monnet * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2990c8741e2bSJesper Dangaard Brouer * possible to both shrink and grow the packet tail. 2991c8741e2bSJesper Dangaard Brouer * Shrink done via *delta* being a negative integer. 29922d020dd7SQuentin Monnet * 299332e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 29942d020dd7SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 29952d020dd7SQuentin Monnet * previously done by the verifier are invalidated and must be 29962d020dd7SQuentin Monnet * performed again, if the helper is used in combination with 29972d020dd7SQuentin Monnet * direct packet access. 29982d020dd7SQuentin Monnet * Return 29992d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 30002d020dd7SQuentin Monnet * 3001bdb7b79bSAndrii Nakryiko * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 30022d020dd7SQuentin Monnet * Description 30032d020dd7SQuentin Monnet * Retrieve the XFRM state (IP transform framework, see also 30042d020dd7SQuentin Monnet * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 30052d020dd7SQuentin Monnet * 30062d020dd7SQuentin Monnet * The retrieved value is stored in the **struct bpf_xfrm_state** 30072d020dd7SQuentin Monnet * pointed by *xfrm_state* and of length *size*. 30082d020dd7SQuentin Monnet * 30092d020dd7SQuentin Monnet * All values for *flags* are reserved for future usage, and must 30102d020dd7SQuentin Monnet * be left at zero. 30112d020dd7SQuentin Monnet * 30122d020dd7SQuentin Monnet * This helper is available only if the kernel was compiled with 30132d020dd7SQuentin Monnet * **CONFIG_XFRM** configuration option. 30142d020dd7SQuentin Monnet * Return 30152d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 3016c195651eSYonghong Song * 3017bdb7b79bSAndrii Nakryiko * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 3018c195651eSYonghong Song * Description 3019c195651eSYonghong Song * Return a user or a kernel stack in bpf program provided buffer. 3020c195651eSYonghong Song * To achieve this, the helper needs *ctx*, which is a pointer 3021c195651eSYonghong Song * to the context on which the tracing program is executed. 3022c195651eSYonghong Song * To store the stacktrace, the bpf program provides *buf* with 3023c195651eSYonghong Song * a nonnegative *size*. 3024c195651eSYonghong Song * 3025c195651eSYonghong Song * The last argument, *flags*, holds the number of stack frames to 3026c195651eSYonghong Song * skip (from 0 to 255), masked with 3027c195651eSYonghong Song * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3028c195651eSYonghong Song * the following flags: 3029c195651eSYonghong Song * 3030c195651eSYonghong Song * **BPF_F_USER_STACK** 3031c195651eSYonghong Song * Collect a user space stack instead of a kernel stack. 3032c195651eSYonghong Song * **BPF_F_USER_BUILD_ID** 3033ca34ce29SDave Marchevsky * Collect (build_id, file_offset) instead of ips for user 3034ca34ce29SDave Marchevsky * stack, only valid if **BPF_F_USER_STACK** is also 3035ca34ce29SDave Marchevsky * specified. 3036ca34ce29SDave Marchevsky * 3037ca34ce29SDave Marchevsky * *file_offset* is an offset relative to the beginning 3038ca34ce29SDave Marchevsky * of the executable or shared object file backing the vma 3039ca34ce29SDave Marchevsky * which the *ip* falls in. It is *not* an offset relative 3040ca34ce29SDave Marchevsky * to that object's base address. Accordingly, it must be 3041ca34ce29SDave Marchevsky * adjusted by adding (sh_addr - sh_offset), where 3042ca34ce29SDave Marchevsky * sh_{addr,offset} correspond to the executable section 3043ca34ce29SDave Marchevsky * containing *file_offset* in the object, for comparisons 3044ca34ce29SDave Marchevsky * to symbols' st_value to be valid. 3045c195651eSYonghong Song * 3046c195651eSYonghong Song * **bpf_get_stack**\ () can collect up to 3047c195651eSYonghong Song * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3048c195651eSYonghong Song * to sufficient large buffer size. Note that 3049c195651eSYonghong Song * this limit can be controlled with the **sysctl** program, and 3050c195651eSYonghong Song * that it should be manually increased in order to profile long 3051c195651eSYonghong Song * user stacks (such as stacks for Java programs). To do so, use: 3052c195651eSYonghong Song * 3053c195651eSYonghong Song * :: 3054c195651eSYonghong Song * 3055c195651eSYonghong Song * # sysctl kernel.perf_event_max_stack=<new value> 3056c195651eSYonghong Song * Return 3057ee2a0988SNamhyung Kim * The non-negative copied *buf* length equal to or less than 3058ee2a0988SNamhyung Kim * *size* on success, or a negative error in case of failure. 30594e1ec56cSDaniel Borkmann * 3060bdb7b79bSAndrii Nakryiko * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 30614e1ec56cSDaniel Borkmann * Description 30624e1ec56cSDaniel Borkmann * This helper is similar to **bpf_skb_load_bytes**\ () in that 30634e1ec56cSDaniel Borkmann * it provides an easy way to load *len* bytes from *offset* 30644e1ec56cSDaniel Borkmann * from the packet associated to *skb*, into the buffer pointed 30654e1ec56cSDaniel Borkmann * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 30664e1ec56cSDaniel Borkmann * a fifth argument *start_header* exists in order to select a 30674e1ec56cSDaniel Borkmann * base offset to start from. *start_header* can be one of: 30684e1ec56cSDaniel Borkmann * 30694e1ec56cSDaniel Borkmann * **BPF_HDR_START_MAC** 30704e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s mac header. 30714e1ec56cSDaniel Borkmann * **BPF_HDR_START_NET** 30724e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s network header. 30734e1ec56cSDaniel Borkmann * 30744e1ec56cSDaniel Borkmann * In general, "direct packet access" is the preferred method to 30754e1ec56cSDaniel Borkmann * access packet data, however, this helper is in particular useful 30764e1ec56cSDaniel Borkmann * in socket filters where *skb*\ **->data** does not always point 30774e1ec56cSDaniel Borkmann * to the start of the mac header and where "direct packet access" 30784e1ec56cSDaniel Borkmann * is not available. 30794e1ec56cSDaniel Borkmann * Return 30804e1ec56cSDaniel Borkmann * 0 on success, or a negative error in case of failure. 30814e1ec56cSDaniel Borkmann * 3082bdb7b79bSAndrii Nakryiko * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 308387f5fc7eSDavid Ahern * Description 308487f5fc7eSDavid Ahern * Do FIB lookup in kernel tables using parameters in *params*. 308587f5fc7eSDavid Ahern * If lookup is successful and result shows packet is to be 308687f5fc7eSDavid Ahern * forwarded, the neighbor tables are searched for the nexthop. 308787f5fc7eSDavid Ahern * If successful (ie., FIB lookup shows forwarding and nexthop 3088fa898d76SDavid Ahern * is resolved), the nexthop address is returned in ipv4_dst 3089fa898d76SDavid Ahern * or ipv6_dst based on family, smac is set to mac address of 3090fa898d76SDavid Ahern * egress device, dmac is set to nexthop mac address, rt_metric 30914c79579bSDavid Ahern * is set to metric from route (IPv4/IPv6 only), and ifindex 30924c79579bSDavid Ahern * is set to the device index of the nexthop from the FIB lookup. 309387f5fc7eSDavid Ahern * 309487f5fc7eSDavid Ahern * *plen* argument is the size of the passed in struct. 30957a279e93SQuentin Monnet * *flags* argument can be a combination of one or more of the 30967a279e93SQuentin Monnet * following values: 309787f5fc7eSDavid Ahern * 30987a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_DIRECT** 30997a279e93SQuentin Monnet * Do a direct table lookup vs full lookup using FIB 31007a279e93SQuentin Monnet * rules. 31017a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_OUTPUT** 31027a279e93SQuentin Monnet * Perform lookup from an egress perspective (default is 31037a279e93SQuentin Monnet * ingress). 310487f5fc7eSDavid Ahern * 310587f5fc7eSDavid Ahern * *ctx* is either **struct xdp_md** for XDP programs or 310687f5fc7eSDavid Ahern * **struct sk_buff** tc cls_act programs. 310787f5fc7eSDavid Ahern * Return 31084c79579bSDavid Ahern * * < 0 if any input argument is invalid 31094c79579bSDavid Ahern * * 0 on success (packet is forwarded, nexthop neighbor exists) 31104c79579bSDavid Ahern * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 31112bae79d2SQuentin Monnet * packet is not forwarded or needs assist from full stack 311281110384SJohn Fastabend * 3113e1850ea9SJesper Dangaard Brouer * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3114e1850ea9SJesper Dangaard Brouer * was exceeded and output params->mtu_result contains the MTU. 3115e1850ea9SJesper Dangaard Brouer * 3116bdb7b79bSAndrii Nakryiko * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 311781110384SJohn Fastabend * Description 311881110384SJohn Fastabend * Add an entry to, or update a sockhash *map* referencing sockets. 311981110384SJohn Fastabend * The *skops* is used as a new value for the entry associated to 312081110384SJohn Fastabend * *key*. *flags* is one of: 312181110384SJohn Fastabend * 312281110384SJohn Fastabend * **BPF_NOEXIST** 312381110384SJohn Fastabend * The entry for *key* must not exist in the map. 312481110384SJohn Fastabend * **BPF_EXIST** 312581110384SJohn Fastabend * The entry for *key* must already exist in the map. 312681110384SJohn Fastabend * **BPF_ANY** 312781110384SJohn Fastabend * No condition on the existence of the entry for *key*. 312881110384SJohn Fastabend * 312981110384SJohn Fastabend * If the *map* has eBPF programs (parser and verdict), those will 313081110384SJohn Fastabend * be inherited by the socket being added. If the socket is 313181110384SJohn Fastabend * already attached to eBPF programs, this results in an error. 313281110384SJohn Fastabend * Return 313381110384SJohn Fastabend * 0 on success, or a negative error in case of failure. 313481110384SJohn Fastabend * 3135bdb7b79bSAndrii Nakryiko * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 313681110384SJohn Fastabend * Description 313781110384SJohn Fastabend * This helper is used in programs implementing policies at the 313881110384SJohn Fastabend * socket level. If the message *msg* is allowed to pass (i.e. if 313981110384SJohn Fastabend * the verdict eBPF program returns **SK_PASS**), redirect it to 314081110384SJohn Fastabend * the socket referenced by *map* (of type 314181110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 314281110384SJohn Fastabend * egress interfaces can be used for redirection. The 314381110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 314481110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 314581110384SJohn Fastabend * egress path otherwise). This is the only flag supported for now. 314681110384SJohn Fastabend * Return 314781110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 314881110384SJohn Fastabend * 3149bdb7b79bSAndrii Nakryiko * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 315081110384SJohn Fastabend * Description 315181110384SJohn Fastabend * This helper is used in programs implementing policies at the 315281110384SJohn Fastabend * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 315349f3d12bSJakub Wilk * if the verdict eBPF program returns **SK_PASS**), redirect it 315481110384SJohn Fastabend * to the socket referenced by *map* (of type 315581110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 315681110384SJohn Fastabend * egress interfaces can be used for redirection. The 315781110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 315881110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 315981110384SJohn Fastabend * egress otherwise). This is the only flag supported for now. 316081110384SJohn Fastabend * Return 316181110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 3162fe94cc29SMathieu Xhonneux * 3163bdb7b79bSAndrii Nakryiko * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3164fe94cc29SMathieu Xhonneux * Description 3165fe94cc29SMathieu Xhonneux * Encapsulate the packet associated to *skb* within a Layer 3 3166fe94cc29SMathieu Xhonneux * protocol header. This header is provided in the buffer at 3167fe94cc29SMathieu Xhonneux * address *hdr*, with *len* its size in bytes. *type* indicates 3168fe94cc29SMathieu Xhonneux * the protocol of the header and can be one of: 3169fe94cc29SMathieu Xhonneux * 3170fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6** 3171fe94cc29SMathieu Xhonneux * IPv6 encapsulation with Segment Routing Header 3172fe94cc29SMathieu Xhonneux * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3173fe94cc29SMathieu Xhonneux * the IPv6 header is computed by the kernel. 3174fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6_INLINE** 3175fe94cc29SMathieu Xhonneux * Only works if *skb* contains an IPv6 packet. Insert a 3176fe94cc29SMathieu Xhonneux * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3177fe94cc29SMathieu Xhonneux * the IPv6 header. 31783e0bd37cSPeter Oskolkov * **BPF_LWT_ENCAP_IP** 31793e0bd37cSPeter Oskolkov * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 31803e0bd37cSPeter Oskolkov * must be IPv4 or IPv6, followed by zero or more 318180867c5eSQuentin Monnet * additional headers, up to **LWT_BPF_MAX_HEADROOM** 318280867c5eSQuentin Monnet * total bytes in all prepended headers. Please note that 318380867c5eSQuentin Monnet * if **skb_is_gso**\ (*skb*) is true, no more than two 318480867c5eSQuentin Monnet * headers can be prepended, and the inner header, if 318580867c5eSQuentin Monnet * present, should be either GRE or UDP/GUE. 31863e0bd37cSPeter Oskolkov * 318780867c5eSQuentin Monnet * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 318880867c5eSQuentin Monnet * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 318980867c5eSQuentin Monnet * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 319080867c5eSQuentin Monnet * **BPF_PROG_TYPE_LWT_XMIT**. 3191fe94cc29SMathieu Xhonneux * 319232e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3193fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3194fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3195fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3196fe94cc29SMathieu Xhonneux * direct packet access. 3197fe94cc29SMathieu Xhonneux * Return 3198fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3199fe94cc29SMathieu Xhonneux * 3200bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3201fe94cc29SMathieu Xhonneux * Description 3202fe94cc29SMathieu Xhonneux * Store *len* bytes from address *from* into the packet 3203fe94cc29SMathieu Xhonneux * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3204fe94cc29SMathieu Xhonneux * inside the outermost IPv6 Segment Routing Header can be 3205fe94cc29SMathieu Xhonneux * modified through this helper. 3206fe94cc29SMathieu Xhonneux * 320732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3208fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3209fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3210fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3211fe94cc29SMathieu Xhonneux * direct packet access. 3212fe94cc29SMathieu Xhonneux * Return 3213fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3214fe94cc29SMathieu Xhonneux * 3215bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3216fe94cc29SMathieu Xhonneux * Description 3217fe94cc29SMathieu Xhonneux * Adjust the size allocated to TLVs in the outermost IPv6 3218fe94cc29SMathieu Xhonneux * Segment Routing Header contained in the packet associated to 3219fe94cc29SMathieu Xhonneux * *skb*, at position *offset* by *delta* bytes. Only offsets 3220fe94cc29SMathieu Xhonneux * after the segments are accepted. *delta* can be as well 3221fe94cc29SMathieu Xhonneux * positive (growing) as negative (shrinking). 3222fe94cc29SMathieu Xhonneux * 322332e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3224fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3225fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3226fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3227fe94cc29SMathieu Xhonneux * direct packet access. 3228fe94cc29SMathieu Xhonneux * Return 3229fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3230fe94cc29SMathieu Xhonneux * 3231bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3232fe94cc29SMathieu Xhonneux * Description 3233fe94cc29SMathieu Xhonneux * Apply an IPv6 Segment Routing action of type *action* to the 3234fe94cc29SMathieu Xhonneux * packet associated to *skb*. Each action takes a parameter 3235fe94cc29SMathieu Xhonneux * contained at address *param*, and of length *param_len* bytes. 3236fe94cc29SMathieu Xhonneux * *action* can be one of: 3237fe94cc29SMathieu Xhonneux * 3238fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_X** 3239fe94cc29SMathieu Xhonneux * End.X action: Endpoint with Layer-3 cross-connect. 3240fe94cc29SMathieu Xhonneux * Type of *param*: **struct in6_addr**. 3241fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_T** 3242fe94cc29SMathieu Xhonneux * End.T action: Endpoint with specific IPv6 table lookup. 3243fe94cc29SMathieu Xhonneux * Type of *param*: **int**. 3244fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6** 3245fe94cc29SMathieu Xhonneux * End.B6 action: Endpoint bound to an SRv6 policy. 324680867c5eSQuentin Monnet * Type of *param*: **struct ipv6_sr_hdr**. 3247fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3248fe94cc29SMathieu Xhonneux * End.B6.Encap action: Endpoint bound to an SRv6 3249fe94cc29SMathieu Xhonneux * encapsulation policy. 325080867c5eSQuentin Monnet * Type of *param*: **struct ipv6_sr_hdr**. 3251fe94cc29SMathieu Xhonneux * 325232e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 3253fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 3254fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 3255fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 3256fe94cc29SMathieu Xhonneux * direct packet access. 3257fe94cc29SMathieu Xhonneux * Return 3258fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 3259f4364dcfSSean Young * 3260bdb7b79bSAndrii Nakryiko * long bpf_rc_repeat(void *ctx) 326162369db2SQuentin Monnet * Description 326262369db2SQuentin Monnet * This helper is used in programs implementing IR decoding, to 326362369db2SQuentin Monnet * report a successfully decoded repeat key message. This delays 326462369db2SQuentin Monnet * the generation of a key up event for previously generated 326562369db2SQuentin Monnet * key down event. 326662369db2SQuentin Monnet * 326762369db2SQuentin Monnet * Some IR protocols like NEC have a special IR message for 326862369db2SQuentin Monnet * repeating last button, for when a button is held down. 326962369db2SQuentin Monnet * 327062369db2SQuentin Monnet * The *ctx* should point to the lirc sample as passed into 327162369db2SQuentin Monnet * the program. 327262369db2SQuentin Monnet * 327362369db2SQuentin Monnet * This helper is only available is the kernel was compiled with 327462369db2SQuentin Monnet * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 327562369db2SQuentin Monnet * "**y**". 327662369db2SQuentin Monnet * Return 327762369db2SQuentin Monnet * 0 327862369db2SQuentin Monnet * 3279bdb7b79bSAndrii Nakryiko * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3280f4364dcfSSean Young * Description 3281f4364dcfSSean Young * This helper is used in programs implementing IR decoding, to 3282f4364dcfSSean Young * report a successfully decoded key press with *scancode*, 3283f4364dcfSSean Young * *toggle* value in the given *protocol*. The scancode will be 3284f4364dcfSSean Young * translated to a keycode using the rc keymap, and reported as 3285f4364dcfSSean Young * an input key down event. After a period a key up event is 3286f4364dcfSSean Young * generated. This period can be extended by calling either 328790b1023fSQuentin Monnet * **bpf_rc_keydown**\ () again with the same values, or calling 328890b1023fSQuentin Monnet * **bpf_rc_repeat**\ (). 3289f4364dcfSSean Young * 3290f4364dcfSSean Young * Some protocols include a toggle bit, in case the button was 3291f4364dcfSSean Young * released and pressed again between consecutive scancodes. 3292f4364dcfSSean Young * 3293f4364dcfSSean Young * The *ctx* should point to the lirc sample as passed into 3294f4364dcfSSean Young * the program. 3295f4364dcfSSean Young * 3296f4364dcfSSean Young * The *protocol* is the decoded protocol number (see 3297f4364dcfSSean Young * **enum rc_proto** for some predefined values). 3298f4364dcfSSean Young * 3299f4364dcfSSean Young * This helper is only available is the kernel was compiled with 3300f4364dcfSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3301f4364dcfSSean Young * "**y**". 3302f4364dcfSSean Young * Return 3303f4364dcfSSean Young * 0 3304f4364dcfSSean Young * 330562369db2SQuentin Monnet * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3306cb20b08eSDaniel Borkmann * Description 3307cb20b08eSDaniel Borkmann * Return the cgroup v2 id of the socket associated with the *skb*. 3308cb20b08eSDaniel Borkmann * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3309cb20b08eSDaniel Borkmann * helper for cgroup v1 by providing a tag resp. identifier that 3310cb20b08eSDaniel Borkmann * can be matched on or used for map lookups e.g. to implement 3311cb20b08eSDaniel Borkmann * policy. The cgroup v2 id of a given path in the hierarchy is 3312cb20b08eSDaniel Borkmann * exposed in user space through the f_handle API in order to get 3313cb20b08eSDaniel Borkmann * to the same 64-bit id. 3314cb20b08eSDaniel Borkmann * 3315cb20b08eSDaniel Borkmann * This helper can be used on TC egress path, but not on ingress, 3316cb20b08eSDaniel Borkmann * and is available only if the kernel was compiled with the 3317cb20b08eSDaniel Borkmann * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3318cb20b08eSDaniel Borkmann * Return 3319cb20b08eSDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 3320bf6fa2c8SYonghong Song * 3321bf6fa2c8SYonghong Song * u64 bpf_get_current_cgroup_id(void) 3322e40fbbf0SUsama Arif * Description 3323e40fbbf0SUsama Arif * Get the current cgroup id based on the cgroup within which 3324e40fbbf0SUsama Arif * the current task is running. 3325bf6fa2c8SYonghong Song * Return 3326bf6fa2c8SYonghong Song * A 64-bit integer containing the current cgroup id based 3327bf6fa2c8SYonghong Song * on the cgroup within which the current task is running. 3328cd339431SRoman Gushchin * 332962369db2SQuentin Monnet * void *bpf_get_local_storage(void *map, u64 flags) 3330cd339431SRoman Gushchin * Description 3331cd339431SRoman Gushchin * Get the pointer to the local storage area. 3332cd339431SRoman Gushchin * The type and the size of the local storage is defined 3333cd339431SRoman Gushchin * by the *map* argument. 3334cd339431SRoman Gushchin * The *flags* meaning is specific for each map type, 3335cd339431SRoman Gushchin * and has to be 0 for cgroup local storage. 3336cd339431SRoman Gushchin * 333790b1023fSQuentin Monnet * Depending on the BPF program type, a local storage area 333890b1023fSQuentin Monnet * can be shared between multiple instances of the BPF program, 3339cd339431SRoman Gushchin * running simultaneously. 3340cd339431SRoman Gushchin * 3341cd339431SRoman Gushchin * A user should care about the synchronization by himself. 334291c960b0SBrendan Jackman * For example, by using the **BPF_ATOMIC** instructions to alter 3343cd339431SRoman Gushchin * the shared data. 3344cd339431SRoman Gushchin * Return 334590b1023fSQuentin Monnet * A pointer to the local storage area. 33462dbb9b9eSMartin KaFai Lau * 3347bdb7b79bSAndrii Nakryiko * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 33482dbb9b9eSMartin KaFai Lau * Description 334990b1023fSQuentin Monnet * Select a **SO_REUSEPORT** socket from a 3350f170acdaSKuniyuki Iwashima * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 335190b1023fSQuentin Monnet * It checks the selected socket is matching the incoming 335290b1023fSQuentin Monnet * request in the socket buffer. 33532dbb9b9eSMartin KaFai Lau * Return 33542dbb9b9eSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 33556acc9b43SJoe Stringer * 335662369db2SQuentin Monnet * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 335762369db2SQuentin Monnet * Description 335862369db2SQuentin Monnet * Return id of cgroup v2 that is ancestor of cgroup associated 335962369db2SQuentin Monnet * with the *skb* at the *ancestor_level*. The root cgroup is at 336062369db2SQuentin Monnet * *ancestor_level* zero and each step down the hierarchy 336162369db2SQuentin Monnet * increments the level. If *ancestor_level* == level of cgroup 336262369db2SQuentin Monnet * associated with *skb*, then return value will be same as that 336362369db2SQuentin Monnet * of **bpf_skb_cgroup_id**\ (). 336462369db2SQuentin Monnet * 336562369db2SQuentin Monnet * The helper is useful to implement policies based on cgroups 336662369db2SQuentin Monnet * that are upper in hierarchy than immediate cgroup associated 336762369db2SQuentin Monnet * with *skb*. 336862369db2SQuentin Monnet * 336962369db2SQuentin Monnet * The format of returned id and helper limitations are same as in 337062369db2SQuentin Monnet * **bpf_skb_cgroup_id**\ (). 337162369db2SQuentin Monnet * Return 337262369db2SQuentin Monnet * The id is returned or 0 in case the id could not be retrieved. 337362369db2SQuentin Monnet * 3374f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 33756acc9b43SJoe Stringer * Description 33766acc9b43SJoe Stringer * Look for TCP socket matching *tuple*, optionally in a child 33776acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 337890b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 33796acc9b43SJoe Stringer * 33806acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 33816acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 33826acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 33836acc9b43SJoe Stringer * 33846acc9b43SJoe Stringer * *tuple_size* must be one of: 33856acc9b43SJoe Stringer * 33866acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 33876acc9b43SJoe Stringer * Look for an IPv4 socket. 33886acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 33896acc9b43SJoe Stringer * Look for an IPv6 socket. 33906acc9b43SJoe Stringer * 3391f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 3392bfdfa517SRandy Dunlap * socket lookup table in the netns associated with the *ctx* 3393f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 3394f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 3395f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 3396f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 3397f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 3398f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 33996acc9b43SJoe Stringer * 34006acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 34016acc9b43SJoe Stringer * be left at zero. 34026acc9b43SJoe Stringer * 34036acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 34046acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 34056acc9b43SJoe Stringer * Return 34060bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 34070bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 340880867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 340980867c5eSQuentin Monnet * tuple. 34106acc9b43SJoe Stringer * 3411f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 34126acc9b43SJoe Stringer * Description 34136acc9b43SJoe Stringer * Look for UDP socket matching *tuple*, optionally in a child 34146acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 341590b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 34166acc9b43SJoe Stringer * 34176acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 34186acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 34196acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 34206acc9b43SJoe Stringer * 34216acc9b43SJoe Stringer * *tuple_size* must be one of: 34226acc9b43SJoe Stringer * 34236acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 34246acc9b43SJoe Stringer * Look for an IPv4 socket. 34256acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 34266acc9b43SJoe Stringer * Look for an IPv6 socket. 34276acc9b43SJoe Stringer * 3428f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 3429bfdfa517SRandy Dunlap * socket lookup table in the netns associated with the *ctx* 3430f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 3431f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 3432f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 3433f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 3434f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 3435f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 34366acc9b43SJoe Stringer * 34376acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 34386acc9b43SJoe Stringer * be left at zero. 34396acc9b43SJoe Stringer * 34406acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 34416acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 34426acc9b43SJoe Stringer * Return 34430bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 34440bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 344580867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 344680867c5eSQuentin Monnet * tuple. 34476acc9b43SJoe Stringer * 3448a5fa25adSMartin KaFai Lau * long bpf_sk_release(void *sock) 34496acc9b43SJoe Stringer * Description 345090b1023fSQuentin Monnet * Release the reference held by *sock*. *sock* must be a 345190b1023fSQuentin Monnet * non-**NULL** pointer that was returned from 345290b1023fSQuentin Monnet * **bpf_sk_lookup_xxx**\ (). 345390b1023fSQuentin Monnet * Return 345490b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 345590b1023fSQuentin Monnet * 3456bdb7b79bSAndrii Nakryiko * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 345762369db2SQuentin Monnet * Description 345862369db2SQuentin Monnet * Push an element *value* in *map*. *flags* is one of: 345962369db2SQuentin Monnet * 346062369db2SQuentin Monnet * **BPF_EXIST** 346162369db2SQuentin Monnet * If the queue/stack is full, the oldest element is 346262369db2SQuentin Monnet * removed to make room for this. 346362369db2SQuentin Monnet * Return 346462369db2SQuentin Monnet * 0 on success, or a negative error in case of failure. 346562369db2SQuentin Monnet * 3466bdb7b79bSAndrii Nakryiko * long bpf_map_pop_elem(struct bpf_map *map, void *value) 346790b1023fSQuentin Monnet * Description 346890b1023fSQuentin Monnet * Pop an element from *map*. 346990b1023fSQuentin Monnet * Return 347090b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 347190b1023fSQuentin Monnet * 3472bdb7b79bSAndrii Nakryiko * long bpf_map_peek_elem(struct bpf_map *map, void *value) 347390b1023fSQuentin Monnet * Description 347490b1023fSQuentin Monnet * Get an element from *map* without removing it. 34756acc9b43SJoe Stringer * Return 34766acc9b43SJoe Stringer * 0 on success, or a negative error in case of failure. 34776fff607eSJohn Fastabend * 3478bdb7b79bSAndrii Nakryiko * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 34796fff607eSJohn Fastabend * Description 348090b1023fSQuentin Monnet * For socket policies, insert *len* bytes into *msg* at offset 34816fff607eSJohn Fastabend * *start*. 34826fff607eSJohn Fastabend * 34836fff607eSJohn Fastabend * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 348490b1023fSQuentin Monnet * *msg* it may want to insert metadata or options into the *msg*. 34856fff607eSJohn Fastabend * This can later be read and used by any of the lower layer BPF 34866fff607eSJohn Fastabend * hooks. 34876fff607eSJohn Fastabend * 34886fff607eSJohn Fastabend * This helper may fail if under memory pressure (a malloc 34896fff607eSJohn Fastabend * fails) in these cases BPF programs will get an appropriate 34906fff607eSJohn Fastabend * error and BPF programs will need to handle them. 34916fff607eSJohn Fastabend * Return 34926fff607eSJohn Fastabend * 0 on success, or a negative error in case of failure. 34937246d8edSJohn Fastabend * 3494bdb7b79bSAndrii Nakryiko * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 34957246d8edSJohn Fastabend * Description 34965f0e5412SAndrii Nakryiko * Will remove *len* bytes from a *msg* starting at byte *start*. 34977246d8edSJohn Fastabend * This may result in **ENOMEM** errors under certain situations if 34987246d8edSJohn Fastabend * an allocation and copy are required due to a full ring buffer. 34997246d8edSJohn Fastabend * However, the helper will try to avoid doing the allocation 35007246d8edSJohn Fastabend * if possible. Other errors can occur if input parameters are 350190b1023fSQuentin Monnet * invalid either due to *start* byte not being valid part of *msg* 35027246d8edSJohn Fastabend * payload and/or *pop* value being to large. 35037246d8edSJohn Fastabend * Return 350490b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 350501d3240aSSean Young * 3506bdb7b79bSAndrii Nakryiko * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 350701d3240aSSean Young * Description 350801d3240aSSean Young * This helper is used in programs implementing IR decoding, to 350901d3240aSSean Young * report a successfully decoded pointer movement. 351001d3240aSSean Young * 351101d3240aSSean Young * The *ctx* should point to the lirc sample as passed into 351201d3240aSSean Young * the program. 351301d3240aSSean Young * 351401d3240aSSean Young * This helper is only available is the kernel was compiled with 351501d3240aSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 351601d3240aSSean Young * "**y**". 351701d3240aSSean Young * Return 351801d3240aSSean Young * 0 351946f8bc92SMartin KaFai Lau * 3520bdb7b79bSAndrii Nakryiko * long bpf_spin_lock(struct bpf_spin_lock *lock) 35210eb09785SQuentin Monnet * Description 35220eb09785SQuentin Monnet * Acquire a spinlock represented by the pointer *lock*, which is 35230eb09785SQuentin Monnet * stored as part of a value of a map. Taking the lock allows to 35240eb09785SQuentin Monnet * safely update the rest of the fields in that value. The 35250eb09785SQuentin Monnet * spinlock can (and must) later be released with a call to 35260eb09785SQuentin Monnet * **bpf_spin_unlock**\ (\ *lock*\ ). 35270eb09785SQuentin Monnet * 35280eb09785SQuentin Monnet * Spinlocks in BPF programs come with a number of restrictions 35290eb09785SQuentin Monnet * and constraints: 35300eb09785SQuentin Monnet * 35310eb09785SQuentin Monnet * * **bpf_spin_lock** objects are only allowed inside maps of 35320eb09785SQuentin Monnet * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 35330eb09785SQuentin Monnet * list could be extended in the future). 35340eb09785SQuentin Monnet * * BTF description of the map is mandatory. 35350eb09785SQuentin Monnet * * The BPF program can take ONE lock at a time, since taking two 35360eb09785SQuentin Monnet * or more could cause dead locks. 35370eb09785SQuentin Monnet * * Only one **struct bpf_spin_lock** is allowed per map element. 35380eb09785SQuentin Monnet * * When the lock is taken, calls (either BPF to BPF or helpers) 35390eb09785SQuentin Monnet * are not allowed. 35400eb09785SQuentin Monnet * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 35410eb09785SQuentin Monnet * allowed inside a spinlock-ed region. 35420eb09785SQuentin Monnet * * The BPF program MUST call **bpf_spin_unlock**\ () to release 35430eb09785SQuentin Monnet * the lock, on all execution paths, before it returns. 35440eb09785SQuentin Monnet * * The BPF program can access **struct bpf_spin_lock** only via 35450eb09785SQuentin Monnet * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 35460eb09785SQuentin Monnet * helpers. Loading or storing data into the **struct 35470eb09785SQuentin Monnet * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 35480eb09785SQuentin Monnet * * To use the **bpf_spin_lock**\ () helper, the BTF description 35490eb09785SQuentin Monnet * of the map value must be a struct and have **struct 35500eb09785SQuentin Monnet * bpf_spin_lock** *anyname*\ **;** field at the top level. 35510eb09785SQuentin Monnet * Nested lock inside another struct is not allowed. 35520eb09785SQuentin Monnet * * The **struct bpf_spin_lock** *lock* field in a map value must 35530eb09785SQuentin Monnet * be aligned on a multiple of 4 bytes in that value. 35540eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 35550eb09785SQuentin Monnet * the **bpf_spin_lock** field to user space. 35560eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 35570eb09785SQuentin Monnet * a BPF program, do not update the **bpf_spin_lock** field. 35580eb09785SQuentin Monnet * * **bpf_spin_lock** cannot be on the stack or inside a 35590eb09785SQuentin Monnet * networking packet (it can only be inside of a map values). 35600eb09785SQuentin Monnet * * **bpf_spin_lock** is available to root only. 35610eb09785SQuentin Monnet * * Tracing programs and socket filter programs cannot use 35620eb09785SQuentin Monnet * **bpf_spin_lock**\ () due to insufficient preemption checks 35630eb09785SQuentin Monnet * (but this may change in the future). 35640eb09785SQuentin Monnet * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 35650eb09785SQuentin Monnet * Return 35660eb09785SQuentin Monnet * 0 35670eb09785SQuentin Monnet * 3568bdb7b79bSAndrii Nakryiko * long bpf_spin_unlock(struct bpf_spin_lock *lock) 35690eb09785SQuentin Monnet * Description 35700eb09785SQuentin Monnet * Release the *lock* previously locked by a call to 35710eb09785SQuentin Monnet * **bpf_spin_lock**\ (\ *lock*\ ). 35720eb09785SQuentin Monnet * Return 35730eb09785SQuentin Monnet * 0 35740eb09785SQuentin Monnet * 357546f8bc92SMartin KaFai Lau * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 357646f8bc92SMartin KaFai Lau * Description 357746f8bc92SMartin KaFai Lau * This helper gets a **struct bpf_sock** pointer such 357862369db2SQuentin Monnet * that all the fields in this **bpf_sock** can be accessed. 357946f8bc92SMartin KaFai Lau * Return 358062369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 358146f8bc92SMartin KaFai Lau * case of failure. 3582655a51e5SMartin KaFai Lau * 3583655a51e5SMartin KaFai Lau * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3584655a51e5SMartin KaFai Lau * Description 3585655a51e5SMartin KaFai Lau * This helper gets a **struct bpf_tcp_sock** pointer from a 3586655a51e5SMartin KaFai Lau * **struct bpf_sock** pointer. 3587655a51e5SMartin KaFai Lau * Return 358862369db2SQuentin Monnet * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3589655a51e5SMartin KaFai Lau * case of failure. 3590f7c917baSbrakmo * 3591bdb7b79bSAndrii Nakryiko * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3592f7c917baSbrakmo * Description 359362369db2SQuentin Monnet * Set ECN (Explicit Congestion Notification) field of IP header 359462369db2SQuentin Monnet * to **CE** (Congestion Encountered) if current value is **ECT** 359562369db2SQuentin Monnet * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 359662369db2SQuentin Monnet * and IPv4. 3597f7c917baSbrakmo * Return 359862369db2SQuentin Monnet * 1 if the **CE** flag is set (either by the current helper call 359962369db2SQuentin Monnet * or because it was already present), 0 if it is not set. 3600dbafd7ddSMartin KaFai Lau * 3601dbafd7ddSMartin KaFai Lau * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3602dbafd7ddSMartin KaFai Lau * Description 360362369db2SQuentin Monnet * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 360462369db2SQuentin Monnet * **bpf_sk_release**\ () is unnecessary and not allowed. 3605dbafd7ddSMartin KaFai Lau * Return 360662369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 3607dbafd7ddSMartin KaFai Lau * case of failure. 3608edbf8c01SLorenz Bauer * 3609edbf8c01SLorenz Bauer * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3610edbf8c01SLorenz Bauer * Description 3611edbf8c01SLorenz Bauer * Look for TCP socket matching *tuple*, optionally in a child 3612edbf8c01SLorenz Bauer * network namespace *netns*. The return value must be checked, 3613edbf8c01SLorenz Bauer * and if non-**NULL**, released via **bpf_sk_release**\ (). 3614edbf8c01SLorenz Bauer * 361580867c5eSQuentin Monnet * This function is identical to **bpf_sk_lookup_tcp**\ (), except 361680867c5eSQuentin Monnet * that it also returns timewait or request sockets. Use 361780867c5eSQuentin Monnet * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 361880867c5eSQuentin Monnet * full structure. 3619edbf8c01SLorenz Bauer * 3620edbf8c01SLorenz Bauer * This helper is available only if the kernel was compiled with 3621edbf8c01SLorenz Bauer * **CONFIG_NET** configuration option. 3622edbf8c01SLorenz Bauer * Return 3623edbf8c01SLorenz Bauer * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3624edbf8c01SLorenz Bauer * For sockets with reuseport option, the **struct bpf_sock** 362580867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 362680867c5eSQuentin Monnet * tuple. 362739904084SLorenz Bauer * 3628c0df236eSMartin KaFai Lau * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 362939904084SLorenz Bauer * Description 363080867c5eSQuentin Monnet * Check whether *iph* and *th* contain a valid SYN cookie ACK for 363180867c5eSQuentin Monnet * the listening socket in *sk*. 363239904084SLorenz Bauer * 363380867c5eSQuentin Monnet * *iph* points to the start of the IPv4 or IPv6 header, while 363480867c5eSQuentin Monnet * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3635ac80287aSMaxim Mikityanskiy * **sizeof**\ (**struct ipv6hdr**). 363639904084SLorenz Bauer * 363780867c5eSQuentin Monnet * *th* points to the start of the TCP header, while *th_len* 3638ac80287aSMaxim Mikityanskiy * contains the length of the TCP header (at least 3639ac80287aSMaxim Mikityanskiy * **sizeof**\ (**struct tcphdr**)). 364039904084SLorenz Bauer * Return 364180867c5eSQuentin Monnet * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 364280867c5eSQuentin Monnet * error otherwise. 3643808649fbSAndrey Ignatov * 3644bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3645808649fbSAndrey Ignatov * Description 3646808649fbSAndrey Ignatov * Get name of sysctl in /proc/sys/ and copy it into provided by 3647808649fbSAndrey Ignatov * program buffer *buf* of size *buf_len*. 3648808649fbSAndrey Ignatov * 3649808649fbSAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 3650808649fbSAndrey Ignatov * 3651808649fbSAndrey Ignatov * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3652808649fbSAndrey Ignatov * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3653808649fbSAndrey Ignatov * only (e.g. "tcp_mem"). 3654808649fbSAndrey Ignatov * Return 3655808649fbSAndrey Ignatov * Number of character copied (not including the trailing NUL). 3656808649fbSAndrey Ignatov * 3657808649fbSAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3658808649fbSAndrey Ignatov * truncated name in this case). 36591d11b301SAndrey Ignatov * 3660bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 36611d11b301SAndrey Ignatov * Description 36621d11b301SAndrey Ignatov * Get current value of sysctl as it is presented in /proc/sys 36631d11b301SAndrey Ignatov * (incl. newline, etc), and copy it as a string into provided 36641d11b301SAndrey Ignatov * by program buffer *buf* of size *buf_len*. 36651d11b301SAndrey Ignatov * 36661d11b301SAndrey Ignatov * The whole value is copied, no matter what file position user 36671d11b301SAndrey Ignatov * space issued e.g. sys_read at. 36681d11b301SAndrey Ignatov * 36691d11b301SAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 36701d11b301SAndrey Ignatov * Return 36711d11b301SAndrey Ignatov * Number of character copied (not including the trailing NUL). 36721d11b301SAndrey Ignatov * 36731d11b301SAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 36741d11b301SAndrey Ignatov * truncated name in this case). 36751d11b301SAndrey Ignatov * 36761d11b301SAndrey Ignatov * **-EINVAL** if current value was unavailable, e.g. because 36771d11b301SAndrey Ignatov * sysctl is uninitialized and read returns -EIO for it. 36784e63acdfSAndrey Ignatov * 3679bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 36804e63acdfSAndrey Ignatov * Description 36814e63acdfSAndrey Ignatov * Get new value being written by user space to sysctl (before 36824e63acdfSAndrey Ignatov * the actual write happens) and copy it as a string into 36834e63acdfSAndrey Ignatov * provided by program buffer *buf* of size *buf_len*. 36844e63acdfSAndrey Ignatov * 36854e63acdfSAndrey Ignatov * User space may write new value at file position > 0. 36864e63acdfSAndrey Ignatov * 36874e63acdfSAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 36884e63acdfSAndrey Ignatov * Return 36894e63acdfSAndrey Ignatov * Number of character copied (not including the trailing NUL). 36904e63acdfSAndrey Ignatov * 36914e63acdfSAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 36924e63acdfSAndrey Ignatov * truncated name in this case). 36934e63acdfSAndrey Ignatov * 36944e63acdfSAndrey Ignatov * **-EINVAL** if sysctl is being read. 36954e63acdfSAndrey Ignatov * 3696bdb7b79bSAndrii Nakryiko * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 36974e63acdfSAndrey Ignatov * Description 36984e63acdfSAndrey Ignatov * Override new value being written by user space to sysctl with 36994e63acdfSAndrey Ignatov * value provided by program in buffer *buf* of size *buf_len*. 37004e63acdfSAndrey Ignatov * 37014e63acdfSAndrey Ignatov * *buf* should contain a string in same form as provided by user 37024e63acdfSAndrey Ignatov * space on sysctl write. 37034e63acdfSAndrey Ignatov * 37044e63acdfSAndrey Ignatov * User space may write new value at file position > 0. To override 37054e63acdfSAndrey Ignatov * the whole sysctl value file position should be set to zero. 37064e63acdfSAndrey Ignatov * Return 37074e63acdfSAndrey Ignatov * 0 on success. 37084e63acdfSAndrey Ignatov * 37094e63acdfSAndrey Ignatov * **-E2BIG** if the *buf_len* is too big. 37104e63acdfSAndrey Ignatov * 37114e63acdfSAndrey Ignatov * **-EINVAL** if sysctl is being read. 3712d7a4cb9bSAndrey Ignatov * 3713bdb7b79bSAndrii Nakryiko * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3714d7a4cb9bSAndrey Ignatov * Description 3715d7a4cb9bSAndrey Ignatov * Convert the initial part of the string from buffer *buf* of 3716d7a4cb9bSAndrey Ignatov * size *buf_len* to a long integer according to the given base 3717d7a4cb9bSAndrey Ignatov * and save the result in *res*. 3718d7a4cb9bSAndrey Ignatov * 3719d7a4cb9bSAndrey Ignatov * The string may begin with an arbitrary amount of white space 372080867c5eSQuentin Monnet * (as determined by **isspace**\ (3)) followed by a single 372180867c5eSQuentin Monnet * optional '**-**' sign. 3722d7a4cb9bSAndrey Ignatov * 3723d7a4cb9bSAndrey Ignatov * Five least significant bits of *flags* encode base, other bits 3724d7a4cb9bSAndrey Ignatov * are currently unused. 3725d7a4cb9bSAndrey Ignatov * 3726d7a4cb9bSAndrey Ignatov * Base must be either 8, 10, 16 or 0 to detect it automatically 372780867c5eSQuentin Monnet * similar to user space **strtol**\ (3). 3728d7a4cb9bSAndrey Ignatov * Return 3729d7a4cb9bSAndrey Ignatov * Number of characters consumed on success. Must be positive but 373080867c5eSQuentin Monnet * no more than *buf_len*. 3731d7a4cb9bSAndrey Ignatov * 3732d7a4cb9bSAndrey Ignatov * **-EINVAL** if no valid digits were found or unsupported base 3733d7a4cb9bSAndrey Ignatov * was provided. 3734d7a4cb9bSAndrey Ignatov * 3735d7a4cb9bSAndrey Ignatov * **-ERANGE** if resulting value was out of range. 3736d7a4cb9bSAndrey Ignatov * 3737bdb7b79bSAndrii Nakryiko * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3738d7a4cb9bSAndrey Ignatov * Description 3739d7a4cb9bSAndrey Ignatov * Convert the initial part of the string from buffer *buf* of 3740d7a4cb9bSAndrey Ignatov * size *buf_len* to an unsigned long integer according to the 3741d7a4cb9bSAndrey Ignatov * given base and save the result in *res*. 3742d7a4cb9bSAndrey Ignatov * 3743d7a4cb9bSAndrey Ignatov * The string may begin with an arbitrary amount of white space 374480867c5eSQuentin Monnet * (as determined by **isspace**\ (3)). 3745d7a4cb9bSAndrey Ignatov * 3746d7a4cb9bSAndrey Ignatov * Five least significant bits of *flags* encode base, other bits 3747d7a4cb9bSAndrey Ignatov * are currently unused. 3748d7a4cb9bSAndrey Ignatov * 3749d7a4cb9bSAndrey Ignatov * Base must be either 8, 10, 16 or 0 to detect it automatically 375080867c5eSQuentin Monnet * similar to user space **strtoul**\ (3). 3751d7a4cb9bSAndrey Ignatov * Return 3752d7a4cb9bSAndrey Ignatov * Number of characters consumed on success. Must be positive but 375380867c5eSQuentin Monnet * no more than *buf_len*. 3754d7a4cb9bSAndrey Ignatov * 3755d7a4cb9bSAndrey Ignatov * **-EINVAL** if no valid digits were found or unsupported base 3756d7a4cb9bSAndrey Ignatov * was provided. 3757d7a4cb9bSAndrey Ignatov * 3758d7a4cb9bSAndrey Ignatov * **-ERANGE** if resulting value was out of range. 37596ac99e8fSMartin KaFai Lau * 376030897832SKP Singh * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 37616ac99e8fSMartin KaFai Lau * Description 376280867c5eSQuentin Monnet * Get a bpf-local-storage from a *sk*. 37636ac99e8fSMartin KaFai Lau * 37646ac99e8fSMartin KaFai Lau * Logically, it could be thought of getting the value from 37656ac99e8fSMartin KaFai Lau * a *map* with *sk* as the **key**. From this 37666ac99e8fSMartin KaFai Lau * perspective, the usage is not much different from 376780867c5eSQuentin Monnet * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 376880867c5eSQuentin Monnet * helper enforces the key must be a full socket and the map must 376980867c5eSQuentin Monnet * be a **BPF_MAP_TYPE_SK_STORAGE** also. 37706ac99e8fSMartin KaFai Lau * 37716ac99e8fSMartin KaFai Lau * Underneath, the value is stored locally at *sk* instead of 377280867c5eSQuentin Monnet * the *map*. The *map* is used as the bpf-local-storage 377380867c5eSQuentin Monnet * "type". The bpf-local-storage "type" (i.e. the *map*) is 377480867c5eSQuentin Monnet * searched against all bpf-local-storages residing at *sk*. 37756ac99e8fSMartin KaFai Lau * 377630897832SKP Singh * *sk* is a kernel **struct sock** pointer for LSM program. 377730897832SKP Singh * *sk* is a **struct bpf_sock** pointer for other program types. 377830897832SKP Singh * 377980867c5eSQuentin Monnet * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 37806ac99e8fSMartin KaFai Lau * used such that a new bpf-local-storage will be 37816ac99e8fSMartin KaFai Lau * created if one does not exist. *value* can be used 378280867c5eSQuentin Monnet * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 37836ac99e8fSMartin KaFai Lau * the initial value of a bpf-local-storage. If *value* is 378480867c5eSQuentin Monnet * **NULL**, the new bpf-local-storage will be zero initialized. 37856ac99e8fSMartin KaFai Lau * Return 37866ac99e8fSMartin KaFai Lau * A bpf-local-storage pointer is returned on success. 37876ac99e8fSMartin KaFai Lau * 37886ac99e8fSMartin KaFai Lau * **NULL** if not found or there was an error in adding 37896ac99e8fSMartin KaFai Lau * a new bpf-local-storage. 37906ac99e8fSMartin KaFai Lau * 379130897832SKP Singh * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 37926ac99e8fSMartin KaFai Lau * Description 379380867c5eSQuentin Monnet * Delete a bpf-local-storage from a *sk*. 37946ac99e8fSMartin KaFai Lau * Return 37956ac99e8fSMartin KaFai Lau * 0 on success. 37966ac99e8fSMartin KaFai Lau * 37976ac99e8fSMartin KaFai Lau * **-ENOENT** if the bpf-local-storage cannot be found. 3798592a3498SMartin KaFai Lau * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 37998b401f9eSYonghong Song * 3800bdb7b79bSAndrii Nakryiko * long bpf_send_signal(u32 sig) 38018b401f9eSYonghong Song * Description 38028482941fSYonghong Song * Send signal *sig* to the process of the current task. 38038482941fSYonghong Song * The signal may be delivered to any of this process's threads. 38048b401f9eSYonghong Song * Return 38058b401f9eSYonghong Song * 0 on success or successfully queued. 38068b401f9eSYonghong Song * 38078b401f9eSYonghong Song * **-EBUSY** if work queue under nmi is full. 38088b401f9eSYonghong Song * 38098b401f9eSYonghong Song * **-EINVAL** if *sig* is invalid. 38108b401f9eSYonghong Song * 38118b401f9eSYonghong Song * **-EPERM** if no permission to send the *sig*. 38128b401f9eSYonghong Song * 38138b401f9eSYonghong Song * **-EAGAIN** if bpf program can try again. 381470d66244SPetar Penkov * 3815c0df236eSMartin KaFai Lau * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 381670d66244SPetar Penkov * Description 381770d66244SPetar Penkov * Try to issue a SYN cookie for the packet with corresponding 381870d66244SPetar Penkov * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 381970d66244SPetar Penkov * 382070d66244SPetar Penkov * *iph* points to the start of the IPv4 or IPv6 header, while 382170d66244SPetar Penkov * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3822ac80287aSMaxim Mikityanskiy * **sizeof**\ (**struct ipv6hdr**). 382370d66244SPetar Penkov * 382470d66244SPetar Penkov * *th* points to the start of the TCP header, while *th_len* 3825ac80287aSMaxim Mikityanskiy * contains the length of the TCP header with options (at least 3826ac80287aSMaxim Mikityanskiy * **sizeof**\ (**struct tcphdr**)). 382770d66244SPetar Penkov * Return 382870d66244SPetar Penkov * On success, lower 32 bits hold the generated SYN cookie in 382970d66244SPetar Penkov * followed by 16 bits which hold the MSS value for that cookie, 383070d66244SPetar Penkov * and the top 16 bits are unused. 383170d66244SPetar Penkov * 383270d66244SPetar Penkov * On failure, the returned value is one of the following: 383370d66244SPetar Penkov * 383470d66244SPetar Penkov * **-EINVAL** SYN cookie cannot be issued due to error 383570d66244SPetar Penkov * 383670d66244SPetar Penkov * **-ENOENT** SYN cookie should not be issued (no SYN flood) 383770d66244SPetar Penkov * 383870d66244SPetar Penkov * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 383970d66244SPetar Penkov * 384070d66244SPetar Penkov * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 3841a7658e1aSAlexei Starovoitov * 3842bdb7b79bSAndrii Nakryiko * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3843a7658e1aSAlexei Starovoitov * Description 3844a7658e1aSAlexei Starovoitov * Write raw *data* blob into a special BPF perf event held by 3845a7658e1aSAlexei Starovoitov * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3846a7658e1aSAlexei Starovoitov * event must have the following attributes: **PERF_SAMPLE_RAW** 3847a7658e1aSAlexei Starovoitov * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3848a7658e1aSAlexei Starovoitov * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3849a7658e1aSAlexei Starovoitov * 3850a7658e1aSAlexei Starovoitov * The *flags* are used to indicate the index in *map* for which 3851a7658e1aSAlexei Starovoitov * the value must be put, masked with **BPF_F_INDEX_MASK**. 3852a7658e1aSAlexei Starovoitov * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3853a7658e1aSAlexei Starovoitov * to indicate that the index of the current CPU core should be 3854a7658e1aSAlexei Starovoitov * used. 3855a7658e1aSAlexei Starovoitov * 3856a7658e1aSAlexei Starovoitov * The value to write, of *size*, is passed through eBPF stack and 3857a7658e1aSAlexei Starovoitov * pointed by *data*. 3858a7658e1aSAlexei Starovoitov * 3859a7658e1aSAlexei Starovoitov * *ctx* is a pointer to in-kernel struct sk_buff. 3860a7658e1aSAlexei Starovoitov * 3861a7658e1aSAlexei Starovoitov * This helper is similar to **bpf_perf_event_output**\ () but 3862a7658e1aSAlexei Starovoitov * restricted to raw_tracepoint bpf programs. 3863a7658e1aSAlexei Starovoitov * Return 3864a7658e1aSAlexei Starovoitov * 0 on success, or a negative error in case of failure. 38656ae08ae3SDaniel Borkmann * 3866bdb7b79bSAndrii Nakryiko * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 38676ae08ae3SDaniel Borkmann * Description 38686ae08ae3SDaniel Borkmann * Safely attempt to read *size* bytes from user space address 38696ae08ae3SDaniel Borkmann * *unsafe_ptr* and store the data in *dst*. 38706ae08ae3SDaniel Borkmann * Return 38716ae08ae3SDaniel Borkmann * 0 on success, or a negative error in case of failure. 38726ae08ae3SDaniel Borkmann * 3873bdb7b79bSAndrii Nakryiko * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 38746ae08ae3SDaniel Borkmann * Description 38756ae08ae3SDaniel Borkmann * Safely attempt to read *size* bytes from kernel space address 38766ae08ae3SDaniel Borkmann * *unsafe_ptr* and store the data in *dst*. 38776ae08ae3SDaniel Borkmann * Return 38786ae08ae3SDaniel Borkmann * 0 on success, or a negative error in case of failure. 38796ae08ae3SDaniel Borkmann * 3880bdb7b79bSAndrii Nakryiko * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 38816ae08ae3SDaniel Borkmann * Description 38826ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe user address 38836ae08ae3SDaniel Borkmann * *unsafe_ptr* to *dst*. The *size* should include the 38846ae08ae3SDaniel Borkmann * terminating NUL byte. In case the string length is smaller than 38856ae08ae3SDaniel Borkmann * *size*, the target is not padded with further NUL bytes. If the 38866ae08ae3SDaniel Borkmann * string length is larger than *size*, just *size*-1 bytes are 38876ae08ae3SDaniel Borkmann * copied and the last byte is set to NUL. 38886ae08ae3SDaniel Borkmann * 3889c6458e72SBrendan Jackman * On success, returns the number of bytes that were written, 3890c6458e72SBrendan Jackman * including the terminal NUL. This makes this helper useful in 3891c6458e72SBrendan Jackman * tracing programs for reading strings, and more importantly to 3892c6458e72SBrendan Jackman * get its length at runtime. See the following snippet: 38936ae08ae3SDaniel Borkmann * 38946ae08ae3SDaniel Borkmann * :: 38956ae08ae3SDaniel Borkmann * 38966ae08ae3SDaniel Borkmann * SEC("kprobe/sys_open") 38976ae08ae3SDaniel Borkmann * void bpf_sys_open(struct pt_regs *ctx) 38986ae08ae3SDaniel Borkmann * { 38996ae08ae3SDaniel Borkmann * char buf[PATHLEN]; // PATHLEN is defined to 256 39006ae08ae3SDaniel Borkmann * int res = bpf_probe_read_user_str(buf, sizeof(buf), 39016ae08ae3SDaniel Borkmann * ctx->di); 39026ae08ae3SDaniel Borkmann * 39036ae08ae3SDaniel Borkmann * // Consume buf, for example push it to 39046ae08ae3SDaniel Borkmann * // userspace via bpf_perf_event_output(); we 39056ae08ae3SDaniel Borkmann * // can use res (the string length) as event 39066ae08ae3SDaniel Borkmann * // size, after checking its boundaries. 39076ae08ae3SDaniel Borkmann * } 39086ae08ae3SDaniel Borkmann * 3909ab8d7809SQuentin Monnet * In comparison, using **bpf_probe_read_user**\ () helper here 39106ae08ae3SDaniel Borkmann * instead to read the string would require to estimate the length 39116ae08ae3SDaniel Borkmann * at compile time, and would often result in copying more memory 39126ae08ae3SDaniel Borkmann * than necessary. 39136ae08ae3SDaniel Borkmann * 39146ae08ae3SDaniel Borkmann * Another useful use case is when parsing individual process 39156ae08ae3SDaniel Borkmann * arguments or individual environment variables navigating 39166ae08ae3SDaniel Borkmann * *current*\ **->mm->arg_start** and *current*\ 39176ae08ae3SDaniel Borkmann * **->mm->env_start**: using this helper and the return value, 39186ae08ae3SDaniel Borkmann * one can quickly iterate at the right offset of the memory area. 39196ae08ae3SDaniel Borkmann * Return 3920c6458e72SBrendan Jackman * On success, the strictly positive length of the output string, 39216ae08ae3SDaniel Borkmann * including the trailing NUL character. On error, a negative 39226ae08ae3SDaniel Borkmann * value. 39236ae08ae3SDaniel Borkmann * 3924bdb7b79bSAndrii Nakryiko * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 39256ae08ae3SDaniel Borkmann * Description 39266ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 3927ab8d7809SQuentin Monnet * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 39286ae08ae3SDaniel Borkmann * Return 39296ae08ae3SDaniel Borkmann * On success, the strictly positive length of the string, including 39306ae08ae3SDaniel Borkmann * the trailing NUL character. On error, a negative value. 3931206057feSMartin KaFai Lau * 3932bdb7b79bSAndrii Nakryiko * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 3933206057feSMartin KaFai Lau * Description 3934ab8d7809SQuentin Monnet * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 3935206057feSMartin KaFai Lau * *rcv_nxt* is the ack_seq to be sent out. 3936206057feSMartin KaFai Lau * Return 3937206057feSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 3938206057feSMartin KaFai Lau * 3939bdb7b79bSAndrii Nakryiko * long bpf_send_signal_thread(u32 sig) 39408482941fSYonghong Song * Description 39418482941fSYonghong Song * Send signal *sig* to the thread corresponding to the current task. 39428482941fSYonghong Song * Return 39438482941fSYonghong Song * 0 on success or successfully queued. 39448482941fSYonghong Song * 39458482941fSYonghong Song * **-EBUSY** if work queue under nmi is full. 39468482941fSYonghong Song * 39478482941fSYonghong Song * **-EINVAL** if *sig* is invalid. 39488482941fSYonghong Song * 39498482941fSYonghong Song * **-EPERM** if no permission to send the *sig*. 39508482941fSYonghong Song * 39518482941fSYonghong Song * **-EAGAIN** if bpf program can try again. 39525576b991SMartin KaFai Lau * 39535576b991SMartin KaFai Lau * u64 bpf_jiffies64(void) 39545576b991SMartin KaFai Lau * Description 39555576b991SMartin KaFai Lau * Obtain the 64bit jiffies 39565576b991SMartin KaFai Lau * Return 39575576b991SMartin KaFai Lau * The 64 bit jiffies 3958fff7b643SDaniel Xu * 3959bdb7b79bSAndrii Nakryiko * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3960fff7b643SDaniel Xu * Description 3961fff7b643SDaniel Xu * For an eBPF program attached to a perf event, retrieve the 3962ab8d7809SQuentin Monnet * branch records (**struct perf_branch_entry**) associated to *ctx* 3963fff7b643SDaniel Xu * and store it in the buffer pointed by *buf* up to size 3964fff7b643SDaniel Xu * *size* bytes. 3965fff7b643SDaniel Xu * Return 3966fff7b643SDaniel Xu * On success, number of bytes written to *buf*. On error, a 3967fff7b643SDaniel Xu * negative value. 3968fff7b643SDaniel Xu * 3969fff7b643SDaniel Xu * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3970fff7b643SDaniel Xu * instead return the number of bytes required to store all the 3971fff7b643SDaniel Xu * branch entries. If this flag is set, *buf* may be NULL. 3972fff7b643SDaniel Xu * 3973fff7b643SDaniel Xu * **-EINVAL** if arguments invalid or **size** not a multiple 3974ab8d7809SQuentin Monnet * of **sizeof**\ (**struct perf_branch_entry**\ ). 3975fff7b643SDaniel Xu * 3976fff7b643SDaniel Xu * **-ENOENT** if architecture does not support branch records. 3977b4490c5cSCarlos Neira * 3978bdb7b79bSAndrii Nakryiko * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3979b4490c5cSCarlos Neira * Description 3980b4490c5cSCarlos Neira * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3981b4490c5cSCarlos Neira * *namespace* will be returned in *nsdata*. 3982ab8d7809SQuentin Monnet * Return 3983ab8d7809SQuentin Monnet * 0 on success, or one of the following in case of failure: 3984b4490c5cSCarlos Neira * 3985b4490c5cSCarlos Neira * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3986b4490c5cSCarlos Neira * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3987b4490c5cSCarlos Neira * 3988b4490c5cSCarlos Neira * **-ENOENT** if pidns does not exists for the current task. 3989b4490c5cSCarlos Neira * 3990bdb7b79bSAndrii Nakryiko * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3991d831ee84SEelco Chaudron * Description 3992d831ee84SEelco Chaudron * Write raw *data* blob into a special BPF perf event held by 3993d831ee84SEelco Chaudron * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3994d831ee84SEelco Chaudron * event must have the following attributes: **PERF_SAMPLE_RAW** 3995d831ee84SEelco Chaudron * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3996d831ee84SEelco Chaudron * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3997d831ee84SEelco Chaudron * 3998d831ee84SEelco Chaudron * The *flags* are used to indicate the index in *map* for which 3999d831ee84SEelco Chaudron * the value must be put, masked with **BPF_F_INDEX_MASK**. 4000d831ee84SEelco Chaudron * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4001d831ee84SEelco Chaudron * to indicate that the index of the current CPU core should be 4002d831ee84SEelco Chaudron * used. 4003d831ee84SEelco Chaudron * 4004d831ee84SEelco Chaudron * The value to write, of *size*, is passed through eBPF stack and 4005d831ee84SEelco Chaudron * pointed by *data*. 4006d831ee84SEelco Chaudron * 4007d831ee84SEelco Chaudron * *ctx* is a pointer to in-kernel struct xdp_buff. 4008d831ee84SEelco Chaudron * 4009d831ee84SEelco Chaudron * This helper is similar to **bpf_perf_eventoutput**\ () but 4010d831ee84SEelco Chaudron * restricted to raw_tracepoint bpf programs. 4011d831ee84SEelco Chaudron * Return 4012d831ee84SEelco Chaudron * 0 on success, or a negative error in case of failure. 4013f318903cSDaniel Borkmann * 4014f318903cSDaniel Borkmann * u64 bpf_get_netns_cookie(void *ctx) 4015f318903cSDaniel Borkmann * Description 4016f318903cSDaniel Borkmann * Retrieve the cookie (generated by the kernel) of the network 4017f318903cSDaniel Borkmann * namespace the input *ctx* is associated with. The network 4018f318903cSDaniel Borkmann * namespace cookie remains stable for its lifetime and provides 4019f318903cSDaniel Borkmann * a global identifier that can be assumed unique. If *ctx* is 4020f318903cSDaniel Borkmann * NULL, then the helper returns the cookie for the initial 4021f318903cSDaniel Borkmann * network namespace. The cookie itself is very similar to that 4022ab8d7809SQuentin Monnet * of **bpf_get_socket_cookie**\ () helper, but for network 4023ab8d7809SQuentin Monnet * namespaces instead of sockets. 4024f318903cSDaniel Borkmann * Return 4025f318903cSDaniel Borkmann * A 8-byte long opaque number. 40260f09abd1SDaniel Borkmann * 40270f09abd1SDaniel Borkmann * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 40280f09abd1SDaniel Borkmann * Description 40290f09abd1SDaniel Borkmann * Return id of cgroup v2 that is ancestor of the cgroup associated 40300f09abd1SDaniel Borkmann * with the current task at the *ancestor_level*. The root cgroup 40310f09abd1SDaniel Borkmann * is at *ancestor_level* zero and each step down the hierarchy 40320f09abd1SDaniel Borkmann * increments the level. If *ancestor_level* == level of cgroup 40330f09abd1SDaniel Borkmann * associated with the current task, then return value will be the 40340f09abd1SDaniel Borkmann * same as that of **bpf_get_current_cgroup_id**\ (). 40350f09abd1SDaniel Borkmann * 40360f09abd1SDaniel Borkmann * The helper is useful to implement policies based on cgroups 40370f09abd1SDaniel Borkmann * that are upper in hierarchy than immediate cgroup associated 40380f09abd1SDaniel Borkmann * with the current task. 40390f09abd1SDaniel Borkmann * 40400f09abd1SDaniel Borkmann * The format of returned id and helper limitations are same as in 40410f09abd1SDaniel Borkmann * **bpf_get_current_cgroup_id**\ (). 40420f09abd1SDaniel Borkmann * Return 40430f09abd1SDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 4044cf7fbe66SJoe Stringer * 404527e5203bSMartin KaFai Lau * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 4046cf7fbe66SJoe Stringer * Description 4047e9ddbb77SJakub Sitnicki * Helper is overloaded depending on BPF program type. This 4048e9ddbb77SJakub Sitnicki * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 4049e9ddbb77SJakub Sitnicki * **BPF_PROG_TYPE_SCHED_ACT** programs. 4050e9ddbb77SJakub Sitnicki * 4051cf7fbe66SJoe Stringer * Assign the *sk* to the *skb*. When combined with appropriate 4052cf7fbe66SJoe Stringer * routing configuration to receive the packet towards the socket, 4053cf7fbe66SJoe Stringer * will cause *skb* to be delivered to the specified socket. 4054cf7fbe66SJoe Stringer * Subsequent redirection of *skb* via **bpf_redirect**\ (), 4055cf7fbe66SJoe Stringer * **bpf_clone_redirect**\ () or other methods outside of BPF may 4056cf7fbe66SJoe Stringer * interfere with successful delivery to the socket. 4057cf7fbe66SJoe Stringer * 4058cf7fbe66SJoe Stringer * This operation is only valid from TC ingress path. 4059cf7fbe66SJoe Stringer * 4060cf7fbe66SJoe Stringer * The *flags* argument must be zero. 4061cf7fbe66SJoe Stringer * Return 4062ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 4063cf7fbe66SJoe Stringer * 4064ab8d7809SQuentin Monnet * **-EINVAL** if specified *flags* are not supported. 4065ab8d7809SQuentin Monnet * 4066ab8d7809SQuentin Monnet * **-ENOENT** if the socket is unavailable for assignment. 4067ab8d7809SQuentin Monnet * 4068ab8d7809SQuentin Monnet * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4069ab8d7809SQuentin Monnet * 4070ab8d7809SQuentin Monnet * **-EOPNOTSUPP** if the operation is not supported, for example 4071ab8d7809SQuentin Monnet * a call from outside of TC ingress. 4072ab8d7809SQuentin Monnet * 4073ab8d7809SQuentin Monnet * **-ESOCKTNOSUPPORT** if the socket type is not supported 4074ab8d7809SQuentin Monnet * (reuseport). 407571d19214SMaciej Żenczykowski * 4076e9ddbb77SJakub Sitnicki * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4077e9ddbb77SJakub Sitnicki * Description 4078e9ddbb77SJakub Sitnicki * Helper is overloaded depending on BPF program type. This 4079e9ddbb77SJakub Sitnicki * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4080e9ddbb77SJakub Sitnicki * 4081e9ddbb77SJakub Sitnicki * Select the *sk* as a result of a socket lookup. 4082e9ddbb77SJakub Sitnicki * 4083e9ddbb77SJakub Sitnicki * For the operation to succeed passed socket must be compatible 4084e9ddbb77SJakub Sitnicki * with the packet description provided by the *ctx* object. 4085e9ddbb77SJakub Sitnicki * 4086e9ddbb77SJakub Sitnicki * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4087e9ddbb77SJakub Sitnicki * be an exact match. While IP family (**AF_INET** or 4088e9ddbb77SJakub Sitnicki * **AF_INET6**) must be compatible, that is IPv6 sockets 4089e9ddbb77SJakub Sitnicki * that are not v6-only can be selected for IPv4 packets. 4090e9ddbb77SJakub Sitnicki * 4091e9ddbb77SJakub Sitnicki * Only TCP listeners and UDP unconnected sockets can be 4092e9ddbb77SJakub Sitnicki * selected. *sk* can also be NULL to reset any previous 4093e9ddbb77SJakub Sitnicki * selection. 4094e9ddbb77SJakub Sitnicki * 4095e9ddbb77SJakub Sitnicki * *flags* argument can combination of following values: 4096e9ddbb77SJakub Sitnicki * 4097e9ddbb77SJakub Sitnicki * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4098e9ddbb77SJakub Sitnicki * socket selection, potentially done by a BPF program 4099e9ddbb77SJakub Sitnicki * that ran before us. 4100e9ddbb77SJakub Sitnicki * 4101e9ddbb77SJakub Sitnicki * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4102e9ddbb77SJakub Sitnicki * load-balancing within reuseport group for the socket 4103e9ddbb77SJakub Sitnicki * being selected. 4104e9ddbb77SJakub Sitnicki * 4105e9ddbb77SJakub Sitnicki * On success *ctx->sk* will point to the selected socket. 4106e9ddbb77SJakub Sitnicki * 4107e9ddbb77SJakub Sitnicki * Return 4108e9ddbb77SJakub Sitnicki * 0 on success, or a negative errno in case of failure. 4109e9ddbb77SJakub Sitnicki * 4110e9ddbb77SJakub Sitnicki * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4111e9ddbb77SJakub Sitnicki * not compatible with packet family (*ctx->family*). 4112e9ddbb77SJakub Sitnicki * 4113e9ddbb77SJakub Sitnicki * * **-EEXIST** if socket has been already selected, 4114e9ddbb77SJakub Sitnicki * potentially by another program, and 4115e9ddbb77SJakub Sitnicki * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4116e9ddbb77SJakub Sitnicki * 4117e9ddbb77SJakub Sitnicki * * **-EINVAL** if unsupported flags were specified. 4118e9ddbb77SJakub Sitnicki * 4119e9ddbb77SJakub Sitnicki * * **-EPROTOTYPE** if socket L4 protocol 4120e9ddbb77SJakub Sitnicki * (*sk->protocol*) doesn't match packet protocol 4121e9ddbb77SJakub Sitnicki * (*ctx->protocol*). 4122e9ddbb77SJakub Sitnicki * 4123e9ddbb77SJakub Sitnicki * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4124e9ddbb77SJakub Sitnicki * state (TCP listening or UDP unconnected). 4125e9ddbb77SJakub Sitnicki * 412671d19214SMaciej Żenczykowski * u64 bpf_ktime_get_boot_ns(void) 412771d19214SMaciej Żenczykowski * Description 412871d19214SMaciej Żenczykowski * Return the time elapsed since system boot, in nanoseconds. 412971d19214SMaciej Żenczykowski * Does include the time the system was suspended. 4130ab8d7809SQuentin Monnet * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 413171d19214SMaciej Żenczykowski * Return 413271d19214SMaciej Żenczykowski * Current *ktime*. 4133492e639fSYonghong Song * 4134bdb7b79bSAndrii Nakryiko * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4135492e639fSYonghong Song * Description 4136ab8d7809SQuentin Monnet * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4137ab8d7809SQuentin Monnet * out the format string. 4138492e639fSYonghong Song * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4139492e639fSYonghong Song * the format string itself. The *data* and *data_len* are format string 4140ab8d7809SQuentin Monnet * arguments. The *data* are a **u64** array and corresponding format string 4141492e639fSYonghong Song * values are stored in the array. For strings and pointers where pointees 4142492e639fSYonghong Song * are accessed, only the pointer values are stored in the *data* array. 4143a42effb0SDave Marchevsky * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4144492e639fSYonghong Song * 4145492e639fSYonghong Song * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4146492e639fSYonghong Song * Reading kernel memory may fail due to either invalid address or 4147492e639fSYonghong Song * valid address but requiring a major memory fault. If reading kernel memory 4148492e639fSYonghong Song * fails, the string for **%s** will be an empty string, and the ip 4149492e639fSYonghong Song * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4150ab8d7809SQuentin Monnet * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4151492e639fSYonghong Song * Return 4152ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 4153492e639fSYonghong Song * 4154ab8d7809SQuentin Monnet * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4155492e639fSYonghong Song * by returning 1 from bpf program. 4156ab8d7809SQuentin Monnet * 4157ab8d7809SQuentin Monnet * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4158ab8d7809SQuentin Monnet * 4159ab8d7809SQuentin Monnet * **-E2BIG** if *fmt* contains too many format specifiers. 4160ab8d7809SQuentin Monnet * 4161ab8d7809SQuentin Monnet * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4162492e639fSYonghong Song * 4163bdb7b79bSAndrii Nakryiko * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4164492e639fSYonghong Song * Description 4165ab8d7809SQuentin Monnet * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4166492e639fSYonghong Song * The *m* represents the seq_file. The *data* and *len* represent the 4167492e639fSYonghong Song * data to write in bytes. 4168492e639fSYonghong Song * Return 4169ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 4170492e639fSYonghong Song * 4171ab8d7809SQuentin Monnet * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4172f307fa2cSAndrey Ignatov * 4173a5fa25adSMartin KaFai Lau * u64 bpf_sk_cgroup_id(void *sk) 4174f307fa2cSAndrey Ignatov * Description 4175f307fa2cSAndrey Ignatov * Return the cgroup v2 id of the socket *sk*. 4176f307fa2cSAndrey Ignatov * 4177a5fa25adSMartin KaFai Lau * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4178f307fa2cSAndrey Ignatov * returned from **bpf_sk_lookup_xxx**\ (), 4179f307fa2cSAndrey Ignatov * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4180f307fa2cSAndrey Ignatov * same as in **bpf_skb_cgroup_id**\ (). 4181f307fa2cSAndrey Ignatov * 4182f307fa2cSAndrey Ignatov * This helper is available only if the kernel was compiled with 4183f307fa2cSAndrey Ignatov * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4184f307fa2cSAndrey Ignatov * Return 4185f307fa2cSAndrey Ignatov * The id is returned or 0 in case the id could not be retrieved. 4186f307fa2cSAndrey Ignatov * 4187a5fa25adSMartin KaFai Lau * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4188f307fa2cSAndrey Ignatov * Description 4189f307fa2cSAndrey Ignatov * Return id of cgroup v2 that is ancestor of cgroup associated 4190f307fa2cSAndrey Ignatov * with the *sk* at the *ancestor_level*. The root cgroup is at 4191f307fa2cSAndrey Ignatov * *ancestor_level* zero and each step down the hierarchy 4192f307fa2cSAndrey Ignatov * increments the level. If *ancestor_level* == level of cgroup 4193f307fa2cSAndrey Ignatov * associated with *sk*, then return value will be same as that 4194f307fa2cSAndrey Ignatov * of **bpf_sk_cgroup_id**\ (). 4195f307fa2cSAndrey Ignatov * 4196f307fa2cSAndrey Ignatov * The helper is useful to implement policies based on cgroups 4197f307fa2cSAndrey Ignatov * that are upper in hierarchy than immediate cgroup associated 4198f307fa2cSAndrey Ignatov * with *sk*. 4199f307fa2cSAndrey Ignatov * 4200f307fa2cSAndrey Ignatov * The format of returned id and helper limitations are same as in 4201f307fa2cSAndrey Ignatov * **bpf_sk_cgroup_id**\ (). 4202f307fa2cSAndrey Ignatov * Return 4203f307fa2cSAndrey Ignatov * The id is returned or 0 in case the id could not be retrieved. 4204457f4436SAndrii Nakryiko * 4205e1613b57SAndrii Nakryiko * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4206457f4436SAndrii Nakryiko * Description 4207457f4436SAndrii Nakryiko * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4208bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4209bcc7f554SQuentin Monnet * of new data availability is sent. 4210bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4211bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 42125c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 42135c507329SPedro Tammela * of new data availability is sent. 42145c507329SPedro Tammela * 42155c507329SPedro Tammela * An adaptive notification is a notification sent whenever the user-space 42165c507329SPedro Tammela * process has caught up and consumed all available payloads. In case the user-space 42175c507329SPedro Tammela * process is still processing a previous payload, then no notification is needed 42185c507329SPedro Tammela * as it will process the newly added payload automatically. 4219457f4436SAndrii Nakryiko * Return 4220bcc7f554SQuentin Monnet * 0 on success, or a negative error in case of failure. 4221457f4436SAndrii Nakryiko * 4222457f4436SAndrii Nakryiko * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4223457f4436SAndrii Nakryiko * Description 4224457f4436SAndrii Nakryiko * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 42255c507329SPedro Tammela * *flags* must be 0. 4226457f4436SAndrii Nakryiko * Return 4227457f4436SAndrii Nakryiko * Valid pointer with *size* bytes of memory available; NULL, 4228457f4436SAndrii Nakryiko * otherwise. 4229457f4436SAndrii Nakryiko * 4230457f4436SAndrii Nakryiko * void bpf_ringbuf_submit(void *data, u64 flags) 4231457f4436SAndrii Nakryiko * Description 4232457f4436SAndrii Nakryiko * Submit reserved ring buffer sample, pointed to by *data*. 4233bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4234bcc7f554SQuentin Monnet * of new data availability is sent. 4235bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4236bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 42375c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 42385c507329SPedro Tammela * of new data availability is sent. 42395c507329SPedro Tammela * 42405c507329SPedro Tammela * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4241457f4436SAndrii Nakryiko * Return 4242457f4436SAndrii Nakryiko * Nothing. Always succeeds. 4243457f4436SAndrii Nakryiko * 4244457f4436SAndrii Nakryiko * void bpf_ringbuf_discard(void *data, u64 flags) 4245457f4436SAndrii Nakryiko * Description 4246457f4436SAndrii Nakryiko * Discard reserved ring buffer sample, pointed to by *data*. 4247bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4248bcc7f554SQuentin Monnet * of new data availability is sent. 4249bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4250bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 42515c507329SPedro Tammela * If **0** is specified in *flags*, an adaptive notification 42525c507329SPedro Tammela * of new data availability is sent. 42535c507329SPedro Tammela * 42545c507329SPedro Tammela * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4255457f4436SAndrii Nakryiko * Return 4256457f4436SAndrii Nakryiko * Nothing. Always succeeds. 4257457f4436SAndrii Nakryiko * 4258457f4436SAndrii Nakryiko * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4259457f4436SAndrii Nakryiko * Description 4260457f4436SAndrii Nakryiko * Query various characteristics of provided ring buffer. What 4261457f4436SAndrii Nakryiko * exactly is queries is determined by *flags*: 4262bcc7f554SQuentin Monnet * 4263bcc7f554SQuentin Monnet * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4264bcc7f554SQuentin Monnet * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4265bcc7f554SQuentin Monnet * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4266bcc7f554SQuentin Monnet * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4267bcc7f554SQuentin Monnet * 4268bcc7f554SQuentin Monnet * Data returned is just a momentary snapshot of actual values 4269457f4436SAndrii Nakryiko * and could be inaccurate, so this facility should be used to 4270457f4436SAndrii Nakryiko * power heuristics and for reporting, not to make 100% correct 4271457f4436SAndrii Nakryiko * calculation. 4272457f4436SAndrii Nakryiko * Return 4273bcc7f554SQuentin Monnet * Requested value, or 0, if *flags* are not recognized. 42747cdec54fSDaniel Borkmann * 4275bdb7b79bSAndrii Nakryiko * long bpf_csum_level(struct sk_buff *skb, u64 level) 42767cdec54fSDaniel Borkmann * Description 42777cdec54fSDaniel Borkmann * Change the skbs checksum level by one layer up or down, or 42787cdec54fSDaniel Borkmann * reset it entirely to none in order to have the stack perform 42797cdec54fSDaniel Borkmann * checksum validation. The level is applicable to the following 42807cdec54fSDaniel Borkmann * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 42817cdec54fSDaniel Borkmann * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 42827cdec54fSDaniel Borkmann * through **bpf_skb_adjust_room**\ () helper with passing in 42837cdec54fSDaniel Borkmann * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 42847cdec54fSDaniel Borkmann * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 42857cdec54fSDaniel Borkmann * the UDP header is removed. Similarly, an encap of the latter 42867cdec54fSDaniel Borkmann * into the former could be accompanied by a helper call to 42877cdec54fSDaniel Borkmann * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 42887cdec54fSDaniel Borkmann * skb is still intended to be processed in higher layers of the 42897cdec54fSDaniel Borkmann * stack instead of just egressing at tc. 42907cdec54fSDaniel Borkmann * 42917cdec54fSDaniel Borkmann * There are three supported level settings at this time: 42927cdec54fSDaniel Borkmann * 42937cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 42947cdec54fSDaniel Borkmann * with CHECKSUM_UNNECESSARY. 42957cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 42967cdec54fSDaniel Borkmann * with CHECKSUM_UNNECESSARY. 42977cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 42987cdec54fSDaniel Borkmann * sets CHECKSUM_NONE to force checksum validation by the stack. 42997cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 43007cdec54fSDaniel Borkmann * skb->csum_level. 43017cdec54fSDaniel Borkmann * Return 43027cdec54fSDaniel Borkmann * 0 on success, or a negative error in case of failure. In the 43037cdec54fSDaniel Borkmann * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 43047cdec54fSDaniel Borkmann * is returned or the error code -EACCES in case the skb is not 43057cdec54fSDaniel Borkmann * subject to CHECKSUM_UNNECESSARY. 4306af7ec138SYonghong Song * 4307af7ec138SYonghong Song * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4308af7ec138SYonghong Song * Description 4309af7ec138SYonghong Song * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4310af7ec138SYonghong Song * Return 4311938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4312478cfbdfSYonghong Song * 4313478cfbdfSYonghong Song * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4314478cfbdfSYonghong Song * Description 4315478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4316478cfbdfSYonghong Song * Return 4317938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4318478cfbdfSYonghong Song * 4319478cfbdfSYonghong Song * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4320478cfbdfSYonghong Song * Description 4321478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4322478cfbdfSYonghong Song * Return 4323938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4324478cfbdfSYonghong Song * 4325478cfbdfSYonghong Song * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4326478cfbdfSYonghong Song * Description 4327478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4328478cfbdfSYonghong Song * Return 4329938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 43300d4fad3eSYonghong Song * 43310d4fad3eSYonghong Song * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 43320d4fad3eSYonghong Song * Description 43330d4fad3eSYonghong Song * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 43340d4fad3eSYonghong Song * Return 4335938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 4336fa28dcb8SSong Liu * 4337fa28dcb8SSong Liu * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4338fa28dcb8SSong Liu * Description 4339fa28dcb8SSong Liu * Return a user or a kernel stack in bpf program provided buffer. 4340fa28dcb8SSong Liu * To achieve this, the helper needs *task*, which is a valid 4341938c3efdSQuentin Monnet * pointer to **struct task_struct**. To store the stacktrace, the 4342fa28dcb8SSong Liu * bpf program provides *buf* with a nonnegative *size*. 4343fa28dcb8SSong Liu * 4344fa28dcb8SSong Liu * The last argument, *flags*, holds the number of stack frames to 4345fa28dcb8SSong Liu * skip (from 0 to 255), masked with 4346fa28dcb8SSong Liu * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4347fa28dcb8SSong Liu * the following flags: 4348fa28dcb8SSong Liu * 4349fa28dcb8SSong Liu * **BPF_F_USER_STACK** 4350fa28dcb8SSong Liu * Collect a user space stack instead of a kernel stack. 4351fa28dcb8SSong Liu * **BPF_F_USER_BUILD_ID** 4352fa28dcb8SSong Liu * Collect buildid+offset instead of ips for user stack, 4353fa28dcb8SSong Liu * only valid if **BPF_F_USER_STACK** is also specified. 4354fa28dcb8SSong Liu * 4355fa28dcb8SSong Liu * **bpf_get_task_stack**\ () can collect up to 4356fa28dcb8SSong Liu * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4357fa28dcb8SSong Liu * to sufficient large buffer size. Note that 4358fa28dcb8SSong Liu * this limit can be controlled with the **sysctl** program, and 4359fa28dcb8SSong Liu * that it should be manually increased in order to profile long 4360fa28dcb8SSong Liu * user stacks (such as stacks for Java programs). To do so, use: 4361fa28dcb8SSong Liu * 4362fa28dcb8SSong Liu * :: 4363fa28dcb8SSong Liu * 4364fa28dcb8SSong Liu * # sysctl kernel.perf_event_max_stack=<new value> 4365fa28dcb8SSong Liu * Return 4366ee2a0988SNamhyung Kim * The non-negative copied *buf* length equal to or less than 4367ee2a0988SNamhyung Kim * *size* on success, or a negative error in case of failure. 4368fa28dcb8SSong Liu * 43690813a841SMartin KaFai Lau * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 43700813a841SMartin KaFai Lau * Description 43710813a841SMartin KaFai Lau * Load header option. Support reading a particular TCP header 4372938c3efdSQuentin Monnet * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 43730813a841SMartin KaFai Lau * 43740813a841SMartin KaFai Lau * If *flags* is 0, it will search the option from the 4375938c3efdSQuentin Monnet * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 43760813a841SMartin KaFai Lau * has details on what skb_data contains under different 4377938c3efdSQuentin Monnet * *skops*\ **->op**. 43780813a841SMartin KaFai Lau * 43790813a841SMartin KaFai Lau * The first byte of the *searchby_res* specifies the 43800813a841SMartin KaFai Lau * kind that it wants to search. 43810813a841SMartin KaFai Lau * 43820813a841SMartin KaFai Lau * If the searching kind is an experimental kind 43830813a841SMartin KaFai Lau * (i.e. 253 or 254 according to RFC6994). It also 43840813a841SMartin KaFai Lau * needs to specify the "magic" which is either 43850813a841SMartin KaFai Lau * 2 bytes or 4 bytes. It then also needs to 43860813a841SMartin KaFai Lau * specify the size of the magic by using 43870813a841SMartin KaFai Lau * the 2nd byte which is "kind-length" of a TCP 43880813a841SMartin KaFai Lau * header option and the "kind-length" also 43890813a841SMartin KaFai Lau * includes the first 2 bytes "kind" and "kind-length" 43900813a841SMartin KaFai Lau * itself as a normal TCP header option also does. 43910813a841SMartin KaFai Lau * 43920813a841SMartin KaFai Lau * For example, to search experimental kind 254 with 43930813a841SMartin KaFai Lau * 2 byte magic 0xeB9F, the searchby_res should be 43940813a841SMartin KaFai Lau * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 43950813a841SMartin KaFai Lau * 43960813a841SMartin KaFai Lau * To search for the standard window scale option (3), 4397938c3efdSQuentin Monnet * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 43980813a841SMartin KaFai Lau * Note, kind-length must be 0 for regular option. 43990813a841SMartin KaFai Lau * 44000813a841SMartin KaFai Lau * Searching for No-Op (0) and End-of-Option-List (1) are 44010813a841SMartin KaFai Lau * not supported. 44020813a841SMartin KaFai Lau * 44030813a841SMartin KaFai Lau * *len* must be at least 2 bytes which is the minimal size 44040813a841SMartin KaFai Lau * of a header option. 44050813a841SMartin KaFai Lau * 44060813a841SMartin KaFai Lau * Supported flags: 4407938c3efdSQuentin Monnet * 44080813a841SMartin KaFai Lau * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 44090813a841SMartin KaFai Lau * saved_syn packet or the just-received syn packet. 44100813a841SMartin KaFai Lau * 44110813a841SMartin KaFai Lau * Return 44120813a841SMartin KaFai Lau * > 0 when found, the header option is copied to *searchby_res*. 4413938c3efdSQuentin Monnet * The return value is the total length copied. On failure, a 4414938c3efdSQuentin Monnet * negative error code is returned: 44150813a841SMartin KaFai Lau * 4416938c3efdSQuentin Monnet * **-EINVAL** if a parameter is invalid. 44170813a841SMartin KaFai Lau * 4418938c3efdSQuentin Monnet * **-ENOMSG** if the option is not found. 44190813a841SMartin KaFai Lau * 4420938c3efdSQuentin Monnet * **-ENOENT** if no syn packet is available when 4421938c3efdSQuentin Monnet * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 44220813a841SMartin KaFai Lau * 4423938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space. Only *len* number of 44240813a841SMartin KaFai Lau * bytes are copied. 44250813a841SMartin KaFai Lau * 4426938c3efdSQuentin Monnet * **-EFAULT** on failure to parse the header options in the 4427938c3efdSQuentin Monnet * packet. 44280813a841SMartin KaFai Lau * 4429938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4430938c3efdSQuentin Monnet * *skops*\ **->op**. 44310813a841SMartin KaFai Lau * 44320813a841SMartin KaFai Lau * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 44330813a841SMartin KaFai Lau * Description 44340813a841SMartin KaFai Lau * Store header option. The data will be copied 44350813a841SMartin KaFai Lau * from buffer *from* with length *len* to the TCP header. 44360813a841SMartin KaFai Lau * 44370813a841SMartin KaFai Lau * The buffer *from* should have the whole option that 44380813a841SMartin KaFai Lau * includes the kind, kind-length, and the actual 44390813a841SMartin KaFai Lau * option data. The *len* must be at least kind-length 44400813a841SMartin KaFai Lau * long. The kind-length does not have to be 4 byte 44410813a841SMartin KaFai Lau * aligned. The kernel will take care of the padding 44420813a841SMartin KaFai Lau * and setting the 4 bytes aligned value to th->doff. 44430813a841SMartin KaFai Lau * 44440813a841SMartin KaFai Lau * This helper will check for duplicated option 44450813a841SMartin KaFai Lau * by searching the same option in the outgoing skb. 44460813a841SMartin KaFai Lau * 44470813a841SMartin KaFai Lau * This helper can only be called during 4448938c3efdSQuentin Monnet * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 44490813a841SMartin KaFai Lau * 44500813a841SMartin KaFai Lau * Return 44510813a841SMartin KaFai Lau * 0 on success, or negative error in case of failure: 44520813a841SMartin KaFai Lau * 4453938c3efdSQuentin Monnet * **-EINVAL** If param is invalid. 44540813a841SMartin KaFai Lau * 4455938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space in the header. 44560813a841SMartin KaFai Lau * Nothing has been written 44570813a841SMartin KaFai Lau * 4458938c3efdSQuentin Monnet * **-EEXIST** if the option already exists. 44590813a841SMartin KaFai Lau * 4460aa75622cSQuentin Monnet * **-EFAULT** on failure to parse the existing header options. 44610813a841SMartin KaFai Lau * 4462938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4463938c3efdSQuentin Monnet * *skops*\ **->op**. 44640813a841SMartin KaFai Lau * 44650813a841SMartin KaFai Lau * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 44660813a841SMartin KaFai Lau * Description 44670813a841SMartin KaFai Lau * Reserve *len* bytes for the bpf header option. The 4468938c3efdSQuentin Monnet * space will be used by **bpf_store_hdr_opt**\ () later in 4469938c3efdSQuentin Monnet * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 44700813a841SMartin KaFai Lau * 4471938c3efdSQuentin Monnet * If **bpf_reserve_hdr_opt**\ () is called multiple times, 44720813a841SMartin KaFai Lau * the total number of bytes will be reserved. 44730813a841SMartin KaFai Lau * 44740813a841SMartin KaFai Lau * This helper can only be called during 4475938c3efdSQuentin Monnet * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 44760813a841SMartin KaFai Lau * 44770813a841SMartin KaFai Lau * Return 44780813a841SMartin KaFai Lau * 0 on success, or negative error in case of failure: 44790813a841SMartin KaFai Lau * 4480938c3efdSQuentin Monnet * **-EINVAL** if a parameter is invalid. 44810813a841SMartin KaFai Lau * 4482938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space in the header. 44830813a841SMartin KaFai Lau * 4484938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 4485938c3efdSQuentin Monnet * *skops*\ **->op**. 44866e22ab9dSJiri Olsa * 44878ea63684SKP Singh * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 44888ea63684SKP Singh * Description 44898ea63684SKP Singh * Get a bpf_local_storage from an *inode*. 44908ea63684SKP Singh * 44918ea63684SKP Singh * Logically, it could be thought of as getting the value from 44928ea63684SKP Singh * a *map* with *inode* as the **key**. From this 44938ea63684SKP Singh * perspective, the usage is not much different from 44948ea63684SKP Singh * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 44958ea63684SKP Singh * helper enforces the key must be an inode and the map must also 44968ea63684SKP Singh * be a **BPF_MAP_TYPE_INODE_STORAGE**. 44978ea63684SKP Singh * 44988ea63684SKP Singh * Underneath, the value is stored locally at *inode* instead of 44998ea63684SKP Singh * the *map*. The *map* is used as the bpf-local-storage 45008ea63684SKP Singh * "type". The bpf-local-storage "type" (i.e. the *map*) is 45018ea63684SKP Singh * searched against all bpf_local_storage residing at *inode*. 45028ea63684SKP Singh * 45038ea63684SKP Singh * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 45048ea63684SKP Singh * used such that a new bpf_local_storage will be 45058ea63684SKP Singh * created if one does not exist. *value* can be used 45068ea63684SKP Singh * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 45078ea63684SKP Singh * the initial value of a bpf_local_storage. If *value* is 45088ea63684SKP Singh * **NULL**, the new bpf_local_storage will be zero initialized. 45098ea63684SKP Singh * Return 45108ea63684SKP Singh * A bpf_local_storage pointer is returned on success. 45118ea63684SKP Singh * 45128ea63684SKP Singh * **NULL** if not found or there was an error in adding 45138ea63684SKP Singh * a new bpf_local_storage. 45148ea63684SKP Singh * 45158ea63684SKP Singh * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 45168ea63684SKP Singh * Description 45178ea63684SKP Singh * Delete a bpf_local_storage from an *inode*. 45188ea63684SKP Singh * Return 45198ea63684SKP Singh * 0 on success. 45208ea63684SKP Singh * 45218ea63684SKP Singh * **-ENOENT** if the bpf_local_storage cannot be found. 45226e22ab9dSJiri Olsa * 45236e22ab9dSJiri Olsa * long bpf_d_path(struct path *path, char *buf, u32 sz) 45246e22ab9dSJiri Olsa * Description 4525938c3efdSQuentin Monnet * Return full path for given **struct path** object, which 4526938c3efdSQuentin Monnet * needs to be the kernel BTF *path* object. The path is 4527938c3efdSQuentin Monnet * returned in the provided buffer *buf* of size *sz* and 45286e22ab9dSJiri Olsa * is zero terminated. 45296e22ab9dSJiri Olsa * 45306e22ab9dSJiri Olsa * Return 45316e22ab9dSJiri Olsa * On success, the strictly positive length of the string, 45326e22ab9dSJiri Olsa * including the trailing NUL character. On error, a negative 45336e22ab9dSJiri Olsa * value. 453407be4c4aSAlexei Starovoitov * 453507be4c4aSAlexei Starovoitov * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 453607be4c4aSAlexei Starovoitov * Description 453707be4c4aSAlexei Starovoitov * Read *size* bytes from user space address *user_ptr* and store 4538938c3efdSQuentin Monnet * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 453907be4c4aSAlexei Starovoitov * Return 454007be4c4aSAlexei Starovoitov * 0 on success, or a negative error in case of failure. 4541c4d0bfb4SAlan Maguire * 4542c4d0bfb4SAlan Maguire * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4543c4d0bfb4SAlan Maguire * Description 4544c4d0bfb4SAlan Maguire * Use BTF to store a string representation of *ptr*->ptr in *str*, 4545c4d0bfb4SAlan Maguire * using *ptr*->type_id. This value should specify the type 4546c4d0bfb4SAlan Maguire * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4547c4d0bfb4SAlan Maguire * can be used to look up vmlinux BTF type ids. Traversing the 4548c4d0bfb4SAlan Maguire * data structure using BTF, the type information and values are 4549c4d0bfb4SAlan Maguire * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4550c4d0bfb4SAlan Maguire * the pointer data is carried out to avoid kernel crashes during 4551c4d0bfb4SAlan Maguire * operation. Smaller types can use string space on the stack; 4552c4d0bfb4SAlan Maguire * larger programs can use map data to store the string 4553c4d0bfb4SAlan Maguire * representation. 4554c4d0bfb4SAlan Maguire * 4555c4d0bfb4SAlan Maguire * The string can be subsequently shared with userspace via 4556c4d0bfb4SAlan Maguire * bpf_perf_event_output() or ring buffer interfaces. 4557c4d0bfb4SAlan Maguire * bpf_trace_printk() is to be avoided as it places too small 4558c4d0bfb4SAlan Maguire * a limit on string size to be useful. 4559c4d0bfb4SAlan Maguire * 4560c4d0bfb4SAlan Maguire * *flags* is a combination of 4561c4d0bfb4SAlan Maguire * 4562c4d0bfb4SAlan Maguire * **BTF_F_COMPACT** 4563c4d0bfb4SAlan Maguire * no formatting around type information 4564c4d0bfb4SAlan Maguire * **BTF_F_NONAME** 4565c4d0bfb4SAlan Maguire * no struct/union member names/types 4566c4d0bfb4SAlan Maguire * **BTF_F_PTR_RAW** 4567c4d0bfb4SAlan Maguire * show raw (unobfuscated) pointer values; 4568c4d0bfb4SAlan Maguire * equivalent to printk specifier %px. 4569c4d0bfb4SAlan Maguire * **BTF_F_ZERO** 4570c4d0bfb4SAlan Maguire * show zero-valued struct/union members; they 4571c4d0bfb4SAlan Maguire * are not displayed by default 4572c4d0bfb4SAlan Maguire * 4573c4d0bfb4SAlan Maguire * Return 4574c4d0bfb4SAlan Maguire * The number of bytes that were written (or would have been 4575c4d0bfb4SAlan Maguire * written if output had to be truncated due to string size), 4576c4d0bfb4SAlan Maguire * or a negative error in cases of failure. 4577eb411377SAlan Maguire * 4578eb411377SAlan Maguire * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4579eb411377SAlan Maguire * Description 4580eb411377SAlan Maguire * Use BTF to write to seq_write a string representation of 4581eb411377SAlan Maguire * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4582eb411377SAlan Maguire * *flags* are identical to those used for bpf_snprintf_btf. 4583eb411377SAlan Maguire * Return 4584eb411377SAlan Maguire * 0 on success or a negative error in case of failure. 4585b426ce83SDaniel Borkmann * 4586b426ce83SDaniel Borkmann * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4587b426ce83SDaniel Borkmann * Description 4588b426ce83SDaniel Borkmann * See **bpf_get_cgroup_classid**\ () for the main description. 4589b426ce83SDaniel Borkmann * This helper differs from **bpf_get_cgroup_classid**\ () in that 4590b426ce83SDaniel Borkmann * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4591b426ce83SDaniel Borkmann * associated socket instead of the current process. 4592b426ce83SDaniel Borkmann * Return 4593b426ce83SDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 4594b4ab3141SDaniel Borkmann * 4595ba452c9eSToke Høiland-Jørgensen * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4596b4ab3141SDaniel Borkmann * Description 4597b4ab3141SDaniel Borkmann * Redirect the packet to another net device of index *ifindex* 4598b4ab3141SDaniel Borkmann * and fill in L2 addresses from neighboring subsystem. This helper 4599b4ab3141SDaniel Borkmann * is somewhat similar to **bpf_redirect**\ (), except that it 4600dd2ce6a5SDaniel Borkmann * populates L2 addresses as well, meaning, internally, the helper 4601ba452c9eSToke Høiland-Jørgensen * relies on the neighbor lookup for the L2 address of the nexthop. 4602ba452c9eSToke Høiland-Jørgensen * 4603ba452c9eSToke Høiland-Jørgensen * The helper will perform a FIB lookup based on the skb's 4604ba452c9eSToke Høiland-Jørgensen * networking header to get the address of the next hop, unless 4605ba452c9eSToke Høiland-Jørgensen * this is supplied by the caller in the *params* argument. The 4606ba452c9eSToke Høiland-Jørgensen * *plen* argument indicates the len of *params* and should be set 4607ba452c9eSToke Høiland-Jørgensen * to 0 if *params* is NULL. 4608dd2ce6a5SDaniel Borkmann * 4609b4ab3141SDaniel Borkmann * The *flags* argument is reserved and must be 0. The helper is 4610dd2ce6a5SDaniel Borkmann * currently only supported for tc BPF program types, and enabled 4611dd2ce6a5SDaniel Borkmann * for IPv4 and IPv6 protocols. 4612b4ab3141SDaniel Borkmann * Return 4613b4ab3141SDaniel Borkmann * The helper returns **TC_ACT_REDIRECT** on success or 4614b4ab3141SDaniel Borkmann * **TC_ACT_SHOT** on error. 4615eaa6bcb7SHao Luo * 4616eaa6bcb7SHao Luo * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4617eaa6bcb7SHao Luo * Description 4618eaa6bcb7SHao Luo * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4619eaa6bcb7SHao Luo * pointer to the percpu kernel variable on *cpu*. A ksym is an 4620eaa6bcb7SHao Luo * extern variable decorated with '__ksym'. For ksym, there is a 4621eaa6bcb7SHao Luo * global var (either static or global) defined of the same name 4622eaa6bcb7SHao Luo * in the kernel. The ksym is percpu if the global var is percpu. 4623eaa6bcb7SHao Luo * The returned pointer points to the global percpu var on *cpu*. 4624eaa6bcb7SHao Luo * 4625eaa6bcb7SHao Luo * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4626eaa6bcb7SHao Luo * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4627eaa6bcb7SHao Luo * happens if *cpu* is larger than nr_cpu_ids. The caller of 4628eaa6bcb7SHao Luo * bpf_per_cpu_ptr() must check the returned value. 4629eaa6bcb7SHao Luo * Return 4630eaa6bcb7SHao Luo * A pointer pointing to the kernel percpu variable on *cpu*, or 4631eaa6bcb7SHao Luo * NULL, if *cpu* is invalid. 463263d9b80dSHao Luo * 463363d9b80dSHao Luo * void *bpf_this_cpu_ptr(const void *percpu_ptr) 463463d9b80dSHao Luo * Description 463563d9b80dSHao Luo * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 463663d9b80dSHao Luo * pointer to the percpu kernel variable on this cpu. See the 463763d9b80dSHao Luo * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 463863d9b80dSHao Luo * 463963d9b80dSHao Luo * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 464063d9b80dSHao Luo * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 464163d9b80dSHao Luo * never return NULL. 464263d9b80dSHao Luo * Return 464363d9b80dSHao Luo * A pointer pointing to the kernel percpu variable on this cpu. 46449aa1206eSDaniel Borkmann * 46459aa1206eSDaniel Borkmann * long bpf_redirect_peer(u32 ifindex, u64 flags) 46469aa1206eSDaniel Borkmann * Description 46479aa1206eSDaniel Borkmann * Redirect the packet to another net device of index *ifindex*. 46489aa1206eSDaniel Borkmann * This helper is somewhat similar to **bpf_redirect**\ (), except 46499aa1206eSDaniel Borkmann * that the redirection happens to the *ifindex*' peer device and 46509aa1206eSDaniel Borkmann * the netns switch takes place from ingress to ingress without 46519aa1206eSDaniel Borkmann * going through the CPU's backlog queue. 46529aa1206eSDaniel Borkmann * 46539aa1206eSDaniel Borkmann * The *flags* argument is reserved and must be 0. The helper is 46549aa1206eSDaniel Borkmann * currently only supported for tc BPF program types at the ingress 46559aa1206eSDaniel Borkmann * hook and for veth device types. The peer device must reside in a 46569aa1206eSDaniel Borkmann * different network namespace. 46579aa1206eSDaniel Borkmann * Return 46589aa1206eSDaniel Borkmann * The helper returns **TC_ACT_REDIRECT** on success or 46599aa1206eSDaniel Borkmann * **TC_ACT_SHOT** on error. 46604cf1bc1fSKP Singh * 46614cf1bc1fSKP Singh * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 46624cf1bc1fSKP Singh * Description 46634cf1bc1fSKP Singh * Get a bpf_local_storage from the *task*. 46644cf1bc1fSKP Singh * 46654cf1bc1fSKP Singh * Logically, it could be thought of as getting the value from 46664cf1bc1fSKP Singh * a *map* with *task* as the **key**. From this 46674cf1bc1fSKP Singh * perspective, the usage is not much different from 46684cf1bc1fSKP Singh * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4669aa75622cSQuentin Monnet * helper enforces the key must be a task_struct and the map must also 46704cf1bc1fSKP Singh * be a **BPF_MAP_TYPE_TASK_STORAGE**. 46714cf1bc1fSKP Singh * 46724cf1bc1fSKP Singh * Underneath, the value is stored locally at *task* instead of 46734cf1bc1fSKP Singh * the *map*. The *map* is used as the bpf-local-storage 46744cf1bc1fSKP Singh * "type". The bpf-local-storage "type" (i.e. the *map*) is 46754cf1bc1fSKP Singh * searched against all bpf_local_storage residing at *task*. 46764cf1bc1fSKP Singh * 46774cf1bc1fSKP Singh * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 46784cf1bc1fSKP Singh * used such that a new bpf_local_storage will be 46794cf1bc1fSKP Singh * created if one does not exist. *value* can be used 46804cf1bc1fSKP Singh * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 46814cf1bc1fSKP Singh * the initial value of a bpf_local_storage. If *value* is 46824cf1bc1fSKP Singh * **NULL**, the new bpf_local_storage will be zero initialized. 46834cf1bc1fSKP Singh * Return 46844cf1bc1fSKP Singh * A bpf_local_storage pointer is returned on success. 46854cf1bc1fSKP Singh * 46864cf1bc1fSKP Singh * **NULL** if not found or there was an error in adding 46874cf1bc1fSKP Singh * a new bpf_local_storage. 46884cf1bc1fSKP Singh * 46894cf1bc1fSKP Singh * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 46904cf1bc1fSKP Singh * Description 46914cf1bc1fSKP Singh * Delete a bpf_local_storage from a *task*. 46924cf1bc1fSKP Singh * Return 46934cf1bc1fSKP Singh * 0 on success. 46944cf1bc1fSKP Singh * 46954cf1bc1fSKP Singh * **-ENOENT** if the bpf_local_storage cannot be found. 46963ca1032aSKP Singh * 46973ca1032aSKP Singh * struct task_struct *bpf_get_current_task_btf(void) 46983ca1032aSKP Singh * Description 46993ca1032aSKP Singh * Return a BTF pointer to the "current" task. 47003ca1032aSKP Singh * This pointer can also be used in helpers that accept an 47013ca1032aSKP Singh * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 47023ca1032aSKP Singh * Return 47033ca1032aSKP Singh * Pointer to the current task. 47043f6719c7SKP Singh * 47053f6719c7SKP Singh * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 47063f6719c7SKP Singh * Description 47073f6719c7SKP Singh * Set or clear certain options on *bprm*: 47083f6719c7SKP Singh * 47093f6719c7SKP Singh * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 47103f6719c7SKP Singh * which sets the **AT_SECURE** auxv for glibc. The bit 47113f6719c7SKP Singh * is cleared if the flag is not specified. 47123f6719c7SKP Singh * Return 47133f6719c7SKP Singh * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4714d0551261SDmitrii Banshchikov * 4715d0551261SDmitrii Banshchikov * u64 bpf_ktime_get_coarse_ns(void) 4716d0551261SDmitrii Banshchikov * Description 4717d0551261SDmitrii Banshchikov * Return a coarse-grained version of the time elapsed since 4718d0551261SDmitrii Banshchikov * system boot, in nanoseconds. Does not include time the system 4719d0551261SDmitrii Banshchikov * was suspended. 4720d0551261SDmitrii Banshchikov * 4721d0551261SDmitrii Banshchikov * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4722d0551261SDmitrii Banshchikov * Return 4723d0551261SDmitrii Banshchikov * Current *ktime*. 472427672f0dSKP Singh * 472527672f0dSKP Singh * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 472627672f0dSKP Singh * Description 4727aa75622cSQuentin Monnet * Returns the stored IMA hash of the *inode* (if it's available). 472827672f0dSKP Singh * If the hash is larger than *size*, then only *size* 472927672f0dSKP Singh * bytes will be copied to *dst* 473027672f0dSKP Singh * Return 473127672f0dSKP Singh * The **hash_algo** is returned on success, 473227672f0dSKP Singh * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 473327672f0dSKP Singh * invalid arguments are passed. 47344f19cab7SFlorent Revest * 47354f19cab7SFlorent Revest * struct socket *bpf_sock_from_file(struct file *file) 47364f19cab7SFlorent Revest * Description 47374f19cab7SFlorent Revest * If the given file represents a socket, returns the associated 47384f19cab7SFlorent Revest * socket. 47394f19cab7SFlorent Revest * Return 47404f19cab7SFlorent Revest * A pointer to a struct socket on success or NULL if the file is 47414f19cab7SFlorent Revest * not a socket. 474234b2021cSJesper Dangaard Brouer * 474334b2021cSJesper Dangaard Brouer * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 474434b2021cSJesper Dangaard Brouer * Description 4745e5e35e75SJesper Dangaard Brouer * Check packet size against exceeding MTU of net device (based 474634b2021cSJesper Dangaard Brouer * on *ifindex*). This helper will likely be used in combination 474734b2021cSJesper Dangaard Brouer * with helpers that adjust/change the packet size. 474834b2021cSJesper Dangaard Brouer * 474934b2021cSJesper Dangaard Brouer * The argument *len_diff* can be used for querying with a planned 475034b2021cSJesper Dangaard Brouer * size change. This allows to check MTU prior to changing packet 4751aa75622cSQuentin Monnet * ctx. Providing a *len_diff* adjustment that is larger than the 475234b2021cSJesper Dangaard Brouer * actual packet size (resulting in negative packet size) will in 4753aa75622cSQuentin Monnet * principle not exceed the MTU, which is why it is not considered 4754aa75622cSQuentin Monnet * a failure. Other BPF helpers are needed for performing the 4755aa75622cSQuentin Monnet * planned size change; therefore the responsibility for catching 4756aa75622cSQuentin Monnet * a negative packet size belongs in those helpers. 475734b2021cSJesper Dangaard Brouer * 475834b2021cSJesper Dangaard Brouer * Specifying *ifindex* zero means the MTU check is performed 475934b2021cSJesper Dangaard Brouer * against the current net device. This is practical if this isn't 476034b2021cSJesper Dangaard Brouer * used prior to redirect. 476134b2021cSJesper Dangaard Brouer * 4762e5e35e75SJesper Dangaard Brouer * On input *mtu_len* must be a valid pointer, else verifier will 4763e5e35e75SJesper Dangaard Brouer * reject BPF program. If the value *mtu_len* is initialized to 4764e5e35e75SJesper Dangaard Brouer * zero then the ctx packet size is use. When value *mtu_len* is 4765e5e35e75SJesper Dangaard Brouer * provided as input this specify the L3 length that the MTU check 4766e5e35e75SJesper Dangaard Brouer * is done against. Remember XDP and TC length operate at L2, but 4767e5e35e75SJesper Dangaard Brouer * this value is L3 as this correlate to MTU and IP-header tot_len 4768e5e35e75SJesper Dangaard Brouer * values which are L3 (similar behavior as bpf_fib_lookup). 4769e5e35e75SJesper Dangaard Brouer * 477034b2021cSJesper Dangaard Brouer * The Linux kernel route table can configure MTUs on a more 477134b2021cSJesper Dangaard Brouer * specific per route level, which is not provided by this helper. 477234b2021cSJesper Dangaard Brouer * For route level MTU checks use the **bpf_fib_lookup**\ () 477334b2021cSJesper Dangaard Brouer * helper. 477434b2021cSJesper Dangaard Brouer * 477534b2021cSJesper Dangaard Brouer * *ctx* is either **struct xdp_md** for XDP programs or 477634b2021cSJesper Dangaard Brouer * **struct sk_buff** for tc cls_act programs. 477734b2021cSJesper Dangaard Brouer * 477834b2021cSJesper Dangaard Brouer * The *flags* argument can be a combination of one or more of the 477934b2021cSJesper Dangaard Brouer * following values: 478034b2021cSJesper Dangaard Brouer * 478134b2021cSJesper Dangaard Brouer * **BPF_MTU_CHK_SEGS** 478234b2021cSJesper Dangaard Brouer * This flag will only works for *ctx* **struct sk_buff**. 478334b2021cSJesper Dangaard Brouer * If packet context contains extra packet segment buffers 478434b2021cSJesper Dangaard Brouer * (often knows as GSO skb), then MTU check is harder to 478534b2021cSJesper Dangaard Brouer * check at this point, because in transmit path it is 478634b2021cSJesper Dangaard Brouer * possible for the skb packet to get re-segmented 478734b2021cSJesper Dangaard Brouer * (depending on net device features). This could still be 478834b2021cSJesper Dangaard Brouer * a MTU violation, so this flag enables performing MTU 478934b2021cSJesper Dangaard Brouer * check against segments, with a different violation 479034b2021cSJesper Dangaard Brouer * return code to tell it apart. Check cannot use len_diff. 479134b2021cSJesper Dangaard Brouer * 479234b2021cSJesper Dangaard Brouer * On return *mtu_len* pointer contains the MTU value of the net 479334b2021cSJesper Dangaard Brouer * device. Remember the net device configured MTU is the L3 size, 4794e5e35e75SJesper Dangaard Brouer * which is returned here and XDP and TC length operate at L2. 479534b2021cSJesper Dangaard Brouer * Helper take this into account for you, but remember when using 4796e5e35e75SJesper Dangaard Brouer * MTU value in your BPF-code. 479734b2021cSJesper Dangaard Brouer * 479834b2021cSJesper Dangaard Brouer * Return 479934b2021cSJesper Dangaard Brouer * * 0 on success, and populate MTU value in *mtu_len* pointer. 480034b2021cSJesper Dangaard Brouer * 480134b2021cSJesper Dangaard Brouer * * < 0 if any input argument is invalid (*mtu_len* not updated) 480234b2021cSJesper Dangaard Brouer * 480334b2021cSJesper Dangaard Brouer * MTU violations return positive values, but also populate MTU 480434b2021cSJesper Dangaard Brouer * value in *mtu_len* pointer, as this can be needed for 480534b2021cSJesper Dangaard Brouer * implementing PMTU handing: 480634b2021cSJesper Dangaard Brouer * 480734b2021cSJesper Dangaard Brouer * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 480834b2021cSJesper Dangaard Brouer * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 480934b2021cSJesper Dangaard Brouer * 481069c087baSYonghong Song * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 481169c087baSYonghong Song * Description 481269c087baSYonghong Song * For each element in **map**, call **callback_fn** function with 481369c087baSYonghong Song * **map**, **callback_ctx** and other map-specific parameters. 481469c087baSYonghong Song * The **callback_fn** should be a static function and 481569c087baSYonghong Song * the **callback_ctx** should be a pointer to the stack. 481669c087baSYonghong Song * The **flags** is used to control certain aspects of the helper. 481769c087baSYonghong Song * Currently, the **flags** must be 0. 481869c087baSYonghong Song * 481969c087baSYonghong Song * The following are a list of supported map types and their 482069c087baSYonghong Song * respective expected callback signatures: 482169c087baSYonghong Song * 482269c087baSYonghong Song * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 482369c087baSYonghong Song * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 482469c087baSYonghong Song * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 482569c087baSYonghong Song * 482669c087baSYonghong Song * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 482769c087baSYonghong Song * 482869c087baSYonghong Song * For per_cpu maps, the map_value is the value on the cpu where the 482969c087baSYonghong Song * bpf_prog is running. 483069c087baSYonghong Song * 483169c087baSYonghong Song * If **callback_fn** return 0, the helper will continue to the next 483269c087baSYonghong Song * element. If return value is 1, the helper will skip the rest of 483369c087baSYonghong Song * elements and return. Other return values are not used now. 483469c087baSYonghong Song * 483569c087baSYonghong Song * Return 483669c087baSYonghong Song * The number of traversed map elements for success, **-EINVAL** for 483769c087baSYonghong Song * invalid **flags**. 48387b15523aSFlorent Revest * 48397b15523aSFlorent Revest * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 48407b15523aSFlorent Revest * Description 48417b15523aSFlorent Revest * Outputs a string into the **str** buffer of size **str_size** 48427b15523aSFlorent Revest * based on a format string stored in a read-only map pointed by 48437b15523aSFlorent Revest * **fmt**. 48447b15523aSFlorent Revest * 48457b15523aSFlorent Revest * Each format specifier in **fmt** corresponds to one u64 element 48467b15523aSFlorent Revest * in the **data** array. For strings and pointers where pointees 48477b15523aSFlorent Revest * are accessed, only the pointer values are stored in the *data* 4848a42effb0SDave Marchevsky * array. The *data_len* is the size of *data* in bytes - must be 4849a42effb0SDave Marchevsky * a multiple of 8. 48507b15523aSFlorent Revest * 48517b15523aSFlorent Revest * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 48527b15523aSFlorent Revest * memory. Reading kernel memory may fail due to either invalid 48537b15523aSFlorent Revest * address or valid address but requiring a major memory fault. If 48547b15523aSFlorent Revest * reading kernel memory fails, the string for **%s** will be an 48557b15523aSFlorent Revest * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 48567b15523aSFlorent Revest * Not returning error to bpf program is consistent with what 48577b15523aSFlorent Revest * **bpf_trace_printk**\ () does for now. 48587b15523aSFlorent Revest * 48597b15523aSFlorent Revest * Return 48607b15523aSFlorent Revest * The strictly positive length of the formatted string, including 48617b15523aSFlorent Revest * the trailing zero character. If the return value is greater than 48627b15523aSFlorent Revest * **str_size**, **str** contains a truncated string, guaranteed to 48637b15523aSFlorent Revest * be zero-terminated except when **str_size** is 0. 48647b15523aSFlorent Revest * 48657b15523aSFlorent Revest * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 486679a7f8bdSAlexei Starovoitov * 486779a7f8bdSAlexei Starovoitov * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 486879a7f8bdSAlexei Starovoitov * Description 486979a7f8bdSAlexei Starovoitov * Execute bpf syscall with given arguments. 487079a7f8bdSAlexei Starovoitov * Return 487179a7f8bdSAlexei Starovoitov * A syscall result. 48723d78417bSAlexei Starovoitov * 48733d78417bSAlexei Starovoitov * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 48743d78417bSAlexei Starovoitov * Description 48753d78417bSAlexei Starovoitov * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 48763d78417bSAlexei Starovoitov * Return 48773d78417bSAlexei Starovoitov * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 48783abea089SAlexei Starovoitov * 48793abea089SAlexei Starovoitov * long bpf_sys_close(u32 fd) 48803abea089SAlexei Starovoitov * Description 48813abea089SAlexei Starovoitov * Execute close syscall for given FD. 48823abea089SAlexei Starovoitov * Return 48833abea089SAlexei Starovoitov * A syscall result. 4884b00628b1SAlexei Starovoitov * 4885b00628b1SAlexei Starovoitov * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 4886b00628b1SAlexei Starovoitov * Description 4887b00628b1SAlexei Starovoitov * Initialize the timer. 4888b00628b1SAlexei Starovoitov * First 4 bits of *flags* specify clockid. 4889b00628b1SAlexei Starovoitov * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 4890b00628b1SAlexei Starovoitov * All other bits of *flags* are reserved. 4891b00628b1SAlexei Starovoitov * The verifier will reject the program if *timer* is not from 4892b00628b1SAlexei Starovoitov * the same *map*. 4893b00628b1SAlexei Starovoitov * Return 4894b00628b1SAlexei Starovoitov * 0 on success. 4895b00628b1SAlexei Starovoitov * **-EBUSY** if *timer* is already initialized. 4896b00628b1SAlexei Starovoitov * **-EINVAL** if invalid *flags* are passed. 4897b00628b1SAlexei Starovoitov * **-EPERM** if *timer* is in a map that doesn't have any user references. 4898b00628b1SAlexei Starovoitov * The user space should either hold a file descriptor to a map with timers 4899b00628b1SAlexei Starovoitov * or pin such map in bpffs. When map is unpinned or file descriptor is 4900b00628b1SAlexei Starovoitov * closed all timers in the map will be cancelled and freed. 4901b00628b1SAlexei Starovoitov * 4902b00628b1SAlexei Starovoitov * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 4903b00628b1SAlexei Starovoitov * Description 4904b00628b1SAlexei Starovoitov * Configure the timer to call *callback_fn* static function. 4905b00628b1SAlexei Starovoitov * Return 4906b00628b1SAlexei Starovoitov * 0 on success. 4907b00628b1SAlexei Starovoitov * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4908b00628b1SAlexei Starovoitov * **-EPERM** if *timer* is in a map that doesn't have any user references. 4909b00628b1SAlexei Starovoitov * The user space should either hold a file descriptor to a map with timers 4910b00628b1SAlexei Starovoitov * or pin such map in bpffs. When map is unpinned or file descriptor is 4911b00628b1SAlexei Starovoitov * closed all timers in the map will be cancelled and freed. 4912b00628b1SAlexei Starovoitov * 4913b00628b1SAlexei Starovoitov * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 4914b00628b1SAlexei Starovoitov * Description 4915b00628b1SAlexei Starovoitov * Set timer expiration N nanoseconds from the current time. The 4916b00628b1SAlexei Starovoitov * configured callback will be invoked in soft irq context on some cpu 4917b00628b1SAlexei Starovoitov * and will not repeat unless another bpf_timer_start() is made. 4918b00628b1SAlexei Starovoitov * In such case the next invocation can migrate to a different cpu. 4919b00628b1SAlexei Starovoitov * Since struct bpf_timer is a field inside map element the map 4920b00628b1SAlexei Starovoitov * owns the timer. The bpf_timer_set_callback() will increment refcnt 4921b00628b1SAlexei Starovoitov * of BPF program to make sure that callback_fn code stays valid. 4922b00628b1SAlexei Starovoitov * When user space reference to a map reaches zero all timers 4923b00628b1SAlexei Starovoitov * in a map are cancelled and corresponding program's refcnts are 4924b00628b1SAlexei Starovoitov * decremented. This is done to make sure that Ctrl-C of a user 4925b00628b1SAlexei Starovoitov * process doesn't leave any timers running. If map is pinned in 4926b00628b1SAlexei Starovoitov * bpffs the callback_fn can re-arm itself indefinitely. 4927b00628b1SAlexei Starovoitov * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 4928b00628b1SAlexei Starovoitov * cancel and free the timer in the given map element. 4929b00628b1SAlexei Starovoitov * The map can contain timers that invoke callback_fn-s from different 4930b00628b1SAlexei Starovoitov * programs. The same callback_fn can serve different timers from 4931b00628b1SAlexei Starovoitov * different maps if key/value layout matches across maps. 4932b00628b1SAlexei Starovoitov * Every bpf_timer_set_callback() can have different callback_fn. 4933b00628b1SAlexei Starovoitov * 4934b00628b1SAlexei Starovoitov * Return 4935b00628b1SAlexei Starovoitov * 0 on success. 4936b00628b1SAlexei Starovoitov * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 4937b00628b1SAlexei Starovoitov * or invalid *flags* are passed. 4938b00628b1SAlexei Starovoitov * 4939b00628b1SAlexei Starovoitov * long bpf_timer_cancel(struct bpf_timer *timer) 4940b00628b1SAlexei Starovoitov * Description 4941b00628b1SAlexei Starovoitov * Cancel the timer and wait for callback_fn to finish if it was running. 4942b00628b1SAlexei Starovoitov * Return 4943b00628b1SAlexei Starovoitov * 0 if the timer was not active. 4944b00628b1SAlexei Starovoitov * 1 if the timer was active. 4945b00628b1SAlexei Starovoitov * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4946b00628b1SAlexei Starovoitov * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 4947b00628b1SAlexei Starovoitov * own timer which would have led to a deadlock otherwise. 49489b99edcaSJiri Olsa * 49499b99edcaSJiri Olsa * u64 bpf_get_func_ip(void *ctx) 49509b99edcaSJiri Olsa * Description 49519ffd9f3fSJiri Olsa * Get address of the traced function (for tracing and kprobe programs). 49529b99edcaSJiri Olsa * Return 49539b99edcaSJiri Olsa * Address of the traced function. 49547adfc6c9SAndrii Nakryiko * 49557adfc6c9SAndrii Nakryiko * u64 bpf_get_attach_cookie(void *ctx) 49567adfc6c9SAndrii Nakryiko * Description 49577adfc6c9SAndrii Nakryiko * Get bpf_cookie value provided (optionally) during the program 49587adfc6c9SAndrii Nakryiko * attachment. It might be different for each individual 49597adfc6c9SAndrii Nakryiko * attachment, even if BPF program itself is the same. 49607adfc6c9SAndrii Nakryiko * Expects BPF program context *ctx* as a first argument. 49617adfc6c9SAndrii Nakryiko * 49627adfc6c9SAndrii Nakryiko * Supported for the following program types: 49637adfc6c9SAndrii Nakryiko * - kprobe/uprobe; 49647adfc6c9SAndrii Nakryiko * - tracepoint; 49657adfc6c9SAndrii Nakryiko * - perf_event. 49667adfc6c9SAndrii Nakryiko * Return 49677adfc6c9SAndrii Nakryiko * Value specified by user at BPF link creation/attachment time 49687adfc6c9SAndrii Nakryiko * or 0, if it was not specified. 4969dd6e10fbSDaniel Xu * 4970dd6e10fbSDaniel Xu * long bpf_task_pt_regs(struct task_struct *task) 4971dd6e10fbSDaniel Xu * Description 4972dd6e10fbSDaniel Xu * Get the struct pt_regs associated with **task**. 4973dd6e10fbSDaniel Xu * Return 4974dd6e10fbSDaniel Xu * A pointer to struct pt_regs. 4975856c02dbSSong Liu * 4976856c02dbSSong Liu * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 4977856c02dbSSong Liu * Description 4978856c02dbSSong Liu * Get branch trace from hardware engines like Intel LBR. The 4979856c02dbSSong Liu * hardware engine is stopped shortly after the helper is 4980856c02dbSSong Liu * called. Therefore, the user need to filter branch entries 4981856c02dbSSong Liu * based on the actual use case. To capture branch trace 4982856c02dbSSong Liu * before the trigger point of the BPF program, the helper 4983856c02dbSSong Liu * should be called at the beginning of the BPF program. 4984856c02dbSSong Liu * 4985856c02dbSSong Liu * The data is stored as struct perf_branch_entry into output 4986856c02dbSSong Liu * buffer *entries*. *size* is the size of *entries* in bytes. 4987856c02dbSSong Liu * *flags* is reserved for now and must be zero. 4988856c02dbSSong Liu * 4989856c02dbSSong Liu * Return 4990856c02dbSSong Liu * On success, number of bytes written to *buf*. On error, a 4991856c02dbSSong Liu * negative value. 4992856c02dbSSong Liu * 4993856c02dbSSong Liu * **-EINVAL** if *flags* is not zero. 4994856c02dbSSong Liu * 4995856c02dbSSong Liu * **-ENOENT** if architecture does not support branch records. 499610aceb62SDave Marchevsky * 499710aceb62SDave Marchevsky * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 499810aceb62SDave Marchevsky * Description 499910aceb62SDave Marchevsky * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 500010aceb62SDave Marchevsky * to format and can handle more format args as a result. 500110aceb62SDave Marchevsky * 500210aceb62SDave Marchevsky * Arguments are to be used as in **bpf_seq_printf**\ () helper. 500310aceb62SDave Marchevsky * Return 500410aceb62SDave Marchevsky * The number of bytes written to the buffer, or a negative error 500510aceb62SDave Marchevsky * in case of failure. 50069eeb3aa3SHengqi Chen * 50079eeb3aa3SHengqi Chen * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 50089eeb3aa3SHengqi Chen * Description 50099eeb3aa3SHengqi Chen * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 50109eeb3aa3SHengqi Chen * Return 50119eeb3aa3SHengqi Chen * *sk* if casting is valid, or **NULL** otherwise. 5012d6aef08aSKumar Kartikeya Dwivedi * 5013d6aef08aSKumar Kartikeya Dwivedi * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 5014d6aef08aSKumar Kartikeya Dwivedi * Description 5015d6aef08aSKumar Kartikeya Dwivedi * Get the address of a kernel symbol, returned in *res*. *res* is 5016d6aef08aSKumar Kartikeya Dwivedi * set to 0 if the symbol is not found. 5017d6aef08aSKumar Kartikeya Dwivedi * Return 5018d6aef08aSKumar Kartikeya Dwivedi * On success, zero. On error, a negative value. 5019d6aef08aSKumar Kartikeya Dwivedi * 5020d6aef08aSKumar Kartikeya Dwivedi * **-EINVAL** if *flags* is not zero. 5021d6aef08aSKumar Kartikeya Dwivedi * 5022d6aef08aSKumar Kartikeya Dwivedi * **-EINVAL** if string *name* is not the same size as *name_sz*. 5023d6aef08aSKumar Kartikeya Dwivedi * 5024d6aef08aSKumar Kartikeya Dwivedi * **-ENOENT** if symbol is not found. 5025d6aef08aSKumar Kartikeya Dwivedi * 5026d6aef08aSKumar Kartikeya Dwivedi * **-EPERM** if caller does not have permission to obtain kernel address. 50277c7e3d31SSong Liu * 50287c7e3d31SSong Liu * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 50297c7e3d31SSong Liu * Description 50307c7e3d31SSong Liu * Find vma of *task* that contains *addr*, call *callback_fn* 50317c7e3d31SSong Liu * function with *task*, *vma*, and *callback_ctx*. 50327c7e3d31SSong Liu * The *callback_fn* should be a static function and 50337c7e3d31SSong Liu * the *callback_ctx* should be a pointer to the stack. 50347c7e3d31SSong Liu * The *flags* is used to control certain aspects of the helper. 50357c7e3d31SSong Liu * Currently, the *flags* must be 0. 50367c7e3d31SSong Liu * 50377c7e3d31SSong Liu * The expected callback signature is 50387c7e3d31SSong Liu * 50397c7e3d31SSong Liu * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 50407c7e3d31SSong Liu * 50417c7e3d31SSong Liu * Return 50427c7e3d31SSong Liu * 0 on success. 50437c7e3d31SSong Liu * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 50447c7e3d31SSong Liu * **-EBUSY** if failed to try lock mmap_lock. 50457c7e3d31SSong Liu * **-EINVAL** for invalid **flags**. 5046e6f2dd0fSJoanne Koong * 5047e6f2dd0fSJoanne Koong * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 5048e6f2dd0fSJoanne Koong * Description 5049e6f2dd0fSJoanne Koong * For **nr_loops**, call **callback_fn** function 5050e6f2dd0fSJoanne Koong * with **callback_ctx** as the context parameter. 5051e6f2dd0fSJoanne Koong * The **callback_fn** should be a static function and 5052e6f2dd0fSJoanne Koong * the **callback_ctx** should be a pointer to the stack. 5053e6f2dd0fSJoanne Koong * The **flags** is used to control certain aspects of the helper. 5054e6f2dd0fSJoanne Koong * Currently, the **flags** must be 0. Currently, nr_loops is 5055e6f2dd0fSJoanne Koong * limited to 1 << 23 (~8 million) loops. 5056e6f2dd0fSJoanne Koong * 5057e6f2dd0fSJoanne Koong * long (\*callback_fn)(u32 index, void \*ctx); 5058e6f2dd0fSJoanne Koong * 5059e6f2dd0fSJoanne Koong * where **index** is the current index in the loop. The index 5060e6f2dd0fSJoanne Koong * is zero-indexed. 5061e6f2dd0fSJoanne Koong * 5062e6f2dd0fSJoanne Koong * If **callback_fn** returns 0, the helper will continue to the next 5063e6f2dd0fSJoanne Koong * loop. If return value is 1, the helper will skip the rest of 5064e6f2dd0fSJoanne Koong * the loops and return. Other return values are not used now, 5065e6f2dd0fSJoanne Koong * and will be rejected by the verifier. 5066e6f2dd0fSJoanne Koong * 5067e6f2dd0fSJoanne Koong * Return 5068e6f2dd0fSJoanne Koong * The number of loops performed, **-EINVAL** for invalid **flags**, 5069e6f2dd0fSJoanne Koong * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5070c5fb1993SHou Tao * 5071c5fb1993SHou Tao * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5072c5fb1993SHou Tao * Description 5073c5fb1993SHou Tao * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5074c5fb1993SHou Tao * to be null-terminated and **s1_sz** is the maximum storage 5075c5fb1993SHou Tao * size of **s1**. **s2** must be a read-only string. 5076c5fb1993SHou Tao * Return 5077c5fb1993SHou Tao * An integer less than, equal to, or greater than zero 5078c5fb1993SHou Tao * if the first **s1_sz** bytes of **s1** is found to be 5079c5fb1993SHou Tao * less than, to match, or be greater than **s2**. 5080f92c1e18SJiri Olsa * 5081f92c1e18SJiri Olsa * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5082f92c1e18SJiri Olsa * Description 508327ed9353SYonghong Song * Get **n**-th argument register (zero based) of the traced function (for tracing programs) 5084f92c1e18SJiri Olsa * returned in **value**. 5085f92c1e18SJiri Olsa * 5086f92c1e18SJiri Olsa * Return 5087f92c1e18SJiri Olsa * 0 on success. 508827ed9353SYonghong Song * **-EINVAL** if n >= argument register count of traced function. 5089f92c1e18SJiri Olsa * 5090f92c1e18SJiri Olsa * long bpf_get_func_ret(void *ctx, u64 *value) 5091f92c1e18SJiri Olsa * Description 5092f92c1e18SJiri Olsa * Get return value of the traced function (for tracing programs) 5093f92c1e18SJiri Olsa * in **value**. 5094f92c1e18SJiri Olsa * 5095f92c1e18SJiri Olsa * Return 5096f92c1e18SJiri Olsa * 0 on success. 5097f92c1e18SJiri Olsa * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5098f92c1e18SJiri Olsa * 5099f92c1e18SJiri Olsa * long bpf_get_func_arg_cnt(void *ctx) 5100f92c1e18SJiri Olsa * Description 510127ed9353SYonghong Song * Get number of registers of the traced function (for tracing programs) where 510227ed9353SYonghong Song * function arguments are stored in these registers. 5103f92c1e18SJiri Olsa * 5104f92c1e18SJiri Olsa * Return 510527ed9353SYonghong Song * The number of argument registers of the traced function. 5106b44123b4SYiFei Zhu * 5107b44123b4SYiFei Zhu * int bpf_get_retval(void) 5108b44123b4SYiFei Zhu * Description 51092172fb80SStanislav Fomichev * Get the BPF program's return value that will be returned to the upper layers. 5110b44123b4SYiFei Zhu * 51112172fb80SStanislav Fomichev * This helper is currently supported by cgroup programs and only by the hooks 51122172fb80SStanislav Fomichev * where BPF program's return value is returned to the userspace via errno. 5113b44123b4SYiFei Zhu * Return 51142172fb80SStanislav Fomichev * The BPF program's return value. 5115b44123b4SYiFei Zhu * 5116b44123b4SYiFei Zhu * int bpf_set_retval(int retval) 5117b44123b4SYiFei Zhu * Description 51182172fb80SStanislav Fomichev * Set the BPF program's return value that will be returned to the upper layers. 5119b44123b4SYiFei Zhu * 51202172fb80SStanislav Fomichev * This helper is currently supported by cgroup programs and only by the hooks 51212172fb80SStanislav Fomichev * where BPF program's return value is returned to the userspace via errno. 51222172fb80SStanislav Fomichev * 51232172fb80SStanislav Fomichev * Note that there is the following corner case where the program exports an error 51242172fb80SStanislav Fomichev * via bpf_set_retval but signals success via 'return 1': 51252172fb80SStanislav Fomichev * 51262172fb80SStanislav Fomichev * bpf_set_retval(-EPERM); 51272172fb80SStanislav Fomichev * return 1; 51282172fb80SStanislav Fomichev * 51292172fb80SStanislav Fomichev * In this case, the BPF program's return value will use helper's -EPERM. This 51302172fb80SStanislav Fomichev * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case. 51312172fb80SStanislav Fomichev * 5132b44123b4SYiFei Zhu * Return 5133b44123b4SYiFei Zhu * 0 on success, or a negative error in case of failure. 51340165cc81SLorenzo Bianconi * 51350165cc81SLorenzo Bianconi * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 51360165cc81SLorenzo Bianconi * Description 51370165cc81SLorenzo Bianconi * Get the total size of a given xdp buff (linear and paged area) 51380165cc81SLorenzo Bianconi * Return 51390165cc81SLorenzo Bianconi * The total size of a given xdp buffer. 51403f364222SLorenzo Bianconi * 51413f364222SLorenzo Bianconi * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 51423f364222SLorenzo Bianconi * Description 51433f364222SLorenzo Bianconi * This helper is provided as an easy way to load data from a 51443f364222SLorenzo Bianconi * xdp buffer. It can be used to load *len* bytes from *offset* from 51453f364222SLorenzo Bianconi * the frame associated to *xdp_md*, into the buffer pointed by 51463f364222SLorenzo Bianconi * *buf*. 51473f364222SLorenzo Bianconi * Return 51483f364222SLorenzo Bianconi * 0 on success, or a negative error in case of failure. 51493f364222SLorenzo Bianconi * 51503f364222SLorenzo Bianconi * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 51513f364222SLorenzo Bianconi * Description 51523f364222SLorenzo Bianconi * Store *len* bytes from buffer *buf* into the frame 51533f364222SLorenzo Bianconi * associated to *xdp_md*, at *offset*. 51543f364222SLorenzo Bianconi * Return 51553f364222SLorenzo Bianconi * 0 on success, or a negative error in case of failure. 5156376040e4SKenny Yu * 5157376040e4SKenny Yu * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5158376040e4SKenny Yu * Description 5159376040e4SKenny Yu * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5160376040e4SKenny Yu * address space, and stores the data in *dst*. *flags* is not 5161376040e4SKenny Yu * used yet and is provided for future extensibility. This helper 5162376040e4SKenny Yu * can only be used by sleepable programs. 5163376040e4SKenny Yu * Return 5164376040e4SKenny Yu * 0 on success, or a negative error in case of failure. On error 5165376040e4SKenny Yu * *dst* buffer is zeroed out. 51668d21ec0eSMartin KaFai Lau * 51679bb984f2SMartin KaFai Lau * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 51688d21ec0eSMartin KaFai Lau * Description 51699bb984f2SMartin KaFai Lau * Change the __sk_buff->tstamp_type to *tstamp_type* 51709bb984f2SMartin KaFai Lau * and set *tstamp* to the __sk_buff->tstamp together. 51718d21ec0eSMartin KaFai Lau * 51729bb984f2SMartin KaFai Lau * If there is no need to change the __sk_buff->tstamp_type, 51739bb984f2SMartin KaFai Lau * the tstamp value can be directly written to __sk_buff->tstamp 51748d21ec0eSMartin KaFai Lau * instead. 51758d21ec0eSMartin KaFai Lau * 51769bb984f2SMartin KaFai Lau * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 51779bb984f2SMartin KaFai Lau * will be kept during bpf_redirect_*(). A non zero 51789bb984f2SMartin KaFai Lau * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 51799bb984f2SMartin KaFai Lau * *tstamp_type*. 51809bb984f2SMartin KaFai Lau * 51819bb984f2SMartin KaFai Lau * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 51829bb984f2SMartin KaFai Lau * with a zero *tstamp*. 51838d21ec0eSMartin KaFai Lau * 51848d21ec0eSMartin KaFai Lau * Only IPv4 and IPv6 skb->protocol are supported. 51858d21ec0eSMartin KaFai Lau * 51868d21ec0eSMartin KaFai Lau * This function is most useful when it needs to set a 51878d21ec0eSMartin KaFai Lau * mono delivery time to __sk_buff->tstamp and then 51888d21ec0eSMartin KaFai Lau * bpf_redirect_*() to the egress of an iface. For example, 51898d21ec0eSMartin KaFai Lau * changing the (rcv) timestamp in __sk_buff->tstamp at 51908d21ec0eSMartin KaFai Lau * ingress to a mono delivery time and then bpf_redirect_*() 51918d21ec0eSMartin KaFai Lau * to sch_fq@phy-dev. 51928d21ec0eSMartin KaFai Lau * Return 51938d21ec0eSMartin KaFai Lau * 0 on success. 51948d21ec0eSMartin KaFai Lau * **-EINVAL** for invalid input 51959bb984f2SMartin KaFai Lau * **-EOPNOTSUPP** for unsupported protocol 5196174b1694SRoberto Sassu * 5197174b1694SRoberto Sassu * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5198174b1694SRoberto Sassu * Description 5199174b1694SRoberto Sassu * Returns a calculated IMA hash of the *file*. 5200174b1694SRoberto Sassu * If the hash is larger than *size*, then only *size* 5201174b1694SRoberto Sassu * bytes will be copied to *dst* 5202174b1694SRoberto Sassu * Return 5203174b1694SRoberto Sassu * The **hash_algo** is returned on success, 5204174b1694SRoberto Sassu * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if 5205174b1694SRoberto Sassu * invalid arguments are passed. 5206c0a5a21cSKumar Kartikeya Dwivedi * 5207c0a5a21cSKumar Kartikeya Dwivedi * void *bpf_kptr_xchg(void *map_value, void *ptr) 5208c0a5a21cSKumar Kartikeya Dwivedi * Description 5209c0a5a21cSKumar Kartikeya Dwivedi * Exchange kptr at pointer *map_value* with *ptr*, and return the 5210c0a5a21cSKumar Kartikeya Dwivedi * old value. *ptr* can be NULL, otherwise it must be a referenced 5211c0a5a21cSKumar Kartikeya Dwivedi * pointer which will be released when this helper is called. 5212c0a5a21cSKumar Kartikeya Dwivedi * Return 5213c0a5a21cSKumar Kartikeya Dwivedi * The old value of kptr (which can be NULL). The returned pointer 5214c0a5a21cSKumar Kartikeya Dwivedi * if not NULL, is a reference which must be released using its 5215c0a5a21cSKumar Kartikeya Dwivedi * corresponding release function, or moved into a BPF map before 5216c0a5a21cSKumar Kartikeya Dwivedi * program exit. 521707343110SFeng Zhou * 521807343110SFeng Zhou * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu) 521907343110SFeng Zhou * Description 522007343110SFeng Zhou * Perform a lookup in *percpu map* for an entry associated to 522107343110SFeng Zhou * *key* on *cpu*. 522207343110SFeng Zhou * Return 522307343110SFeng Zhou * Map value associated to *key* on *cpu*, or **NULL** if no entry 522407343110SFeng Zhou * was found or *cpu* is invalid. 52253bc253c2SGeliang Tang * 52263bc253c2SGeliang Tang * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk) 52273bc253c2SGeliang Tang * Description 52283bc253c2SGeliang Tang * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 52293bc253c2SGeliang Tang * Return 52303bc253c2SGeliang Tang * *sk* if casting is valid, or **NULL** otherwise. 5231263ae152SJoanne Koong * 5232263ae152SJoanne Koong * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr) 5233263ae152SJoanne Koong * Description 5234263ae152SJoanne Koong * Get a dynptr to local memory *data*. 5235263ae152SJoanne Koong * 5236263ae152SJoanne Koong * *data* must be a ptr to a map value. 5237263ae152SJoanne Koong * The maximum *size* supported is DYNPTR_MAX_SIZE. 5238263ae152SJoanne Koong * *flags* is currently unused. 5239263ae152SJoanne Koong * Return 5240263ae152SJoanne Koong * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 5241263ae152SJoanne Koong * -EINVAL if flags is not 0. 5242bc34dee6SJoanne Koong * 5243bc34dee6SJoanne Koong * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr) 5244bc34dee6SJoanne Koong * Description 5245bc34dee6SJoanne Koong * Reserve *size* bytes of payload in a ring buffer *ringbuf* 5246bc34dee6SJoanne Koong * through the dynptr interface. *flags* must be 0. 5247bc34dee6SJoanne Koong * 5248bc34dee6SJoanne Koong * Please note that a corresponding bpf_ringbuf_submit_dynptr or 5249bc34dee6SJoanne Koong * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 5250bc34dee6SJoanne Koong * reservation fails. This is enforced by the verifier. 5251bc34dee6SJoanne Koong * Return 5252bc34dee6SJoanne Koong * 0 on success, or a negative error in case of failure. 5253bc34dee6SJoanne Koong * 5254bc34dee6SJoanne Koong * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags) 5255bc34dee6SJoanne Koong * Description 5256bc34dee6SJoanne Koong * Submit reserved ring buffer sample, pointed to by *data*, 5257bc34dee6SJoanne Koong * through the dynptr interface. This is a no-op if the dynptr is 5258bc34dee6SJoanne Koong * invalid/null. 5259bc34dee6SJoanne Koong * 5260bc34dee6SJoanne Koong * For more information on *flags*, please see 5261bc34dee6SJoanne Koong * 'bpf_ringbuf_submit'. 5262bc34dee6SJoanne Koong * Return 5263bc34dee6SJoanne Koong * Nothing. Always succeeds. 5264bc34dee6SJoanne Koong * 5265bc34dee6SJoanne Koong * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags) 5266bc34dee6SJoanne Koong * Description 5267bc34dee6SJoanne Koong * Discard reserved ring buffer sample through the dynptr 5268bc34dee6SJoanne Koong * interface. This is a no-op if the dynptr is invalid/null. 5269bc34dee6SJoanne Koong * 5270bc34dee6SJoanne Koong * For more information on *flags*, please see 5271bc34dee6SJoanne Koong * 'bpf_ringbuf_discard'. 5272bc34dee6SJoanne Koong * Return 5273bc34dee6SJoanne Koong * Nothing. Always succeeds. 527413bbbfbeSJoanne Koong * 5275f8d3da4eSJoanne Koong * long bpf_dynptr_read(void *dst, u32 len, struct bpf_dynptr *src, u32 offset, u64 flags) 527613bbbfbeSJoanne Koong * Description 527713bbbfbeSJoanne Koong * Read *len* bytes from *src* into *dst*, starting from *offset* 527813bbbfbeSJoanne Koong * into *src*. 5279f8d3da4eSJoanne Koong * *flags* is currently unused. 528013bbbfbeSJoanne Koong * Return 528113bbbfbeSJoanne Koong * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5282f8d3da4eSJoanne Koong * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 5283f8d3da4eSJoanne Koong * *flags* is not 0. 528413bbbfbeSJoanne Koong * 5285f8d3da4eSJoanne Koong * long bpf_dynptr_write(struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags) 528613bbbfbeSJoanne Koong * Description 528713bbbfbeSJoanne Koong * Write *len* bytes from *src* into *dst*, starting from *offset* 528813bbbfbeSJoanne Koong * into *dst*. 5289f8d3da4eSJoanne Koong * *flags* is currently unused. 529013bbbfbeSJoanne Koong * Return 529113bbbfbeSJoanne Koong * 0 on success, -E2BIG if *offset* + *len* exceeds the length 529213bbbfbeSJoanne Koong * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 5293f8d3da4eSJoanne Koong * is a read-only dynptr or if *flags* is not 0. 529434d4ef57SJoanne Koong * 529534d4ef57SJoanne Koong * void *bpf_dynptr_data(struct bpf_dynptr *ptr, u32 offset, u32 len) 529634d4ef57SJoanne Koong * Description 529734d4ef57SJoanne Koong * Get a pointer to the underlying dynptr data. 529834d4ef57SJoanne Koong * 529934d4ef57SJoanne Koong * *len* must be a statically known value. The returned data slice 530034d4ef57SJoanne Koong * is invalidated whenever the dynptr is invalidated. 530134d4ef57SJoanne Koong * Return 530234d4ef57SJoanne Koong * Pointer to the underlying dynptr data, NULL if the dynptr is 530334d4ef57SJoanne Koong * read-only, if the dynptr is invalid, or if the offset and length 530434d4ef57SJoanne Koong * is out of bounds. 530533bf9885SMaxim Mikityanskiy * 530633bf9885SMaxim Mikityanskiy * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len) 530733bf9885SMaxim Mikityanskiy * Description 530833bf9885SMaxim Mikityanskiy * Try to issue a SYN cookie for the packet with corresponding 530933bf9885SMaxim Mikityanskiy * IPv4/TCP headers, *iph* and *th*, without depending on a 531033bf9885SMaxim Mikityanskiy * listening socket. 531133bf9885SMaxim Mikityanskiy * 531233bf9885SMaxim Mikityanskiy * *iph* points to the IPv4 header. 531333bf9885SMaxim Mikityanskiy * 531433bf9885SMaxim Mikityanskiy * *th* points to the start of the TCP header, while *th_len* 531533bf9885SMaxim Mikityanskiy * contains the length of the TCP header (at least 531633bf9885SMaxim Mikityanskiy * **sizeof**\ (**struct tcphdr**)). 531733bf9885SMaxim Mikityanskiy * Return 531833bf9885SMaxim Mikityanskiy * On success, lower 32 bits hold the generated SYN cookie in 531933bf9885SMaxim Mikityanskiy * followed by 16 bits which hold the MSS value for that cookie, 532033bf9885SMaxim Mikityanskiy * and the top 16 bits are unused. 532133bf9885SMaxim Mikityanskiy * 532233bf9885SMaxim Mikityanskiy * On failure, the returned value is one of the following: 532333bf9885SMaxim Mikityanskiy * 532433bf9885SMaxim Mikityanskiy * **-EINVAL** if *th_len* is invalid. 532533bf9885SMaxim Mikityanskiy * 532633bf9885SMaxim Mikityanskiy * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len) 532733bf9885SMaxim Mikityanskiy * Description 532833bf9885SMaxim Mikityanskiy * Try to issue a SYN cookie for the packet with corresponding 532933bf9885SMaxim Mikityanskiy * IPv6/TCP headers, *iph* and *th*, without depending on a 533033bf9885SMaxim Mikityanskiy * listening socket. 533133bf9885SMaxim Mikityanskiy * 533233bf9885SMaxim Mikityanskiy * *iph* points to the IPv6 header. 533333bf9885SMaxim Mikityanskiy * 533433bf9885SMaxim Mikityanskiy * *th* points to the start of the TCP header, while *th_len* 533533bf9885SMaxim Mikityanskiy * contains the length of the TCP header (at least 533633bf9885SMaxim Mikityanskiy * **sizeof**\ (**struct tcphdr**)). 533733bf9885SMaxim Mikityanskiy * Return 533833bf9885SMaxim Mikityanskiy * On success, lower 32 bits hold the generated SYN cookie in 533933bf9885SMaxim Mikityanskiy * followed by 16 bits which hold the MSS value for that cookie, 534033bf9885SMaxim Mikityanskiy * and the top 16 bits are unused. 534133bf9885SMaxim Mikityanskiy * 534233bf9885SMaxim Mikityanskiy * On failure, the returned value is one of the following: 534333bf9885SMaxim Mikityanskiy * 534433bf9885SMaxim Mikityanskiy * **-EINVAL** if *th_len* is invalid. 534533bf9885SMaxim Mikityanskiy * 534633bf9885SMaxim Mikityanskiy * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 534733bf9885SMaxim Mikityanskiy * 534833bf9885SMaxim Mikityanskiy * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th) 534933bf9885SMaxim Mikityanskiy * Description 535033bf9885SMaxim Mikityanskiy * Check whether *iph* and *th* contain a valid SYN cookie ACK 535133bf9885SMaxim Mikityanskiy * without depending on a listening socket. 535233bf9885SMaxim Mikityanskiy * 535333bf9885SMaxim Mikityanskiy * *iph* points to the IPv4 header. 535433bf9885SMaxim Mikityanskiy * 535533bf9885SMaxim Mikityanskiy * *th* points to the TCP header. 535633bf9885SMaxim Mikityanskiy * Return 535733bf9885SMaxim Mikityanskiy * 0 if *iph* and *th* are a valid SYN cookie ACK. 535833bf9885SMaxim Mikityanskiy * 535933bf9885SMaxim Mikityanskiy * On failure, the returned value is one of the following: 536033bf9885SMaxim Mikityanskiy * 536133bf9885SMaxim Mikityanskiy * **-EACCES** if the SYN cookie is not valid. 536233bf9885SMaxim Mikityanskiy * 536333bf9885SMaxim Mikityanskiy * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th) 536433bf9885SMaxim Mikityanskiy * Description 536533bf9885SMaxim Mikityanskiy * Check whether *iph* and *th* contain a valid SYN cookie ACK 536633bf9885SMaxim Mikityanskiy * without depending on a listening socket. 536733bf9885SMaxim Mikityanskiy * 536833bf9885SMaxim Mikityanskiy * *iph* points to the IPv6 header. 536933bf9885SMaxim Mikityanskiy * 537033bf9885SMaxim Mikityanskiy * *th* points to the TCP header. 537133bf9885SMaxim Mikityanskiy * Return 537233bf9885SMaxim Mikityanskiy * 0 if *iph* and *th* are a valid SYN cookie ACK. 537333bf9885SMaxim Mikityanskiy * 537433bf9885SMaxim Mikityanskiy * On failure, the returned value is one of the following: 537533bf9885SMaxim Mikityanskiy * 537633bf9885SMaxim Mikityanskiy * **-EACCES** if the SYN cookie is not valid. 537733bf9885SMaxim Mikityanskiy * 537833bf9885SMaxim Mikityanskiy * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5379c8996c98SJesper Dangaard Brouer * 5380c8996c98SJesper Dangaard Brouer * u64 bpf_ktime_get_tai_ns(void) 5381c8996c98SJesper Dangaard Brouer * Description 5382c8996c98SJesper Dangaard Brouer * A nonsettable system-wide clock derived from wall-clock time but 5383c8996c98SJesper Dangaard Brouer * ignoring leap seconds. This clock does not experience 5384c8996c98SJesper Dangaard Brouer * discontinuities and backwards jumps caused by NTP inserting leap 5385c8996c98SJesper Dangaard Brouer * seconds as CLOCK_REALTIME does. 5386c8996c98SJesper Dangaard Brouer * 5387c8996c98SJesper Dangaard Brouer * See: **clock_gettime**\ (**CLOCK_TAI**) 5388c8996c98SJesper Dangaard Brouer * Return 5389c8996c98SJesper Dangaard Brouer * Current *ktime*. 5390c8996c98SJesper Dangaard Brouer * 53917a4b28c6SDaniel Borkmann */ 5392ebb676daSThomas Graf #define __BPF_FUNC_MAPPER(FN) \ 5393ebb676daSThomas Graf FN(unspec), \ 5394ebb676daSThomas Graf FN(map_lookup_elem), \ 5395ebb676daSThomas Graf FN(map_update_elem), \ 5396ebb676daSThomas Graf FN(map_delete_elem), \ 5397ebb676daSThomas Graf FN(probe_read), \ 5398ebb676daSThomas Graf FN(ktime_get_ns), \ 5399ebb676daSThomas Graf FN(trace_printk), \ 5400ebb676daSThomas Graf FN(get_prandom_u32), \ 5401ebb676daSThomas Graf FN(get_smp_processor_id), \ 5402ebb676daSThomas Graf FN(skb_store_bytes), \ 5403ebb676daSThomas Graf FN(l3_csum_replace), \ 5404ebb676daSThomas Graf FN(l4_csum_replace), \ 5405ebb676daSThomas Graf FN(tail_call), \ 5406ebb676daSThomas Graf FN(clone_redirect), \ 5407ebb676daSThomas Graf FN(get_current_pid_tgid), \ 5408ebb676daSThomas Graf FN(get_current_uid_gid), \ 5409ebb676daSThomas Graf FN(get_current_comm), \ 5410ebb676daSThomas Graf FN(get_cgroup_classid), \ 5411ebb676daSThomas Graf FN(skb_vlan_push), \ 5412ebb676daSThomas Graf FN(skb_vlan_pop), \ 5413ebb676daSThomas Graf FN(skb_get_tunnel_key), \ 5414ebb676daSThomas Graf FN(skb_set_tunnel_key), \ 5415ebb676daSThomas Graf FN(perf_event_read), \ 5416ebb676daSThomas Graf FN(redirect), \ 5417ebb676daSThomas Graf FN(get_route_realm), \ 5418ebb676daSThomas Graf FN(perf_event_output), \ 5419ebb676daSThomas Graf FN(skb_load_bytes), \ 5420ebb676daSThomas Graf FN(get_stackid), \ 5421ebb676daSThomas Graf FN(csum_diff), \ 5422ebb676daSThomas Graf FN(skb_get_tunnel_opt), \ 5423ebb676daSThomas Graf FN(skb_set_tunnel_opt), \ 5424ebb676daSThomas Graf FN(skb_change_proto), \ 5425ebb676daSThomas Graf FN(skb_change_type), \ 5426ebb676daSThomas Graf FN(skb_under_cgroup), \ 5427ebb676daSThomas Graf FN(get_hash_recalc), \ 5428ebb676daSThomas Graf FN(get_current_task), \ 5429ebb676daSThomas Graf FN(probe_write_user), \ 5430ebb676daSThomas Graf FN(current_task_under_cgroup), \ 5431ebb676daSThomas Graf FN(skb_change_tail), \ 5432ebb676daSThomas Graf FN(skb_pull_data), \ 5433ebb676daSThomas Graf FN(csum_update), \ 5434ebb676daSThomas Graf FN(set_hash_invalid), \ 54353a0af8fdSThomas Graf FN(get_numa_node_id), \ 543617bedab2SMartin KaFai Lau FN(skb_change_head), \ 5437a5e8c070SGianluca Borello FN(xdp_adjust_head), \ 543891b8270fSChenbo Feng FN(probe_read_str), \ 54396acc5c29SChenbo Feng FN(get_socket_cookie), \ 5440ded092cdSDaniel Borkmann FN(get_socket_uid), \ 54418c4b4c7eSLawrence Brakmo FN(set_hash), \ 54422be7e212SDaniel Borkmann FN(setsockopt), \ 544397f91a7cSJohn Fastabend FN(skb_adjust_room), \ 5444174a79ffSJohn Fastabend FN(redirect_map), \ 5445174a79ffSJohn Fastabend FN(sk_redirect_map), \ 5446174a79ffSJohn Fastabend FN(sock_map_update), \ 5447908432caSYonghong Song FN(xdp_adjust_meta), \ 54484bebdc7aSYonghong Song FN(perf_event_read_value), \ 5449cd86d1fdSLawrence Brakmo FN(perf_prog_read_value), \ 54509802d865SJosef Bacik FN(getsockopt), \ 5451b13d8807SLawrence Brakmo FN(override_return), \ 54524f738adbSJohn Fastabend FN(sock_ops_cb_flags_set), \ 54532a100317SJohn Fastabend FN(msg_redirect_map), \ 545491843d54SJohn Fastabend FN(msg_apply_bytes), \ 5455015632bbSJohn Fastabend FN(msg_cork_bytes), \ 5456d74bad4eSAndrey Ignatov FN(msg_pull_data), \ 5457b32cc5b9SNikita V. Shirokov FN(bind), \ 545812bed760SEyal Birger FN(xdp_adjust_tail), \ 5459c195651eSYonghong Song FN(skb_get_xfrm_state), \ 54604e1ec56cSDaniel Borkmann FN(get_stack), \ 546187f5fc7eSDavid Ahern FN(skb_load_bytes_relative), \ 546281110384SJohn Fastabend FN(fib_lookup), \ 546381110384SJohn Fastabend FN(sock_hash_update), \ 546481110384SJohn Fastabend FN(msg_redirect_hash), \ 5465fe94cc29SMathieu Xhonneux FN(sk_redirect_hash), \ 5466fe94cc29SMathieu Xhonneux FN(lwt_push_encap), \ 5467fe94cc29SMathieu Xhonneux FN(lwt_seg6_store_bytes), \ 5468fe94cc29SMathieu Xhonneux FN(lwt_seg6_adjust_srh), \ 5469f4364dcfSSean Young FN(lwt_seg6_action), \ 5470f4364dcfSSean Young FN(rc_repeat), \ 5471cb20b08eSDaniel Borkmann FN(rc_keydown), \ 5472bf6fa2c8SYonghong Song FN(skb_cgroup_id), \ 5473cd339431SRoman Gushchin FN(get_current_cgroup_id), \ 54742dbb9b9eSMartin KaFai Lau FN(get_local_storage), \ 547577236281SAndrey Ignatov FN(sk_select_reuseport), \ 54766acc9b43SJoe Stringer FN(skb_ancestor_cgroup_id), \ 54776acc9b43SJoe Stringer FN(sk_lookup_tcp), \ 54786acc9b43SJoe Stringer FN(sk_lookup_udp), \ 5479f1a2e44aSMauricio Vasquez B FN(sk_release), \ 5480f1a2e44aSMauricio Vasquez B FN(map_push_elem), \ 5481f1a2e44aSMauricio Vasquez B FN(map_pop_elem), \ 54826fff607eSJohn Fastabend FN(map_peek_elem), \ 54837246d8edSJohn Fastabend FN(msg_push_data), \ 548401d3240aSSean Young FN(msg_pop_data), \ 5485d83525caSAlexei Starovoitov FN(rc_pointer_rel), \ 5486d83525caSAlexei Starovoitov FN(spin_lock), \ 548746f8bc92SMartin KaFai Lau FN(spin_unlock), \ 5488655a51e5SMartin KaFai Lau FN(sk_fullsock), \ 5489f7c917baSbrakmo FN(tcp_sock), \ 5490dbafd7ddSMartin KaFai Lau FN(skb_ecn_set_ce), \ 5491edbf8c01SLorenz Bauer FN(get_listener_sock), \ 549239904084SLorenz Bauer FN(skc_lookup_tcp), \ 5493808649fbSAndrey Ignatov FN(tcp_check_syncookie), \ 54941d11b301SAndrey Ignatov FN(sysctl_get_name), \ 54954e63acdfSAndrey Ignatov FN(sysctl_get_current_value), \ 54964e63acdfSAndrey Ignatov FN(sysctl_get_new_value), \ 5497d7a4cb9bSAndrey Ignatov FN(sysctl_set_new_value), \ 5498d7a4cb9bSAndrey Ignatov FN(strtol), \ 54996ac99e8fSMartin KaFai Lau FN(strtoul), \ 55006ac99e8fSMartin KaFai Lau FN(sk_storage_get), \ 55018b401f9eSYonghong Song FN(sk_storage_delete), \ 550270d66244SPetar Penkov FN(send_signal), \ 5503a7658e1aSAlexei Starovoitov FN(tcp_gen_syncookie), \ 55046ae08ae3SDaniel Borkmann FN(skb_output), \ 55056ae08ae3SDaniel Borkmann FN(probe_read_user), \ 55066ae08ae3SDaniel Borkmann FN(probe_read_kernel), \ 55076ae08ae3SDaniel Borkmann FN(probe_read_user_str), \ 5508206057feSMartin KaFai Lau FN(probe_read_kernel_str), \ 55098482941fSYonghong Song FN(tcp_send_ack), \ 55105576b991SMartin KaFai Lau FN(send_signal_thread), \ 5511fff7b643SDaniel Xu FN(jiffies64), \ 5512b4490c5cSCarlos Neira FN(read_branch_records), \ 5513d831ee84SEelco Chaudron FN(get_ns_current_pid_tgid), \ 5514f318903cSDaniel Borkmann FN(xdp_output), \ 55150f09abd1SDaniel Borkmann FN(get_netns_cookie), \ 5516cf7fbe66SJoe Stringer FN(get_current_ancestor_cgroup_id), \ 551771d19214SMaciej Żenczykowski FN(sk_assign), \ 5518492e639fSYonghong Song FN(ktime_get_boot_ns), \ 5519492e639fSYonghong Song FN(seq_printf), \ 5520f307fa2cSAndrey Ignatov FN(seq_write), \ 5521f307fa2cSAndrey Ignatov FN(sk_cgroup_id), \ 5522457f4436SAndrii Nakryiko FN(sk_ancestor_cgroup_id), \ 5523457f4436SAndrii Nakryiko FN(ringbuf_output), \ 5524457f4436SAndrii Nakryiko FN(ringbuf_reserve), \ 5525457f4436SAndrii Nakryiko FN(ringbuf_submit), \ 5526457f4436SAndrii Nakryiko FN(ringbuf_discard), \ 55277cdec54fSDaniel Borkmann FN(ringbuf_query), \ 5528af7ec138SYonghong Song FN(csum_level), \ 5529478cfbdfSYonghong Song FN(skc_to_tcp6_sock), \ 5530478cfbdfSYonghong Song FN(skc_to_tcp_sock), \ 5531478cfbdfSYonghong Song FN(skc_to_tcp_timewait_sock), \ 55320d4fad3eSYonghong Song FN(skc_to_tcp_request_sock), \ 5533fa28dcb8SSong Liu FN(skc_to_udp6_sock), \ 5534fa28dcb8SSong Liu FN(get_task_stack), \ 55350813a841SMartin KaFai Lau FN(load_hdr_opt), \ 55360813a841SMartin KaFai Lau FN(store_hdr_opt), \ 55378ea63684SKP Singh FN(reserve_hdr_opt), \ 55388ea63684SKP Singh FN(inode_storage_get), \ 55398ea63684SKP Singh FN(inode_storage_delete), \ 55406e22ab9dSJiri Olsa FN(d_path), \ 554107be4c4aSAlexei Starovoitov FN(copy_from_user), \ 5542c4d0bfb4SAlan Maguire FN(snprintf_btf), \ 5543eb411377SAlan Maguire FN(seq_printf_btf), \ 5544b426ce83SDaniel Borkmann FN(skb_cgroup_classid), \ 5545b4ab3141SDaniel Borkmann FN(redirect_neigh), \ 5546b7906b70SAndrii Nakryiko FN(per_cpu_ptr), \ 5547b7906b70SAndrii Nakryiko FN(this_cpu_ptr), \ 55489aa1206eSDaniel Borkmann FN(redirect_peer), \ 55494cf1bc1fSKP Singh FN(task_storage_get), \ 55504cf1bc1fSKP Singh FN(task_storage_delete), \ 55513ca1032aSKP Singh FN(get_current_task_btf), \ 55523f6719c7SKP Singh FN(bprm_opts_set), \ 5553d0551261SDmitrii Banshchikov FN(ktime_get_coarse_ns), \ 555427672f0dSKP Singh FN(ima_inode_hash), \ 55554f19cab7SFlorent Revest FN(sock_from_file), \ 555634b2021cSJesper Dangaard Brouer FN(check_mtu), \ 555769c087baSYonghong Song FN(for_each_map_elem), \ 55587b15523aSFlorent Revest FN(snprintf), \ 555979a7f8bdSAlexei Starovoitov FN(sys_bpf), \ 55603d78417bSAlexei Starovoitov FN(btf_find_by_name_kind), \ 55613abea089SAlexei Starovoitov FN(sys_close), \ 5562b00628b1SAlexei Starovoitov FN(timer_init), \ 5563b00628b1SAlexei Starovoitov FN(timer_set_callback), \ 5564b00628b1SAlexei Starovoitov FN(timer_start), \ 5565b00628b1SAlexei Starovoitov FN(timer_cancel), \ 55669b99edcaSJiri Olsa FN(get_func_ip), \ 55677adfc6c9SAndrii Nakryiko FN(get_attach_cookie), \ 5568dd6e10fbSDaniel Xu FN(task_pt_regs), \ 5569856c02dbSSong Liu FN(get_branch_snapshot), \ 557010aceb62SDave Marchevsky FN(trace_vprintk), \ 55719eeb3aa3SHengqi Chen FN(skc_to_unix_sock), \ 5572d6aef08aSKumar Kartikeya Dwivedi FN(kallsyms_lookup_name), \ 55737c7e3d31SSong Liu FN(find_vma), \ 5574e6f2dd0fSJoanne Koong FN(loop), \ 5575c5fb1993SHou Tao FN(strncmp), \ 5576f92c1e18SJiri Olsa FN(get_func_arg), \ 5577f92c1e18SJiri Olsa FN(get_func_ret), \ 5578f92c1e18SJiri Olsa FN(get_func_arg_cnt), \ 5579b44123b4SYiFei Zhu FN(get_retval), \ 5580b44123b4SYiFei Zhu FN(set_retval), \ 55810165cc81SLorenzo Bianconi FN(xdp_get_buff_len), \ 55823f364222SLorenzo Bianconi FN(xdp_load_bytes), \ 55833f364222SLorenzo Bianconi FN(xdp_store_bytes), \ 5584376040e4SKenny Yu FN(copy_from_user_task), \ 55859bb984f2SMartin KaFai Lau FN(skb_set_tstamp), \ 5586174b1694SRoberto Sassu FN(ima_file_hash), \ 5587c0a5a21cSKumar Kartikeya Dwivedi FN(kptr_xchg), \ 558807343110SFeng Zhou FN(map_lookup_percpu_elem), \ 55893bc253c2SGeliang Tang FN(skc_to_mptcp_sock), \ 5590263ae152SJoanne Koong FN(dynptr_from_mem), \ 5591bc34dee6SJoanne Koong FN(ringbuf_reserve_dynptr), \ 5592bc34dee6SJoanne Koong FN(ringbuf_submit_dynptr), \ 5593bc34dee6SJoanne Koong FN(ringbuf_discard_dynptr), \ 559413bbbfbeSJoanne Koong FN(dynptr_read), \ 559513bbbfbeSJoanne Koong FN(dynptr_write), \ 559634d4ef57SJoanne Koong FN(dynptr_data), \ 559733bf9885SMaxim Mikityanskiy FN(tcp_raw_gen_syncookie_ipv4), \ 559833bf9885SMaxim Mikityanskiy FN(tcp_raw_gen_syncookie_ipv6), \ 559933bf9885SMaxim Mikityanskiy FN(tcp_raw_check_syncookie_ipv4), \ 560033bf9885SMaxim Mikityanskiy FN(tcp_raw_check_syncookie_ipv6), \ 5601c8996c98SJesper Dangaard Brouer FN(ktime_get_tai_ns), \ 5602fa28dcb8SSong Liu /* */ 56037a4b28c6SDaniel Borkmann 5604ebb676daSThomas Graf /* integer value in 'imm' field of BPF_CALL instruction selects which helper 5605ebb676daSThomas Graf * function eBPF program intends to call 56062d0e30c3SDaniel Borkmann */ 5607ebb676daSThomas Graf #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 5608ebb676daSThomas Graf enum bpf_func_id { 5609ebb676daSThomas Graf __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 561009756af4SAlexei Starovoitov __BPF_FUNC_MAX_ID, 561109756af4SAlexei Starovoitov }; 5612ebb676daSThomas Graf #undef __BPF_ENUM_FN 561309756af4SAlexei Starovoitov 5614781c53bcSDaniel Borkmann /* All flags used by eBPF helper functions, placed here. */ 5615781c53bcSDaniel Borkmann 5616781c53bcSDaniel Borkmann /* BPF_FUNC_skb_store_bytes flags. */ 56171aae4bddSAndrii Nakryiko enum { 56181aae4bddSAndrii Nakryiko BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 56191aae4bddSAndrii Nakryiko BPF_F_INVALIDATE_HASH = (1ULL << 1), 56201aae4bddSAndrii Nakryiko }; 5621781c53bcSDaniel Borkmann 5622781c53bcSDaniel Borkmann /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 5623781c53bcSDaniel Borkmann * First 4 bits are for passing the header field size. 5624781c53bcSDaniel Borkmann */ 56251aae4bddSAndrii Nakryiko enum { 56261aae4bddSAndrii Nakryiko BPF_F_HDR_FIELD_MASK = 0xfULL, 56271aae4bddSAndrii Nakryiko }; 5628781c53bcSDaniel Borkmann 5629781c53bcSDaniel Borkmann /* BPF_FUNC_l4_csum_replace flags. */ 56301aae4bddSAndrii Nakryiko enum { 56311aae4bddSAndrii Nakryiko BPF_F_PSEUDO_HDR = (1ULL << 4), 56321aae4bddSAndrii Nakryiko BPF_F_MARK_MANGLED_0 = (1ULL << 5), 56331aae4bddSAndrii Nakryiko BPF_F_MARK_ENFORCE = (1ULL << 6), 56341aae4bddSAndrii Nakryiko }; 5635781c53bcSDaniel Borkmann 5636781c53bcSDaniel Borkmann /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 56371aae4bddSAndrii Nakryiko enum { 56381aae4bddSAndrii Nakryiko BPF_F_INGRESS = (1ULL << 0), 56391aae4bddSAndrii Nakryiko }; 5640781c53bcSDaniel Borkmann 5641c6c33454SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 56421aae4bddSAndrii Nakryiko enum { 56431aae4bddSAndrii Nakryiko BPF_F_TUNINFO_IPV6 = (1ULL << 0), 56441aae4bddSAndrii Nakryiko }; 5645c6c33454SDaniel Borkmann 5646c195651eSYonghong Song /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 56471aae4bddSAndrii Nakryiko enum { 56481aae4bddSAndrii Nakryiko BPF_F_SKIP_FIELD_MASK = 0xffULL, 56491aae4bddSAndrii Nakryiko BPF_F_USER_STACK = (1ULL << 8), 5650c195651eSYonghong Song /* flags used by BPF_FUNC_get_stackid only. */ 56511aae4bddSAndrii Nakryiko BPF_F_FAST_STACK_CMP = (1ULL << 9), 56521aae4bddSAndrii Nakryiko BPF_F_REUSE_STACKID = (1ULL << 10), 5653c195651eSYonghong Song /* flags used by BPF_FUNC_get_stack only. */ 56541aae4bddSAndrii Nakryiko BPF_F_USER_BUILD_ID = (1ULL << 11), 56551aae4bddSAndrii Nakryiko }; 5656d5a3b1f6SAlexei Starovoitov 56572da897e5SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key flags. */ 56581aae4bddSAndrii Nakryiko enum { 56591aae4bddSAndrii Nakryiko BPF_F_ZERO_CSUM_TX = (1ULL << 1), 56601aae4bddSAndrii Nakryiko BPF_F_DONT_FRAGMENT = (1ULL << 2), 56611aae4bddSAndrii Nakryiko BPF_F_SEQ_NUMBER = (1ULL << 3), 56621aae4bddSAndrii Nakryiko }; 56632da897e5SDaniel Borkmann 566444c51472SShmulik Ladkani /* BPF_FUNC_skb_get_tunnel_key flags. */ 566544c51472SShmulik Ladkani enum { 566644c51472SShmulik Ladkani BPF_F_TUNINFO_FLAGS = (1ULL << 4), 566744c51472SShmulik Ladkani }; 566844c51472SShmulik Ladkani 5669908432caSYonghong Song /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 5670908432caSYonghong Song * BPF_FUNC_perf_event_read_value flags. 5671908432caSYonghong Song */ 56721aae4bddSAndrii Nakryiko enum { 56731aae4bddSAndrii Nakryiko BPF_F_INDEX_MASK = 0xffffffffULL, 56741aae4bddSAndrii Nakryiko BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 5675555c8a86SDaniel Borkmann /* BPF_FUNC_perf_event_output for sk_buff input context. */ 56761aae4bddSAndrii Nakryiko BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 56771aae4bddSAndrii Nakryiko }; 56781e33759cSDaniel Borkmann 5679f71c6143SJoe Stringer /* Current network namespace */ 56801aae4bddSAndrii Nakryiko enum { 56811aae4bddSAndrii Nakryiko BPF_F_CURRENT_NETNS = (-1L), 56821aae4bddSAndrii Nakryiko }; 5683f71c6143SJoe Stringer 56847cdec54fSDaniel Borkmann /* BPF_FUNC_csum_level level values. */ 56857cdec54fSDaniel Borkmann enum { 56867cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_QUERY, 56877cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_INC, 56887cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_DEC, 56897cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_RESET, 56907cdec54fSDaniel Borkmann }; 56917cdec54fSDaniel Borkmann 56922278f6ccSWillem de Bruijn /* BPF_FUNC_skb_adjust_room flags. */ 56931aae4bddSAndrii Nakryiko enum { 56941aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 56951aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 56961aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 56971aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 56981aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 5699836e66c2SDaniel Borkmann BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 5700d01b59c9SXuesen Huang BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 57011aae4bddSAndrii Nakryiko }; 57022278f6ccSWillem de Bruijn 57031aae4bddSAndrii Nakryiko enum { 57041aae4bddSAndrii Nakryiko BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 57051aae4bddSAndrii Nakryiko BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 57061aae4bddSAndrii Nakryiko }; 570758dfc900SAlan Maguire 570858dfc900SAlan Maguire #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 570958dfc900SAlan Maguire BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 571058dfc900SAlan Maguire << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 5711868d5235SWillem de Bruijn 5712808649fbSAndrey Ignatov /* BPF_FUNC_sysctl_get_name flags. */ 57131aae4bddSAndrii Nakryiko enum { 57141aae4bddSAndrii Nakryiko BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 57151aae4bddSAndrii Nakryiko }; 5716808649fbSAndrey Ignatov 5717f836a56eSKP Singh /* BPF_FUNC_<kernel_obj>_storage_get flags */ 57181aae4bddSAndrii Nakryiko enum { 5719f836a56eSKP Singh BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 5720f836a56eSKP Singh /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 5721f836a56eSKP Singh * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 5722f836a56eSKP Singh */ 5723f836a56eSKP Singh BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 57241aae4bddSAndrii Nakryiko }; 57256ac99e8fSMartin KaFai Lau 5726fff7b643SDaniel Xu /* BPF_FUNC_read_branch_records flags. */ 57271aae4bddSAndrii Nakryiko enum { 57281aae4bddSAndrii Nakryiko BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 57291aae4bddSAndrii Nakryiko }; 5730fff7b643SDaniel Xu 5731457f4436SAndrii Nakryiko /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 5732457f4436SAndrii Nakryiko * BPF_FUNC_bpf_ringbuf_output flags. 5733457f4436SAndrii Nakryiko */ 5734457f4436SAndrii Nakryiko enum { 5735457f4436SAndrii Nakryiko BPF_RB_NO_WAKEUP = (1ULL << 0), 5736457f4436SAndrii Nakryiko BPF_RB_FORCE_WAKEUP = (1ULL << 1), 5737457f4436SAndrii Nakryiko }; 5738457f4436SAndrii Nakryiko 5739457f4436SAndrii Nakryiko /* BPF_FUNC_bpf_ringbuf_query flags */ 5740457f4436SAndrii Nakryiko enum { 5741457f4436SAndrii Nakryiko BPF_RB_AVAIL_DATA = 0, 5742457f4436SAndrii Nakryiko BPF_RB_RING_SIZE = 1, 5743457f4436SAndrii Nakryiko BPF_RB_CONS_POS = 2, 5744457f4436SAndrii Nakryiko BPF_RB_PROD_POS = 3, 5745457f4436SAndrii Nakryiko }; 5746457f4436SAndrii Nakryiko 5747457f4436SAndrii Nakryiko /* BPF ring buffer constants */ 5748457f4436SAndrii Nakryiko enum { 5749457f4436SAndrii Nakryiko BPF_RINGBUF_BUSY_BIT = (1U << 31), 5750457f4436SAndrii Nakryiko BPF_RINGBUF_DISCARD_BIT = (1U << 30), 5751457f4436SAndrii Nakryiko BPF_RINGBUF_HDR_SZ = 8, 5752457f4436SAndrii Nakryiko }; 5753457f4436SAndrii Nakryiko 5754e9ddbb77SJakub Sitnicki /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 5755e9ddbb77SJakub Sitnicki enum { 5756e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 5757e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 5758e9ddbb77SJakub Sitnicki }; 5759e9ddbb77SJakub Sitnicki 57602be7e212SDaniel Borkmann /* Mode for BPF_FUNC_skb_adjust_room helper. */ 57612be7e212SDaniel Borkmann enum bpf_adj_room_mode { 57622be7e212SDaniel Borkmann BPF_ADJ_ROOM_NET, 576314aa3192SWillem de Bruijn BPF_ADJ_ROOM_MAC, 57642be7e212SDaniel Borkmann }; 57652be7e212SDaniel Borkmann 57664e1ec56cSDaniel Borkmann /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 57674e1ec56cSDaniel Borkmann enum bpf_hdr_start_off { 57684e1ec56cSDaniel Borkmann BPF_HDR_START_MAC, 57694e1ec56cSDaniel Borkmann BPF_HDR_START_NET, 57704e1ec56cSDaniel Borkmann }; 57714e1ec56cSDaniel Borkmann 5772fe94cc29SMathieu Xhonneux /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 5773fe94cc29SMathieu Xhonneux enum bpf_lwt_encap_mode { 5774fe94cc29SMathieu Xhonneux BPF_LWT_ENCAP_SEG6, 57753e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_SEG6_INLINE, 57763e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_IP, 5777fe94cc29SMathieu Xhonneux }; 5778fe94cc29SMathieu Xhonneux 57793f6719c7SKP Singh /* Flags for bpf_bprm_opts_set helper */ 57803f6719c7SKP Singh enum { 57813f6719c7SKP Singh BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 57823f6719c7SKP Singh }; 57833f6719c7SKP Singh 5784e624d4edSHangbin Liu /* Flags for bpf_redirect_map helper */ 5785e624d4edSHangbin Liu enum { 5786e624d4edSHangbin Liu BPF_F_BROADCAST = (1ULL << 3), 5787e624d4edSHangbin Liu BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 5788e624d4edSHangbin Liu }; 5789e624d4edSHangbin Liu 5790b7df9adaSDaniel Borkmann #define __bpf_md_ptr(type, name) \ 5791b7df9adaSDaniel Borkmann union { \ 5792b7df9adaSDaniel Borkmann type name; \ 5793b7df9adaSDaniel Borkmann __u64 :64; \ 5794b7df9adaSDaniel Borkmann } __attribute__((aligned(8))) 5795b7df9adaSDaniel Borkmann 57968d21ec0eSMartin KaFai Lau enum { 57979bb984f2SMartin KaFai Lau BPF_SKB_TSTAMP_UNSPEC, 57989bb984f2SMartin KaFai Lau BPF_SKB_TSTAMP_DELIVERY_MONO, /* tstamp has mono delivery time */ 57999bb984f2SMartin KaFai Lau /* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle, 58009bb984f2SMartin KaFai Lau * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC 58019bb984f2SMartin KaFai Lau * and try to deduce it by ingress, egress or skb->sk->sk_clockid. 58029bb984f2SMartin KaFai Lau */ 58038d21ec0eSMartin KaFai Lau }; 58048d21ec0eSMartin KaFai Lau 58059bac3d6dSAlexei Starovoitov /* user accessible mirror of in-kernel sk_buff. 58069bac3d6dSAlexei Starovoitov * new fields can only be added to the end of this structure 58079bac3d6dSAlexei Starovoitov */ 58089bac3d6dSAlexei Starovoitov struct __sk_buff { 58099bac3d6dSAlexei Starovoitov __u32 len; 58109bac3d6dSAlexei Starovoitov __u32 pkt_type; 58119bac3d6dSAlexei Starovoitov __u32 mark; 58129bac3d6dSAlexei Starovoitov __u32 queue_mapping; 5813c2497395SAlexei Starovoitov __u32 protocol; 5814c2497395SAlexei Starovoitov __u32 vlan_present; 5815c2497395SAlexei Starovoitov __u32 vlan_tci; 581627cd5452SMichal Sekletar __u32 vlan_proto; 5817bcad5718SDaniel Borkmann __u32 priority; 581837e82c2fSAlexei Starovoitov __u32 ingress_ifindex; 581937e82c2fSAlexei Starovoitov __u32 ifindex; 5820d691f9e8SAlexei Starovoitov __u32 tc_index; 5821d691f9e8SAlexei Starovoitov __u32 cb[5]; 5822ba7591d8SDaniel Borkmann __u32 hash; 5823045efa82SDaniel Borkmann __u32 tc_classid; 5824969bf05eSAlexei Starovoitov __u32 data; 5825969bf05eSAlexei Starovoitov __u32 data_end; 5826b1d9fc41SDaniel Borkmann __u32 napi_id; 58278a31db56SJohn Fastabend 5828de8f3a83SDaniel Borkmann /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 58298a31db56SJohn Fastabend __u32 family; 58308a31db56SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 58318a31db56SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 58328a31db56SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 58338a31db56SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 58348a31db56SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 58358a31db56SJohn Fastabend __u32 local_port; /* stored in host byte order */ 5836de8f3a83SDaniel Borkmann /* ... here. */ 5837de8f3a83SDaniel Borkmann 5838de8f3a83SDaniel Borkmann __u32 data_meta; 5839b7df9adaSDaniel Borkmann __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 5840f11216b2SVlad Dumitrescu __u64 tstamp; 5841e3da08d0SPetar Penkov __u32 wire_len; 5842d9ff286aSEric Dumazet __u32 gso_segs; 584346f8bc92SMartin KaFai Lau __bpf_md_ptr(struct bpf_sock *, sk); 5844cf62089bSWillem de Bruijn __u32 gso_size; 58459bb984f2SMartin KaFai Lau __u8 tstamp_type; 58468d21ec0eSMartin KaFai Lau __u32 :24; /* Padding, future use. */ 5847f64c4aceSVadim Fedorenko __u64 hwtstamp; 58489bac3d6dSAlexei Starovoitov }; 58499bac3d6dSAlexei Starovoitov 5850d3aa45ceSAlexei Starovoitov struct bpf_tunnel_key { 5851d3aa45ceSAlexei Starovoitov __u32 tunnel_id; 5852c6c33454SDaniel Borkmann union { 5853d3aa45ceSAlexei Starovoitov __u32 remote_ipv4; 5854c6c33454SDaniel Borkmann __u32 remote_ipv6[4]; 5855c6c33454SDaniel Borkmann }; 5856c6c33454SDaniel Borkmann __u8 tunnel_tos; 5857c6c33454SDaniel Borkmann __u8 tunnel_ttl; 585844c51472SShmulik Ladkani union { 585944c51472SShmulik Ladkani __u16 tunnel_ext; /* compat */ 586044c51472SShmulik Ladkani __be16 tunnel_flags; 586144c51472SShmulik Ladkani }; 58624018ab18SDaniel Borkmann __u32 tunnel_label; 586326101f5aSKaixi Fan union { 586426101f5aSKaixi Fan __u32 local_ipv4; 586526101f5aSKaixi Fan __u32 local_ipv6[4]; 586626101f5aSKaixi Fan }; 5867d3aa45ceSAlexei Starovoitov }; 5868d3aa45ceSAlexei Starovoitov 586912bed760SEyal Birger /* user accessible mirror of in-kernel xfrm_state. 587012bed760SEyal Birger * new fields can only be added to the end of this structure 587112bed760SEyal Birger */ 587212bed760SEyal Birger struct bpf_xfrm_state { 587312bed760SEyal Birger __u32 reqid; 587412bed760SEyal Birger __u32 spi; /* Stored in network byte order */ 587512bed760SEyal Birger __u16 family; 58761fbc2e0cSDaniel Borkmann __u16 ext; /* Padding, future use. */ 587712bed760SEyal Birger union { 587812bed760SEyal Birger __u32 remote_ipv4; /* Stored in network byte order */ 587912bed760SEyal Birger __u32 remote_ipv6[4]; /* Stored in network byte order */ 588012bed760SEyal Birger }; 588112bed760SEyal Birger }; 588212bed760SEyal Birger 58833a0af8fdSThomas Graf /* Generic BPF return codes which all BPF program types may support. 58843a0af8fdSThomas Graf * The values are binary compatible with their TC_ACT_* counter-part to 58853a0af8fdSThomas Graf * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 58863a0af8fdSThomas Graf * programs. 58873a0af8fdSThomas Graf * 58883a0af8fdSThomas Graf * XDP is handled seprately, see XDP_*. 58893a0af8fdSThomas Graf */ 58903a0af8fdSThomas Graf enum bpf_ret_code { 58913a0af8fdSThomas Graf BPF_OK = 0, 58923a0af8fdSThomas Graf /* 1 reserved */ 58933a0af8fdSThomas Graf BPF_DROP = 2, 58943a0af8fdSThomas Graf /* 3-6 reserved */ 58953a0af8fdSThomas Graf BPF_REDIRECT = 7, 58963e0bd37cSPeter Oskolkov /* >127 are reserved for prog type specific return codes. 58973e0bd37cSPeter Oskolkov * 58983e0bd37cSPeter Oskolkov * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 58993e0bd37cSPeter Oskolkov * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 59003e0bd37cSPeter Oskolkov * changed and should be routed based on its new L3 header. 59013e0bd37cSPeter Oskolkov * (This is an L3 redirect, as opposed to L2 redirect 59023e0bd37cSPeter Oskolkov * represented by BPF_REDIRECT above). 59033e0bd37cSPeter Oskolkov */ 59043e0bd37cSPeter Oskolkov BPF_LWT_REROUTE = 128, 590591350fe1SShmulik Ladkani /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR 590691350fe1SShmulik Ladkani * to indicate that no custom dissection was performed, and 590791350fe1SShmulik Ladkani * fallback to standard dissector is requested. 590891350fe1SShmulik Ladkani */ 590991350fe1SShmulik Ladkani BPF_FLOW_DISSECTOR_CONTINUE = 129, 59103a0af8fdSThomas Graf }; 59113a0af8fdSThomas Graf 591261023658SDavid Ahern struct bpf_sock { 591361023658SDavid Ahern __u32 bound_dev_if; 5914aa4c1037SDavid Ahern __u32 family; 5915aa4c1037SDavid Ahern __u32 type; 5916aa4c1037SDavid Ahern __u32 protocol; 5917482dca93SDavid Ahern __u32 mark; 5918482dca93SDavid Ahern __u32 priority; 5919aa65d696SMartin KaFai Lau /* IP address also allows 1 and 2 bytes access */ 5920aa65d696SMartin KaFai Lau __u32 src_ip4; 5921aa65d696SMartin KaFai Lau __u32 src_ip6[4]; 5922aa65d696SMartin KaFai Lau __u32 src_port; /* host byte order */ 59234421a582SJakub Sitnicki __be16 dst_port; /* network byte order */ 59244421a582SJakub Sitnicki __u16 :16; /* zero padding */ 5925aa65d696SMartin KaFai Lau __u32 dst_ip4; 5926aa65d696SMartin KaFai Lau __u32 dst_ip6[4]; 5927aa65d696SMartin KaFai Lau __u32 state; 5928c3c16f2eSAmritha Nambiar __s32 rx_queue_mapping; 592961023658SDavid Ahern }; 593061023658SDavid Ahern 5931655a51e5SMartin KaFai Lau struct bpf_tcp_sock { 5932655a51e5SMartin KaFai Lau __u32 snd_cwnd; /* Sending congestion window */ 5933655a51e5SMartin KaFai Lau __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 5934655a51e5SMartin KaFai Lau __u32 rtt_min; 5935655a51e5SMartin KaFai Lau __u32 snd_ssthresh; /* Slow start size threshold */ 5936655a51e5SMartin KaFai Lau __u32 rcv_nxt; /* What we want to receive next */ 5937655a51e5SMartin KaFai Lau __u32 snd_nxt; /* Next sequence we send */ 5938655a51e5SMartin KaFai Lau __u32 snd_una; /* First byte we want an ack for */ 5939655a51e5SMartin KaFai Lau __u32 mss_cache; /* Cached effective mss, not including SACKS */ 5940655a51e5SMartin KaFai Lau __u32 ecn_flags; /* ECN status bits. */ 5941655a51e5SMartin KaFai Lau __u32 rate_delivered; /* saved rate sample: packets delivered */ 5942655a51e5SMartin KaFai Lau __u32 rate_interval_us; /* saved rate sample: time elapsed */ 5943655a51e5SMartin KaFai Lau __u32 packets_out; /* Packets which are "in flight" */ 5944655a51e5SMartin KaFai Lau __u32 retrans_out; /* Retransmitted packets out */ 5945655a51e5SMartin KaFai Lau __u32 total_retrans; /* Total retransmits for entire connection */ 5946655a51e5SMartin KaFai Lau __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 5947655a51e5SMartin KaFai Lau * total number of segments in. 5948655a51e5SMartin KaFai Lau */ 5949655a51e5SMartin KaFai Lau __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 5950655a51e5SMartin KaFai Lau * total number of data segments in. 5951655a51e5SMartin KaFai Lau */ 5952655a51e5SMartin KaFai Lau __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 5953655a51e5SMartin KaFai Lau * The total number of segments sent. 5954655a51e5SMartin KaFai Lau */ 5955655a51e5SMartin KaFai Lau __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 5956655a51e5SMartin KaFai Lau * total number of data segments sent. 5957655a51e5SMartin KaFai Lau */ 5958655a51e5SMartin KaFai Lau __u32 lost_out; /* Lost packets */ 5959655a51e5SMartin KaFai Lau __u32 sacked_out; /* SACK'd packets */ 5960655a51e5SMartin KaFai Lau __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 5961655a51e5SMartin KaFai Lau * sum(delta(rcv_nxt)), or how many bytes 5962655a51e5SMartin KaFai Lau * were acked. 5963655a51e5SMartin KaFai Lau */ 5964655a51e5SMartin KaFai Lau __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 5965655a51e5SMartin KaFai Lau * sum(delta(snd_una)), or how many bytes 5966655a51e5SMartin KaFai Lau * were acked. 5967655a51e5SMartin KaFai Lau */ 59680357746dSStanislav Fomichev __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 59690357746dSStanislav Fomichev * total number of DSACK blocks received 59700357746dSStanislav Fomichev */ 59710357746dSStanislav Fomichev __u32 delivered; /* Total data packets delivered incl. rexmits */ 59720357746dSStanislav Fomichev __u32 delivered_ce; /* Like the above but only ECE marked packets */ 5973c2cb5e82SStanislav Fomichev __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 5974655a51e5SMartin KaFai Lau }; 5975655a51e5SMartin KaFai Lau 59766acc9b43SJoe Stringer struct bpf_sock_tuple { 59776acc9b43SJoe Stringer union { 59786acc9b43SJoe Stringer struct { 59796acc9b43SJoe Stringer __be32 saddr; 59806acc9b43SJoe Stringer __be32 daddr; 59816acc9b43SJoe Stringer __be16 sport; 59826acc9b43SJoe Stringer __be16 dport; 59836acc9b43SJoe Stringer } ipv4; 59846acc9b43SJoe Stringer struct { 59856acc9b43SJoe Stringer __be32 saddr[4]; 59866acc9b43SJoe Stringer __be32 daddr[4]; 59876acc9b43SJoe Stringer __be16 sport; 59886acc9b43SJoe Stringer __be16 dport; 59896acc9b43SJoe Stringer } ipv6; 59906acc9b43SJoe Stringer }; 59916acc9b43SJoe Stringer }; 59926acc9b43SJoe Stringer 5993fada7fdcSJonathan Lemon struct bpf_xdp_sock { 5994fada7fdcSJonathan Lemon __u32 queue_id; 5995fada7fdcSJonathan Lemon }; 5996fada7fdcSJonathan Lemon 599717bedab2SMartin KaFai Lau #define XDP_PACKET_HEADROOM 256 599817bedab2SMartin KaFai Lau 59996a773a15SBrenden Blanco /* User return codes for XDP prog type. 60006a773a15SBrenden Blanco * A valid XDP program must return one of these defined values. All other 60019beb8bedSDaniel Borkmann * return codes are reserved for future use. Unknown return codes will 60029beb8bedSDaniel Borkmann * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 60036a773a15SBrenden Blanco */ 60046a773a15SBrenden Blanco enum xdp_action { 60056a773a15SBrenden Blanco XDP_ABORTED = 0, 60066a773a15SBrenden Blanco XDP_DROP, 60076a773a15SBrenden Blanco XDP_PASS, 60086ce96ca3SBrenden Blanco XDP_TX, 6009814abfabSJohn Fastabend XDP_REDIRECT, 60106a773a15SBrenden Blanco }; 60116a773a15SBrenden Blanco 60126a773a15SBrenden Blanco /* user accessible metadata for XDP packet hook 60136a773a15SBrenden Blanco * new fields must be added to the end of this structure 60146a773a15SBrenden Blanco */ 60156a773a15SBrenden Blanco struct xdp_md { 60166a773a15SBrenden Blanco __u32 data; 60176a773a15SBrenden Blanco __u32 data_end; 6018de8f3a83SDaniel Borkmann __u32 data_meta; 6019daaf24c6SJesper Dangaard Brouer /* Below access go through struct xdp_rxq_info */ 602002dd3291SJesper Dangaard Brouer __u32 ingress_ifindex; /* rxq->dev->ifindex */ 602102dd3291SJesper Dangaard Brouer __u32 rx_queue_index; /* rxq->queue_index */ 602264b59025SDavid Ahern 602364b59025SDavid Ahern __u32 egress_ifindex; /* txq->dev->ifindex */ 60246a773a15SBrenden Blanco }; 60256a773a15SBrenden Blanco 6026281920b7SJesper Dangaard Brouer /* DEVMAP map-value layout 6027281920b7SJesper Dangaard Brouer * 6028281920b7SJesper Dangaard Brouer * The struct data-layout of map-value is a configuration interface. 6029281920b7SJesper Dangaard Brouer * New members can only be added to the end of this structure. 6030281920b7SJesper Dangaard Brouer */ 6031281920b7SJesper Dangaard Brouer struct bpf_devmap_val { 6032281920b7SJesper Dangaard Brouer __u32 ifindex; /* device index */ 6033281920b7SJesper Dangaard Brouer union { 6034281920b7SJesper Dangaard Brouer int fd; /* prog fd on map write */ 6035281920b7SJesper Dangaard Brouer __u32 id; /* prog id on map read */ 6036281920b7SJesper Dangaard Brouer } bpf_prog; 6037281920b7SJesper Dangaard Brouer }; 6038281920b7SJesper Dangaard Brouer 6039644bfe51SLorenzo Bianconi /* CPUMAP map-value layout 6040644bfe51SLorenzo Bianconi * 6041644bfe51SLorenzo Bianconi * The struct data-layout of map-value is a configuration interface. 6042644bfe51SLorenzo Bianconi * New members can only be added to the end of this structure. 6043644bfe51SLorenzo Bianconi */ 6044644bfe51SLorenzo Bianconi struct bpf_cpumap_val { 6045644bfe51SLorenzo Bianconi __u32 qsize; /* queue size to remote target CPU */ 604692164774SLorenzo Bianconi union { 604792164774SLorenzo Bianconi int fd; /* prog fd on map write */ 604892164774SLorenzo Bianconi __u32 id; /* prog id on map read */ 604992164774SLorenzo Bianconi } bpf_prog; 6050644bfe51SLorenzo Bianconi }; 6051644bfe51SLorenzo Bianconi 6052174a79ffSJohn Fastabend enum sk_action { 6053bfa64075SJohn Fastabend SK_DROP = 0, 6054bfa64075SJohn Fastabend SK_PASS, 6055174a79ffSJohn Fastabend }; 6056174a79ffSJohn Fastabend 60574f738adbSJohn Fastabend /* user accessible metadata for SK_MSG packet hook, new fields must 60584f738adbSJohn Fastabend * be added to the end of this structure 60594f738adbSJohn Fastabend */ 60604f738adbSJohn Fastabend struct sk_msg_md { 6061b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 6062b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 6063303def35SJohn Fastabend 6064303def35SJohn Fastabend __u32 family; 6065303def35SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 6066303def35SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 6067303def35SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 6068303def35SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 6069303def35SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 6070303def35SJohn Fastabend __u32 local_port; /* stored in host byte order */ 60713bdbd022SJohn Fastabend __u32 size; /* Total size of sk_msg */ 607213d70f5aSJohn Fastabend 607313d70f5aSJohn Fastabend __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 60744f738adbSJohn Fastabend }; 60754f738adbSJohn Fastabend 60762dbb9b9eSMartin KaFai Lau struct sk_reuseport_md { 60772dbb9b9eSMartin KaFai Lau /* 60782dbb9b9eSMartin KaFai Lau * Start of directly accessible data. It begins from 60792dbb9b9eSMartin KaFai Lau * the tcp/udp header. 60802dbb9b9eSMartin KaFai Lau */ 6081b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 6082b7df9adaSDaniel Borkmann /* End of directly accessible data */ 6083b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 60842dbb9b9eSMartin KaFai Lau /* 60852dbb9b9eSMartin KaFai Lau * Total length of packet (starting from the tcp/udp header). 60862dbb9b9eSMartin KaFai Lau * Note that the directly accessible bytes (data_end - data) 60872dbb9b9eSMartin KaFai Lau * could be less than this "len". Those bytes could be 60882dbb9b9eSMartin KaFai Lau * indirectly read by a helper "bpf_skb_load_bytes()". 60892dbb9b9eSMartin KaFai Lau */ 60902dbb9b9eSMartin KaFai Lau __u32 len; 60912dbb9b9eSMartin KaFai Lau /* 60922dbb9b9eSMartin KaFai Lau * Eth protocol in the mac header (network byte order). e.g. 60932dbb9b9eSMartin KaFai Lau * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 60942dbb9b9eSMartin KaFai Lau */ 60952dbb9b9eSMartin KaFai Lau __u32 eth_protocol; 60962dbb9b9eSMartin KaFai Lau __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 60972dbb9b9eSMartin KaFai Lau __u32 bind_inany; /* Is sock bound to an INANY address? */ 60982dbb9b9eSMartin KaFai Lau __u32 hash; /* A hash of the packet 4 tuples */ 6099d5e4ddaeSKuniyuki Iwashima /* When reuse->migrating_sk is NULL, it is selecting a sk for the 6100d5e4ddaeSKuniyuki Iwashima * new incoming connection request (e.g. selecting a listen sk for 6101d5e4ddaeSKuniyuki Iwashima * the received SYN in the TCP case). reuse->sk is one of the sk 6102d5e4ddaeSKuniyuki Iwashima * in the reuseport group. The bpf prog can use reuse->sk to learn 6103d5e4ddaeSKuniyuki Iwashima * the local listening ip/port without looking into the skb. 6104d5e4ddaeSKuniyuki Iwashima * 6105d5e4ddaeSKuniyuki Iwashima * When reuse->migrating_sk is not NULL, reuse->sk is closed and 6106d5e4ddaeSKuniyuki Iwashima * reuse->migrating_sk is the socket that needs to be migrated 6107d5e4ddaeSKuniyuki Iwashima * to another listening socket. migrating_sk could be a fullsock 6108d5e4ddaeSKuniyuki Iwashima * sk that is fully established or a reqsk that is in-the-middle 6109d5e4ddaeSKuniyuki Iwashima * of 3-way handshake. 6110d5e4ddaeSKuniyuki Iwashima */ 6111e0610476SKuniyuki Iwashima __bpf_md_ptr(struct bpf_sock *, sk); 6112d5e4ddaeSKuniyuki Iwashima __bpf_md_ptr(struct bpf_sock *, migrating_sk); 61132dbb9b9eSMartin KaFai Lau }; 61142dbb9b9eSMartin KaFai Lau 61151e270976SMartin KaFai Lau #define BPF_TAG_SIZE 8 61161e270976SMartin KaFai Lau 61171e270976SMartin KaFai Lau struct bpf_prog_info { 61181e270976SMartin KaFai Lau __u32 type; 61191e270976SMartin KaFai Lau __u32 id; 61201e270976SMartin KaFai Lau __u8 tag[BPF_TAG_SIZE]; 61211e270976SMartin KaFai Lau __u32 jited_prog_len; 61221e270976SMartin KaFai Lau __u32 xlated_prog_len; 61231e270976SMartin KaFai Lau __aligned_u64 jited_prog_insns; 61241e270976SMartin KaFai Lau __aligned_u64 xlated_prog_insns; 6125cb4d2b3fSMartin KaFai Lau __u64 load_time; /* ns since boottime */ 6126cb4d2b3fSMartin KaFai Lau __u32 created_by_uid; 6127cb4d2b3fSMartin KaFai Lau __u32 nr_map_ids; 6128cb4d2b3fSMartin KaFai Lau __aligned_u64 map_ids; 6129067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 6130675fc275SJakub Kicinski __u32 ifindex; 6131b85fab0eSJiri Olsa __u32 gpl_compatible:1; 61320472301aSBaruch Siach __u32 :31; /* alignment pad */ 6133675fc275SJakub Kicinski __u64 netns_dev; 6134675fc275SJakub Kicinski __u64 netns_ino; 6135dbecd738SSandipan Das __u32 nr_jited_ksyms; 6136815581c1SSandipan Das __u32 nr_jited_func_lens; 6137dbecd738SSandipan Das __aligned_u64 jited_ksyms; 6138815581c1SSandipan Das __aligned_u64 jited_func_lens; 6139838e9690SYonghong Song __u32 btf_id; 6140838e9690SYonghong Song __u32 func_info_rec_size; 6141838e9690SYonghong Song __aligned_u64 func_info; 614211d8b82dSYonghong Song __u32 nr_func_info; 614311d8b82dSYonghong Song __u32 nr_line_info; 6144c454a46bSMartin KaFai Lau __aligned_u64 line_info; 6145c454a46bSMartin KaFai Lau __aligned_u64 jited_line_info; 614611d8b82dSYonghong Song __u32 nr_jited_line_info; 6147c454a46bSMartin KaFai Lau __u32 line_info_rec_size; 6148c454a46bSMartin KaFai Lau __u32 jited_line_info_rec_size; 6149c872bdb3SSong Liu __u32 nr_prog_tags; 6150c872bdb3SSong Liu __aligned_u64 prog_tags; 61515f8f8b93SAlexei Starovoitov __u64 run_time_ns; 61525f8f8b93SAlexei Starovoitov __u64 run_cnt; 61539ed9e9baSAlexei Starovoitov __u64 recursion_misses; 6154aba64c7dSDave Marchevsky __u32 verified_insns; 6155b79c9fc9SStanislav Fomichev __u32 attach_btf_obj_id; 6156b79c9fc9SStanislav Fomichev __u32 attach_btf_id; 61571e270976SMartin KaFai Lau } __attribute__((aligned(8))); 61581e270976SMartin KaFai Lau 61591e270976SMartin KaFai Lau struct bpf_map_info { 61601e270976SMartin KaFai Lau __u32 type; 61611e270976SMartin KaFai Lau __u32 id; 61621e270976SMartin KaFai Lau __u32 key_size; 61631e270976SMartin KaFai Lau __u32 value_size; 61641e270976SMartin KaFai Lau __u32 max_entries; 61651e270976SMartin KaFai Lau __u32 map_flags; 6166067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 616752775b33SJakub Kicinski __u32 ifindex; 616885d33df3SMartin KaFai Lau __u32 btf_vmlinux_value_type_id; 616952775b33SJakub Kicinski __u64 netns_dev; 617052775b33SJakub Kicinski __u64 netns_ino; 617178958fcaSMartin KaFai Lau __u32 btf_id; 61729b2cf328SMartin KaFai Lau __u32 btf_key_type_id; 61739b2cf328SMartin KaFai Lau __u32 btf_value_type_id; 61748845b468SJoanne Koong __u32 :32; /* alignment pad */ 61759330986cSJoanne Koong __u64 map_extra; 61761e270976SMartin KaFai Lau } __attribute__((aligned(8))); 61771e270976SMartin KaFai Lau 617862dab84cSMartin KaFai Lau struct bpf_btf_info { 617962dab84cSMartin KaFai Lau __aligned_u64 btf; 618062dab84cSMartin KaFai Lau __u32 btf_size; 618162dab84cSMartin KaFai Lau __u32 id; 618253297220SAndrii Nakryiko __aligned_u64 name; 618353297220SAndrii Nakryiko __u32 name_len; 618453297220SAndrii Nakryiko __u32 kernel_btf; 618562dab84cSMartin KaFai Lau } __attribute__((aligned(8))); 618662dab84cSMartin KaFai Lau 6187f2e10bffSAndrii Nakryiko struct bpf_link_info { 6188f2e10bffSAndrii Nakryiko __u32 type; 6189f2e10bffSAndrii Nakryiko __u32 id; 6190f2e10bffSAndrii Nakryiko __u32 prog_id; 6191f2e10bffSAndrii Nakryiko union { 6192f2e10bffSAndrii Nakryiko struct { 6193f2e10bffSAndrii Nakryiko __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 6194f2e10bffSAndrii Nakryiko __u32 tp_name_len; /* in/out: tp_name buffer len */ 6195f2e10bffSAndrii Nakryiko } raw_tracepoint; 6196f2e10bffSAndrii Nakryiko struct { 6197f2e10bffSAndrii Nakryiko __u32 attach_type; 6198441e8c66SToke Høiland-Jørgensen __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 6199441e8c66SToke Høiland-Jørgensen __u32 target_btf_id; /* BTF type id inside the object */ 6200f2e10bffSAndrii Nakryiko } tracing; 6201f2e10bffSAndrii Nakryiko struct { 6202f2e10bffSAndrii Nakryiko __u64 cgroup_id; 6203f2e10bffSAndrii Nakryiko __u32 attach_type; 6204f2e10bffSAndrii Nakryiko } cgroup; 62057f045a49SJakub Sitnicki struct { 62066b0a249aSYonghong Song __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 62076b0a249aSYonghong Song __u32 target_name_len; /* in/out: target_name buffer len */ 6208d4ccaf58SHao Luo 6209d4ccaf58SHao Luo /* If the iter specific field is 32 bits, it can be put 6210d4ccaf58SHao Luo * in the first or second union. Otherwise it should be 6211d4ccaf58SHao Luo * put in the second union. 6212d4ccaf58SHao Luo */ 62136b0a249aSYonghong Song union { 6214b0c9eb37SYonghong Song struct { 62156b0a249aSYonghong Song __u32 map_id; 62166b0a249aSYonghong Song } map; 6217b0c9eb37SYonghong Song }; 6218d4ccaf58SHao Luo union { 6219d4ccaf58SHao Luo struct { 6220d4ccaf58SHao Luo __u64 cgroup_id; 6221d4ccaf58SHao Luo __u32 order; 6222d4ccaf58SHao Luo } cgroup; 6223d4ccaf58SHao Luo }; 62246b0a249aSYonghong Song } iter; 62256b0a249aSYonghong Song struct { 62267f045a49SJakub Sitnicki __u32 netns_ino; 62277f045a49SJakub Sitnicki __u32 attach_type; 62287f045a49SJakub Sitnicki } netns; 6229c1931c97SAndrii Nakryiko struct { 6230c1931c97SAndrii Nakryiko __u32 ifindex; 6231c1931c97SAndrii Nakryiko } xdp; 6232f2e10bffSAndrii Nakryiko }; 6233f2e10bffSAndrii Nakryiko } __attribute__((aligned(8))); 6234f2e10bffSAndrii Nakryiko 62354fbac77dSAndrey Ignatov /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 62364fbac77dSAndrey Ignatov * by user and intended to be used by socket (e.g. to bind to, depends on 6237bfdfa517SRandy Dunlap * attach type). 62384fbac77dSAndrey Ignatov */ 62394fbac77dSAndrey Ignatov struct bpf_sock_addr { 62404fbac77dSAndrey Ignatov __u32 user_family; /* Allows 4-byte read, but no write. */ 62414fbac77dSAndrey Ignatov __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 62424fbac77dSAndrey Ignatov * Stored in network byte order. 62434fbac77dSAndrey Ignatov */ 6244d4ecfeb1SStanislav Fomichev __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 62454fbac77dSAndrey Ignatov * Stored in network byte order. 62464fbac77dSAndrey Ignatov */ 62477aebfa1bSAndrey Ignatov __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 62484fbac77dSAndrey Ignatov * Stored in network byte order 62494fbac77dSAndrey Ignatov */ 62504fbac77dSAndrey Ignatov __u32 family; /* Allows 4-byte read, but no write */ 62514fbac77dSAndrey Ignatov __u32 type; /* Allows 4-byte read, but no write */ 62524fbac77dSAndrey Ignatov __u32 protocol; /* Allows 4-byte read, but no write */ 6253600c70baSStanislav Fomichev __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 62541cedee13SAndrey Ignatov * Stored in network byte order. 62551cedee13SAndrey Ignatov */ 6256d4ecfeb1SStanislav Fomichev __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 62571cedee13SAndrey Ignatov * Stored in network byte order. 62581cedee13SAndrey Ignatov */ 6259fb85c4a7SStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 62604fbac77dSAndrey Ignatov }; 62614fbac77dSAndrey Ignatov 626240304b2aSLawrence Brakmo /* User bpf_sock_ops struct to access socket values and specify request ops 626340304b2aSLawrence Brakmo * and their replies. 626440304b2aSLawrence Brakmo * Some of this fields are in network (bigendian) byte order and may need 626540304b2aSLawrence Brakmo * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 626640304b2aSLawrence Brakmo * New fields can only be added at the end of this structure 626740304b2aSLawrence Brakmo */ 626840304b2aSLawrence Brakmo struct bpf_sock_ops { 626940304b2aSLawrence Brakmo __u32 op; 627040304b2aSLawrence Brakmo union { 6271de525be2SLawrence Brakmo __u32 args[4]; /* Optionally passed to bpf program */ 6272de525be2SLawrence Brakmo __u32 reply; /* Returned by bpf program */ 6273de525be2SLawrence Brakmo __u32 replylong[4]; /* Optionally returned by bpf prog */ 627440304b2aSLawrence Brakmo }; 627540304b2aSLawrence Brakmo __u32 family; 627640304b2aSLawrence Brakmo __u32 remote_ip4; /* Stored in network byte order */ 627740304b2aSLawrence Brakmo __u32 local_ip4; /* Stored in network byte order */ 627840304b2aSLawrence Brakmo __u32 remote_ip6[4]; /* Stored in network byte order */ 627940304b2aSLawrence Brakmo __u32 local_ip6[4]; /* Stored in network byte order */ 628040304b2aSLawrence Brakmo __u32 remote_port; /* Stored in network byte order */ 628140304b2aSLawrence Brakmo __u32 local_port; /* stored in host byte order */ 6282f19397a5SLawrence Brakmo __u32 is_fullsock; /* Some TCP fields are only valid if 6283f19397a5SLawrence Brakmo * there is a full socket. If not, the 6284f19397a5SLawrence Brakmo * fields read as zero. 6285f19397a5SLawrence Brakmo */ 6286f19397a5SLawrence Brakmo __u32 snd_cwnd; 6287f19397a5SLawrence Brakmo __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 6288b13d8807SLawrence Brakmo __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 628944f0e430SLawrence Brakmo __u32 state; 629044f0e430SLawrence Brakmo __u32 rtt_min; 629144f0e430SLawrence Brakmo __u32 snd_ssthresh; 629244f0e430SLawrence Brakmo __u32 rcv_nxt; 629344f0e430SLawrence Brakmo __u32 snd_nxt; 629444f0e430SLawrence Brakmo __u32 snd_una; 629544f0e430SLawrence Brakmo __u32 mss_cache; 629644f0e430SLawrence Brakmo __u32 ecn_flags; 629744f0e430SLawrence Brakmo __u32 rate_delivered; 629844f0e430SLawrence Brakmo __u32 rate_interval_us; 629944f0e430SLawrence Brakmo __u32 packets_out; 630044f0e430SLawrence Brakmo __u32 retrans_out; 630144f0e430SLawrence Brakmo __u32 total_retrans; 630244f0e430SLawrence Brakmo __u32 segs_in; 630344f0e430SLawrence Brakmo __u32 data_segs_in; 630444f0e430SLawrence Brakmo __u32 segs_out; 630544f0e430SLawrence Brakmo __u32 data_segs_out; 630644f0e430SLawrence Brakmo __u32 lost_out; 630744f0e430SLawrence Brakmo __u32 sacked_out; 630844f0e430SLawrence Brakmo __u32 sk_txhash; 630944f0e430SLawrence Brakmo __u64 bytes_received; 631044f0e430SLawrence Brakmo __u64 bytes_acked; 63111314ef56SStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 63120813a841SMartin KaFai Lau /* [skb_data, skb_data_end) covers the whole TCP header. 63130813a841SMartin KaFai Lau * 63140813a841SMartin KaFai Lau * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 63150813a841SMartin KaFai Lau * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 63160813a841SMartin KaFai Lau * header has not been written. 63170813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 63180813a841SMartin KaFai Lau * been written so far. 63190813a841SMartin KaFai Lau * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 63200813a841SMartin KaFai Lau * the 3WHS. 63210813a841SMartin KaFai Lau * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 63220813a841SMartin KaFai Lau * the 3WHS. 63230813a841SMartin KaFai Lau * 63240813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also be used to read a particular option. 63250813a841SMartin KaFai Lau */ 63260813a841SMartin KaFai Lau __bpf_md_ptr(void *, skb_data); 63270813a841SMartin KaFai Lau __bpf_md_ptr(void *, skb_data_end); 63280813a841SMartin KaFai Lau __u32 skb_len; /* The total length of a packet. 63290813a841SMartin KaFai Lau * It includes the header, options, 63300813a841SMartin KaFai Lau * and payload. 63310813a841SMartin KaFai Lau */ 63320813a841SMartin KaFai Lau __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 63330813a841SMartin KaFai Lau * an easy way to check for tcp_flags 63340813a841SMartin KaFai Lau * without parsing skb_data. 63350813a841SMartin KaFai Lau * 63360813a841SMartin KaFai Lau * In particular, the skb_tcp_flags 63370813a841SMartin KaFai Lau * will still be available in 63380813a841SMartin KaFai Lau * BPF_SOCK_OPS_HDR_OPT_LEN even though 63390813a841SMartin KaFai Lau * the outgoing header has not 63400813a841SMartin KaFai Lau * been written yet. 63410813a841SMartin KaFai Lau */ 634240304b2aSLawrence Brakmo }; 634340304b2aSLawrence Brakmo 6344b13d8807SLawrence Brakmo /* Definitions for bpf_sock_ops_cb_flags */ 63451aae4bddSAndrii Nakryiko enum { 63461aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 63471aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 63481aae4bddSAndrii Nakryiko BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 63491aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 63500813a841SMartin KaFai Lau /* Call bpf for all received TCP headers. The bpf prog will be 63510813a841SMartin KaFai Lau * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 63520813a841SMartin KaFai Lau * 63530813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 63540813a841SMartin KaFai Lau * for the header option related helpers that will be useful 63550813a841SMartin KaFai Lau * to the bpf programs. 63560813a841SMartin KaFai Lau * 63570813a841SMartin KaFai Lau * It could be used at the client/active side (i.e. connect() side) 63580813a841SMartin KaFai Lau * when the server told it that the server was in syncookie 63590813a841SMartin KaFai Lau * mode and required the active side to resend the bpf-written 63600813a841SMartin KaFai Lau * options. The active side can keep writing the bpf-options until 63610813a841SMartin KaFai Lau * it received a valid packet from the server side to confirm 63620813a841SMartin KaFai Lau * the earlier packet (and options) has been received. The later 63630813a841SMartin KaFai Lau * example patch is using it like this at the active side when the 63640813a841SMartin KaFai Lau * server is in syncookie mode. 63650813a841SMartin KaFai Lau * 63660813a841SMartin KaFai Lau * The bpf prog will usually turn this off in the common cases. 63670813a841SMartin KaFai Lau */ 636800d211a4SMartin KaFai Lau BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 63690813a841SMartin KaFai Lau /* Call bpf when kernel has received a header option that 63700813a841SMartin KaFai Lau * the kernel cannot handle. The bpf prog will be called under 63710813a841SMartin KaFai Lau * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 63720813a841SMartin KaFai Lau * 63730813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 63740813a841SMartin KaFai Lau * for the header option related helpers that will be useful 63750813a841SMartin KaFai Lau * to the bpf programs. 63760813a841SMartin KaFai Lau */ 637700d211a4SMartin KaFai Lau BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 63780813a841SMartin KaFai Lau /* Call bpf when the kernel is writing header options for the 63790813a841SMartin KaFai Lau * outgoing packet. The bpf prog will first be called 63800813a841SMartin KaFai Lau * to reserve space in a skb under 63810813a841SMartin KaFai Lau * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 63820813a841SMartin KaFai Lau * the bpf prog will be called to write the header option(s) 63830813a841SMartin KaFai Lau * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 63840813a841SMartin KaFai Lau * 63850813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 63860813a841SMartin KaFai Lau * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 63870813a841SMartin KaFai Lau * related helpers that will be useful to the bpf programs. 63880813a841SMartin KaFai Lau * 63890813a841SMartin KaFai Lau * The kernel gets its chance to reserve space and write 63900813a841SMartin KaFai Lau * options first before the BPF program does. 63910813a841SMartin KaFai Lau */ 6392331fca43SMartin KaFai Lau BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 63931aae4bddSAndrii Nakryiko /* Mask of all currently supported cb flags */ 6394331fca43SMartin KaFai Lau BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 63951aae4bddSAndrii Nakryiko }; 6396b13d8807SLawrence Brakmo 639740304b2aSLawrence Brakmo /* List of known BPF sock_ops operators. 639840304b2aSLawrence Brakmo * New entries can only be added at the end 639940304b2aSLawrence Brakmo */ 640040304b2aSLawrence Brakmo enum { 640140304b2aSLawrence Brakmo BPF_SOCK_OPS_VOID, 64028550f328SLawrence Brakmo BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 64038550f328SLawrence Brakmo * -1 if default value should be used 64048550f328SLawrence Brakmo */ 640513d3b1ebSLawrence Brakmo BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 640613d3b1ebSLawrence Brakmo * window (in packets) or -1 if default 640713d3b1ebSLawrence Brakmo * value should be used 640813d3b1ebSLawrence Brakmo */ 64099872a4bdSLawrence Brakmo BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 64109872a4bdSLawrence Brakmo * active connection is initialized 64119872a4bdSLawrence Brakmo */ 64129872a4bdSLawrence Brakmo BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 64139872a4bdSLawrence Brakmo * active connection is 64149872a4bdSLawrence Brakmo * established 64159872a4bdSLawrence Brakmo */ 64169872a4bdSLawrence Brakmo BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 64179872a4bdSLawrence Brakmo * passive connection is 64189872a4bdSLawrence Brakmo * established 64199872a4bdSLawrence Brakmo */ 642091b5b21cSLawrence Brakmo BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 642191b5b21cSLawrence Brakmo * needs ECN 642291b5b21cSLawrence Brakmo */ 6423e6546ef6SLawrence Brakmo BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 6424e6546ef6SLawrence Brakmo * based on the path and may be 6425e6546ef6SLawrence Brakmo * dependent on the congestion control 6426e6546ef6SLawrence Brakmo * algorithm. In general it indicates 6427e6546ef6SLawrence Brakmo * a congestion threshold. RTTs above 6428e6546ef6SLawrence Brakmo * this indicate congestion 6429e6546ef6SLawrence Brakmo */ 6430f89013f6SLawrence Brakmo BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 6431f89013f6SLawrence Brakmo * Arg1: value of icsk_retransmits 6432f89013f6SLawrence Brakmo * Arg2: value of icsk_rto 6433f89013f6SLawrence Brakmo * Arg3: whether RTO has expired 6434f89013f6SLawrence Brakmo */ 6435a31ad29eSLawrence Brakmo BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6436a31ad29eSLawrence Brakmo * Arg1: sequence number of 1st byte 6437a31ad29eSLawrence Brakmo * Arg2: # segments 6438a31ad29eSLawrence Brakmo * Arg3: return value of 6439a31ad29eSLawrence Brakmo * tcp_transmit_skb (0 => success) 6440a31ad29eSLawrence Brakmo */ 6441d4487491SLawrence Brakmo BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6442d4487491SLawrence Brakmo * Arg1: old_state 6443d4487491SLawrence Brakmo * Arg2: new_state 6444d4487491SLawrence Brakmo */ 6445f333ee0cSAndrey Ignatov BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6446f333ee0cSAndrey Ignatov * socket transition to LISTEN state. 6447f333ee0cSAndrey Ignatov */ 644823729ff2SStanislav Fomichev BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 644923729ff2SStanislav Fomichev */ 64500813a841SMartin KaFai Lau BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 64510813a841SMartin KaFai Lau * It will be called to handle 64520813a841SMartin KaFai Lau * the packets received at 64530813a841SMartin KaFai Lau * an already established 64540813a841SMartin KaFai Lau * connection. 64550813a841SMartin KaFai Lau * 64560813a841SMartin KaFai Lau * sock_ops->skb_data: 64570813a841SMartin KaFai Lau * Referring to the received skb. 64580813a841SMartin KaFai Lau * It covers the TCP header only. 64590813a841SMartin KaFai Lau * 64600813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also 64610813a841SMartin KaFai Lau * be used to search for a 64620813a841SMartin KaFai Lau * particular option. 64630813a841SMartin KaFai Lau */ 64640813a841SMartin KaFai Lau BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 64650813a841SMartin KaFai Lau * header option later in 64660813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 64670813a841SMartin KaFai Lau * Arg1: bool want_cookie. (in 64680813a841SMartin KaFai Lau * writing SYNACK only) 64690813a841SMartin KaFai Lau * 64700813a841SMartin KaFai Lau * sock_ops->skb_data: 64710813a841SMartin KaFai Lau * Not available because no header has 64720813a841SMartin KaFai Lau * been written yet. 64730813a841SMartin KaFai Lau * 64740813a841SMartin KaFai Lau * sock_ops->skb_tcp_flags: 64750813a841SMartin KaFai Lau * The tcp_flags of the 64760813a841SMartin KaFai Lau * outgoing skb. (e.g. SYN, ACK, FIN). 64770813a841SMartin KaFai Lau * 64780813a841SMartin KaFai Lau * bpf_reserve_hdr_opt() should 64790813a841SMartin KaFai Lau * be used to reserve space. 64800813a841SMartin KaFai Lau */ 64810813a841SMartin KaFai Lau BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 64820813a841SMartin KaFai Lau * Arg1: bool want_cookie. (in 64830813a841SMartin KaFai Lau * writing SYNACK only) 64840813a841SMartin KaFai Lau * 64850813a841SMartin KaFai Lau * sock_ops->skb_data: 64860813a841SMartin KaFai Lau * Referring to the outgoing skb. 64870813a841SMartin KaFai Lau * It covers the TCP header 64880813a841SMartin KaFai Lau * that has already been written 64890813a841SMartin KaFai Lau * by the kernel and the 64900813a841SMartin KaFai Lau * earlier bpf-progs. 64910813a841SMartin KaFai Lau * 64920813a841SMartin KaFai Lau * sock_ops->skb_tcp_flags: 64930813a841SMartin KaFai Lau * The tcp_flags of the outgoing 64940813a841SMartin KaFai Lau * skb. (e.g. SYN, ACK, FIN). 64950813a841SMartin KaFai Lau * 64960813a841SMartin KaFai Lau * bpf_store_hdr_opt() should 64970813a841SMartin KaFai Lau * be used to write the 64980813a841SMartin KaFai Lau * option. 64990813a841SMartin KaFai Lau * 65000813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also 65010813a841SMartin KaFai Lau * be used to search for a 65020813a841SMartin KaFai Lau * particular option that 65030813a841SMartin KaFai Lau * has already been written 65040813a841SMartin KaFai Lau * by the kernel or the 65050813a841SMartin KaFai Lau * earlier bpf-progs. 65060813a841SMartin KaFai Lau */ 6507d4487491SLawrence Brakmo }; 6508d4487491SLawrence Brakmo 6509d4487491SLawrence Brakmo /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 6510d4487491SLawrence Brakmo * changes between the TCP and BPF versions. Ideally this should never happen. 6511d4487491SLawrence Brakmo * If it does, we need to add code to convert them before calling 6512d4487491SLawrence Brakmo * the BPF sock_ops function. 6513d4487491SLawrence Brakmo */ 6514d4487491SLawrence Brakmo enum { 6515d4487491SLawrence Brakmo BPF_TCP_ESTABLISHED = 1, 6516d4487491SLawrence Brakmo BPF_TCP_SYN_SENT, 6517d4487491SLawrence Brakmo BPF_TCP_SYN_RECV, 6518d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT1, 6519d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT2, 6520d4487491SLawrence Brakmo BPF_TCP_TIME_WAIT, 6521d4487491SLawrence Brakmo BPF_TCP_CLOSE, 6522d4487491SLawrence Brakmo BPF_TCP_CLOSE_WAIT, 6523d4487491SLawrence Brakmo BPF_TCP_LAST_ACK, 6524d4487491SLawrence Brakmo BPF_TCP_LISTEN, 6525d4487491SLawrence Brakmo BPF_TCP_CLOSING, /* Now a valid state */ 6526d4487491SLawrence Brakmo BPF_TCP_NEW_SYN_RECV, 6527d4487491SLawrence Brakmo 6528d4487491SLawrence Brakmo BPF_TCP_MAX_STATES /* Leave at the end! */ 652940304b2aSLawrence Brakmo }; 653040304b2aSLawrence Brakmo 65311aae4bddSAndrii Nakryiko enum { 65321aae4bddSAndrii Nakryiko TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 65331aae4bddSAndrii Nakryiko TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 65342b8ee4f0SMartin KaFai Lau TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 6535ca584ba0SMartin KaFai Lau TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 65360813a841SMartin KaFai Lau /* Copy the SYN pkt to optval 65370813a841SMartin KaFai Lau * 65380813a841SMartin KaFai Lau * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 65390813a841SMartin KaFai Lau * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 65400813a841SMartin KaFai Lau * to only getting from the saved_syn. It can either get the 65410813a841SMartin KaFai Lau * syn packet from: 65420813a841SMartin KaFai Lau * 65430813a841SMartin KaFai Lau * 1. the just-received SYN packet (only available when writing the 65440813a841SMartin KaFai Lau * SYNACK). It will be useful when it is not necessary to 65450813a841SMartin KaFai Lau * save the SYN packet for latter use. It is also the only way 65460813a841SMartin KaFai Lau * to get the SYN during syncookie mode because the syn 65470813a841SMartin KaFai Lau * packet cannot be saved during syncookie. 65480813a841SMartin KaFai Lau * 65490813a841SMartin KaFai Lau * OR 65500813a841SMartin KaFai Lau * 65510813a841SMartin KaFai Lau * 2. the earlier saved syn which was done by 65520813a841SMartin KaFai Lau * bpf_setsockopt(TCP_SAVE_SYN). 65530813a841SMartin KaFai Lau * 65540813a841SMartin KaFai Lau * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 65550813a841SMartin KaFai Lau * SYN packet is obtained. 65560813a841SMartin KaFai Lau * 65570813a841SMartin KaFai Lau * If the bpf-prog does not need the IP[46] header, the 65580813a841SMartin KaFai Lau * bpf-prog can avoid parsing the IP header by using 65590813a841SMartin KaFai Lau * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 65600813a841SMartin KaFai Lau * IP[46] and TCP header by using TCP_BPF_SYN_IP. 65610813a841SMartin KaFai Lau * 65620813a841SMartin KaFai Lau * >0: Total number of bytes copied 65630813a841SMartin KaFai Lau * -ENOSPC: Not enough space in optval. Only optlen number of 65640813a841SMartin KaFai Lau * bytes is copied. 65650813a841SMartin KaFai Lau * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 65660813a841SMartin KaFai Lau * is not saved by setsockopt(TCP_SAVE_SYN). 65670813a841SMartin KaFai Lau */ 65680813a841SMartin KaFai Lau TCP_BPF_SYN = 1005, /* Copy the TCP header */ 65690813a841SMartin KaFai Lau TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 6570267cf9faSMartin KaFai Lau TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 65710813a841SMartin KaFai Lau }; 65720813a841SMartin KaFai Lau 65730813a841SMartin KaFai Lau enum { 65740813a841SMartin KaFai Lau BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 65750813a841SMartin KaFai Lau }; 65760813a841SMartin KaFai Lau 65770813a841SMartin KaFai Lau /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 65780813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 65790813a841SMartin KaFai Lau */ 65800813a841SMartin KaFai Lau enum { 65810813a841SMartin KaFai Lau BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 65820813a841SMartin KaFai Lau * total option spaces 65830813a841SMartin KaFai Lau * required for an established 65840813a841SMartin KaFai Lau * sk in order to calculate the 65850813a841SMartin KaFai Lau * MSS. No skb is actually 65860813a841SMartin KaFai Lau * sent. 65870813a841SMartin KaFai Lau */ 65880813a841SMartin KaFai Lau BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 65890813a841SMartin KaFai Lau * when sending a SYN. 65900813a841SMartin KaFai Lau */ 65911aae4bddSAndrii Nakryiko }; 6592fc747810SLawrence Brakmo 6593908432caSYonghong Song struct bpf_perf_event_value { 6594908432caSYonghong Song __u64 counter; 6595908432caSYonghong Song __u64 enabled; 6596908432caSYonghong Song __u64 running; 6597908432caSYonghong Song }; 6598908432caSYonghong Song 65991aae4bddSAndrii Nakryiko enum { 66001aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 66011aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_READ = (1ULL << 1), 66021aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_WRITE = (1ULL << 2), 66031aae4bddSAndrii Nakryiko }; 6604ebc614f6SRoman Gushchin 66051aae4bddSAndrii Nakryiko enum { 66061aae4bddSAndrii Nakryiko BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 66071aae4bddSAndrii Nakryiko BPF_DEVCG_DEV_CHAR = (1ULL << 1), 66081aae4bddSAndrii Nakryiko }; 6609ebc614f6SRoman Gushchin 6610ebc614f6SRoman Gushchin struct bpf_cgroup_dev_ctx { 661106ef0ccbSYonghong Song /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 661206ef0ccbSYonghong Song __u32 access_type; 6613ebc614f6SRoman Gushchin __u32 major; 6614ebc614f6SRoman Gushchin __u32 minor; 6615ebc614f6SRoman Gushchin }; 6616ebc614f6SRoman Gushchin 6617c4f6699dSAlexei Starovoitov struct bpf_raw_tracepoint_args { 6618c4f6699dSAlexei Starovoitov __u64 args[0]; 6619c4f6699dSAlexei Starovoitov }; 6620c4f6699dSAlexei Starovoitov 662187f5fc7eSDavid Ahern /* DIRECT: Skip the FIB rules and go to FIB table associated with device 662287f5fc7eSDavid Ahern * OUTPUT: Do lookup from egress perspective; default is ingress 662387f5fc7eSDavid Ahern */ 66241aae4bddSAndrii Nakryiko enum { 66251aae4bddSAndrii Nakryiko BPF_FIB_LOOKUP_DIRECT = (1U << 0), 66261aae4bddSAndrii Nakryiko BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 66271aae4bddSAndrii Nakryiko }; 662887f5fc7eSDavid Ahern 66294c79579bSDavid Ahern enum { 66304c79579bSDavid Ahern BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 66314c79579bSDavid Ahern BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 66324c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 66334c79579bSDavid Ahern BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 66344c79579bSDavid Ahern BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 66354c79579bSDavid Ahern BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 66364c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 66374c79579bSDavid Ahern BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 66384c79579bSDavid Ahern BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 66394c79579bSDavid Ahern }; 66404c79579bSDavid Ahern 664187f5fc7eSDavid Ahern struct bpf_fib_lookup { 6642fa898d76SDavid Ahern /* input: network family for lookup (AF_INET, AF_INET6) 6643fa898d76SDavid Ahern * output: network family of egress nexthop 6644fa898d76SDavid Ahern */ 6645fa898d76SDavid Ahern __u8 family; 664687f5fc7eSDavid Ahern 664787f5fc7eSDavid Ahern /* set if lookup is to consider L4 data - e.g., FIB rules */ 664887f5fc7eSDavid Ahern __u8 l4_protocol; 664987f5fc7eSDavid Ahern __be16 sport; 665087f5fc7eSDavid Ahern __be16 dport; 665187f5fc7eSDavid Ahern 6652e1850ea9SJesper Dangaard Brouer union { /* used for MTU check */ 6653e1850ea9SJesper Dangaard Brouer /* input to lookup */ 6654e1850ea9SJesper Dangaard Brouer __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 66554c79579bSDavid Ahern 6656e1850ea9SJesper Dangaard Brouer /* output: MTU value */ 6657e1850ea9SJesper Dangaard Brouer __u16 mtu_result; 6658e1850ea9SJesper Dangaard Brouer }; 66594c79579bSDavid Ahern /* input: L3 device index for lookup 66604c79579bSDavid Ahern * output: device index from FIB lookup 66614c79579bSDavid Ahern */ 66624c79579bSDavid Ahern __u32 ifindex; 666387f5fc7eSDavid Ahern 666487f5fc7eSDavid Ahern union { 666587f5fc7eSDavid Ahern /* inputs to lookup */ 666687f5fc7eSDavid Ahern __u8 tos; /* AF_INET */ 6667bd3a08aaSDavid Ahern __be32 flowinfo; /* AF_INET6, flow_label + priority */ 666887f5fc7eSDavid Ahern 6669fa898d76SDavid Ahern /* output: metric of fib result (IPv4/IPv6 only) */ 667087f5fc7eSDavid Ahern __u32 rt_metric; 667187f5fc7eSDavid Ahern }; 667287f5fc7eSDavid Ahern 667387f5fc7eSDavid Ahern union { 667487f5fc7eSDavid Ahern __be32 ipv4_src; 667587f5fc7eSDavid Ahern __u32 ipv6_src[4]; /* in6_addr; network order */ 667687f5fc7eSDavid Ahern }; 667787f5fc7eSDavid Ahern 6678fa898d76SDavid Ahern /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 6679fa898d76SDavid Ahern * network header. output: bpf_fib_lookup sets to gateway address 6680fa898d76SDavid Ahern * if FIB lookup returns gateway route 668187f5fc7eSDavid Ahern */ 668287f5fc7eSDavid Ahern union { 668387f5fc7eSDavid Ahern __be32 ipv4_dst; 668487f5fc7eSDavid Ahern __u32 ipv6_dst[4]; /* in6_addr; network order */ 668587f5fc7eSDavid Ahern }; 668687f5fc7eSDavid Ahern 668787f5fc7eSDavid Ahern /* output */ 668887f5fc7eSDavid Ahern __be16 h_vlan_proto; 668987f5fc7eSDavid Ahern __be16 h_vlan_TCI; 669087f5fc7eSDavid Ahern __u8 smac[6]; /* ETH_ALEN */ 669187f5fc7eSDavid Ahern __u8 dmac[6]; /* ETH_ALEN */ 669287f5fc7eSDavid Ahern }; 669387f5fc7eSDavid Ahern 6694ba452c9eSToke Høiland-Jørgensen struct bpf_redir_neigh { 6695ba452c9eSToke Høiland-Jørgensen /* network family for lookup (AF_INET, AF_INET6) */ 6696ba452c9eSToke Høiland-Jørgensen __u32 nh_family; 6697ba452c9eSToke Høiland-Jørgensen /* network address of nexthop; skips fib lookup to find gateway */ 6698ba452c9eSToke Høiland-Jørgensen union { 6699ba452c9eSToke Høiland-Jørgensen __be32 ipv4_nh; 6700ba452c9eSToke Høiland-Jørgensen __u32 ipv6_nh[4]; /* in6_addr; network order */ 6701ba452c9eSToke Høiland-Jørgensen }; 6702ba452c9eSToke Høiland-Jørgensen }; 6703ba452c9eSToke Høiland-Jørgensen 670434b2021cSJesper Dangaard Brouer /* bpf_check_mtu flags*/ 670534b2021cSJesper Dangaard Brouer enum bpf_check_mtu_flags { 670634b2021cSJesper Dangaard Brouer BPF_MTU_CHK_SEGS = (1U << 0), 670734b2021cSJesper Dangaard Brouer }; 670834b2021cSJesper Dangaard Brouer 670934b2021cSJesper Dangaard Brouer enum bpf_check_mtu_ret { 671034b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 671134b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 671234b2021cSJesper Dangaard Brouer BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 671334b2021cSJesper Dangaard Brouer }; 671434b2021cSJesper Dangaard Brouer 671541bdc4b4SYonghong Song enum bpf_task_fd_type { 671641bdc4b4SYonghong Song BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 671741bdc4b4SYonghong Song BPF_FD_TYPE_TRACEPOINT, /* tp name */ 671841bdc4b4SYonghong Song BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 671941bdc4b4SYonghong Song BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 672041bdc4b4SYonghong Song BPF_FD_TYPE_UPROBE, /* filename + offset */ 672141bdc4b4SYonghong Song BPF_FD_TYPE_URETPROBE, /* filename + offset */ 672241bdc4b4SYonghong Song }; 672341bdc4b4SYonghong Song 67241aae4bddSAndrii Nakryiko enum { 67251aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 67261aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 67271aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 67281aae4bddSAndrii Nakryiko }; 6729086f9568SStanislav Fomichev 6730d58e468bSPetar Penkov struct bpf_flow_keys { 6731d58e468bSPetar Penkov __u16 nhoff; 6732d58e468bSPetar Penkov __u16 thoff; 6733d58e468bSPetar Penkov __u16 addr_proto; /* ETH_P_* of valid addrs */ 6734d58e468bSPetar Penkov __u8 is_frag; 6735d58e468bSPetar Penkov __u8 is_first_frag; 6736d58e468bSPetar Penkov __u8 is_encap; 6737d58e468bSPetar Penkov __u8 ip_proto; 6738d58e468bSPetar Penkov __be16 n_proto; 6739d58e468bSPetar Penkov __be16 sport; 6740d58e468bSPetar Penkov __be16 dport; 6741d58e468bSPetar Penkov union { 6742d58e468bSPetar Penkov struct { 6743d58e468bSPetar Penkov __be32 ipv4_src; 6744d58e468bSPetar Penkov __be32 ipv4_dst; 6745d58e468bSPetar Penkov }; 6746d58e468bSPetar Penkov struct { 6747d58e468bSPetar Penkov __u32 ipv6_src[4]; /* in6_addr; network order */ 6748d58e468bSPetar Penkov __u32 ipv6_dst[4]; /* in6_addr; network order */ 6749d58e468bSPetar Penkov }; 6750d58e468bSPetar Penkov }; 6751086f9568SStanislav Fomichev __u32 flags; 675271c99e32SStanislav Fomichev __be32 flow_label; 6753d58e468bSPetar Penkov }; 6754d58e468bSPetar Penkov 6755838e9690SYonghong Song struct bpf_func_info { 6756d30d42e0SMartin KaFai Lau __u32 insn_off; 6757838e9690SYonghong Song __u32 type_id; 6758838e9690SYonghong Song }; 6759838e9690SYonghong Song 6760c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 6761c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 6762c454a46bSMartin KaFai Lau 6763c454a46bSMartin KaFai Lau struct bpf_line_info { 6764c454a46bSMartin KaFai Lau __u32 insn_off; 6765c454a46bSMartin KaFai Lau __u32 file_name_off; 6766c454a46bSMartin KaFai Lau __u32 line_off; 6767c454a46bSMartin KaFai Lau __u32 line_col; 6768c454a46bSMartin KaFai Lau }; 6769c454a46bSMartin KaFai Lau 6770d83525caSAlexei Starovoitov struct bpf_spin_lock { 6771d83525caSAlexei Starovoitov __u32 val; 6772d83525caSAlexei Starovoitov }; 67737b146cebSAndrey Ignatov 6774b00628b1SAlexei Starovoitov struct bpf_timer { 6775b00628b1SAlexei Starovoitov __u64 :64; 6776b00628b1SAlexei Starovoitov __u64 :64; 6777b00628b1SAlexei Starovoitov } __attribute__((aligned(8))); 6778b00628b1SAlexei Starovoitov 677997e03f52SJoanne Koong struct bpf_dynptr { 678097e03f52SJoanne Koong __u64 :64; 678197e03f52SJoanne Koong __u64 :64; 678297e03f52SJoanne Koong } __attribute__((aligned(8))); 678397e03f52SJoanne Koong 67847b146cebSAndrey Ignatov struct bpf_sysctl { 67857b146cebSAndrey Ignatov __u32 write; /* Sysctl is being read (= 0) or written (= 1). 67867b146cebSAndrey Ignatov * Allows 1,2,4-byte read, but no write. 67877b146cebSAndrey Ignatov */ 6788e1550bfeSAndrey Ignatov __u32 file_pos; /* Sysctl file position to read from, write to. 6789e1550bfeSAndrey Ignatov * Allows 1,2,4-byte read an 4-byte write. 6790e1550bfeSAndrey Ignatov */ 67917b146cebSAndrey Ignatov }; 67927b146cebSAndrey Ignatov 67930d01da6aSStanislav Fomichev struct bpf_sockopt { 67940d01da6aSStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 67950d01da6aSStanislav Fomichev __bpf_md_ptr(void *, optval); 67960d01da6aSStanislav Fomichev __bpf_md_ptr(void *, optval_end); 67970d01da6aSStanislav Fomichev 67980d01da6aSStanislav Fomichev __s32 level; 67990d01da6aSStanislav Fomichev __s32 optname; 68000d01da6aSStanislav Fomichev __s32 optlen; 68010d01da6aSStanislav Fomichev __s32 retval; 68020d01da6aSStanislav Fomichev }; 68030d01da6aSStanislav Fomichev 6804b4490c5cSCarlos Neira struct bpf_pidns_info { 6805b4490c5cSCarlos Neira __u32 pid; 6806b4490c5cSCarlos Neira __u32 tgid; 6807b4490c5cSCarlos Neira }; 6808e9ddbb77SJakub Sitnicki 6809e9ddbb77SJakub Sitnicki /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 6810e9ddbb77SJakub Sitnicki struct bpf_sk_lookup { 68117c32e8f8SLorenz Bauer union { 6812e9ddbb77SJakub Sitnicki __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 68137c32e8f8SLorenz Bauer __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 68147c32e8f8SLorenz Bauer }; 6815e9ddbb77SJakub Sitnicki 6816e9ddbb77SJakub Sitnicki __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 6817e9ddbb77SJakub Sitnicki __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 6818e9ddbb77SJakub Sitnicki __u32 remote_ip4; /* Network byte order */ 6819e9ddbb77SJakub Sitnicki __u32 remote_ip6[4]; /* Network byte order */ 68209a69e2b3SJakub Sitnicki __be16 remote_port; /* Network byte order */ 68219a69e2b3SJakub Sitnicki __u16 :16; /* Zero padding */ 6822e9ddbb77SJakub Sitnicki __u32 local_ip4; /* Network byte order */ 6823e9ddbb77SJakub Sitnicki __u32 local_ip6[4]; /* Network byte order */ 6824e9ddbb77SJakub Sitnicki __u32 local_port; /* Host byte order */ 6825f8931565SMark Pashmfouroush __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 6826e9ddbb77SJakub Sitnicki }; 6827e9ddbb77SJakub Sitnicki 6828c4d0bfb4SAlan Maguire /* 6829c4d0bfb4SAlan Maguire * struct btf_ptr is used for typed pointer representation; the 6830c4d0bfb4SAlan Maguire * type id is used to render the pointer data as the appropriate type 6831c4d0bfb4SAlan Maguire * via the bpf_snprintf_btf() helper described above. A flags field - 6832c4d0bfb4SAlan Maguire * potentially to specify additional details about the BTF pointer 6833c4d0bfb4SAlan Maguire * (rather than its mode of display) - is included for future use. 6834c4d0bfb4SAlan Maguire * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 6835c4d0bfb4SAlan Maguire */ 6836c4d0bfb4SAlan Maguire struct btf_ptr { 6837c4d0bfb4SAlan Maguire void *ptr; 6838c4d0bfb4SAlan Maguire __u32 type_id; 6839c4d0bfb4SAlan Maguire __u32 flags; /* BTF ptr flags; unused at present. */ 6840c4d0bfb4SAlan Maguire }; 6841c4d0bfb4SAlan Maguire 6842c4d0bfb4SAlan Maguire /* 6843c4d0bfb4SAlan Maguire * Flags to control bpf_snprintf_btf() behaviour. 6844c4d0bfb4SAlan Maguire * - BTF_F_COMPACT: no formatting around type information 6845c4d0bfb4SAlan Maguire * - BTF_F_NONAME: no struct/union member names/types 6846c4d0bfb4SAlan Maguire * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 6847c4d0bfb4SAlan Maguire * equivalent to %px. 6848c4d0bfb4SAlan Maguire * - BTF_F_ZERO: show zero-valued struct/union members; they 6849c4d0bfb4SAlan Maguire * are not displayed by default 6850c4d0bfb4SAlan Maguire */ 6851c4d0bfb4SAlan Maguire enum { 6852c4d0bfb4SAlan Maguire BTF_F_COMPACT = (1ULL << 0), 6853c4d0bfb4SAlan Maguire BTF_F_NONAME = (1ULL << 1), 6854c4d0bfb4SAlan Maguire BTF_F_PTR_RAW = (1ULL << 2), 6855c4d0bfb4SAlan Maguire BTF_F_ZERO = (1ULL << 3), 6856c4d0bfb4SAlan Maguire }; 6857c4d0bfb4SAlan Maguire 685846334a0cSAlexei Starovoitov /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 685946334a0cSAlexei Starovoitov * has to be adjusted by relocations. It is emitted by llvm and passed to 686046334a0cSAlexei Starovoitov * libbpf and later to the kernel. 686146334a0cSAlexei Starovoitov */ 686246334a0cSAlexei Starovoitov enum bpf_core_relo_kind { 686346334a0cSAlexei Starovoitov BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 686446334a0cSAlexei Starovoitov BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 686546334a0cSAlexei Starovoitov BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 686646334a0cSAlexei Starovoitov BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 686746334a0cSAlexei Starovoitov BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 686846334a0cSAlexei Starovoitov BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 686946334a0cSAlexei Starovoitov BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 687046334a0cSAlexei Starovoitov BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 687146334a0cSAlexei Starovoitov BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 687246334a0cSAlexei Starovoitov BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 687346334a0cSAlexei Starovoitov BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 687446334a0cSAlexei Starovoitov BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 68753c660a5dSDaniel Müller BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */ 687646334a0cSAlexei Starovoitov }; 687746334a0cSAlexei Starovoitov 6878fbd94c7aSAlexei Starovoitov /* 6879fbd94c7aSAlexei Starovoitov * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 6880fbd94c7aSAlexei Starovoitov * and from libbpf to the kernel. 6881fbd94c7aSAlexei Starovoitov * 6882fbd94c7aSAlexei Starovoitov * CO-RE relocation captures the following data: 6883fbd94c7aSAlexei Starovoitov * - insn_off - instruction offset (in bytes) within a BPF program that needs 6884fbd94c7aSAlexei Starovoitov * its insn->imm field to be relocated with actual field info; 6885fbd94c7aSAlexei Starovoitov * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 6886fbd94c7aSAlexei Starovoitov * type or field; 6887fbd94c7aSAlexei Starovoitov * - access_str_off - offset into corresponding .BTF string section. String 6888fbd94c7aSAlexei Starovoitov * interpretation depends on specific relocation kind: 6889fbd94c7aSAlexei Starovoitov * - for field-based relocations, string encodes an accessed field using 6890fbd94c7aSAlexei Starovoitov * a sequence of field and array indices, separated by colon (:). It's 6891fbd94c7aSAlexei Starovoitov * conceptually very close to LLVM's getelementptr ([0]) instruction's 6892fbd94c7aSAlexei Starovoitov * arguments for identifying offset to a field. 6893fbd94c7aSAlexei Starovoitov * - for type-based relocations, strings is expected to be just "0"; 6894fbd94c7aSAlexei Starovoitov * - for enum value-based relocations, string contains an index of enum 6895fbd94c7aSAlexei Starovoitov * value within its enum type; 6896fbd94c7aSAlexei Starovoitov * - kind - one of enum bpf_core_relo_kind; 6897fbd94c7aSAlexei Starovoitov * 6898fbd94c7aSAlexei Starovoitov * Example: 6899fbd94c7aSAlexei Starovoitov * struct sample { 6900fbd94c7aSAlexei Starovoitov * int a; 6901fbd94c7aSAlexei Starovoitov * struct { 6902fbd94c7aSAlexei Starovoitov * int b[10]; 6903fbd94c7aSAlexei Starovoitov * }; 6904fbd94c7aSAlexei Starovoitov * }; 6905fbd94c7aSAlexei Starovoitov * 6906fbd94c7aSAlexei Starovoitov * struct sample *s = ...; 6907fbd94c7aSAlexei Starovoitov * int *x = &s->a; // encoded as "0:0" (a is field #0) 6908fbd94c7aSAlexei Starovoitov * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 6909fbd94c7aSAlexei Starovoitov * // b is field #0 inside anon struct, accessing elem #5) 6910fbd94c7aSAlexei Starovoitov * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 6911fbd94c7aSAlexei Starovoitov * 6912fbd94c7aSAlexei Starovoitov * type_id for all relocs in this example will capture BTF type id of 6913fbd94c7aSAlexei Starovoitov * `struct sample`. 6914fbd94c7aSAlexei Starovoitov * 6915fbd94c7aSAlexei Starovoitov * Such relocation is emitted when using __builtin_preserve_access_index() 6916fbd94c7aSAlexei Starovoitov * Clang built-in, passing expression that captures field address, e.g.: 6917fbd94c7aSAlexei Starovoitov * 6918fbd94c7aSAlexei Starovoitov * bpf_probe_read(&dst, sizeof(dst), 6919fbd94c7aSAlexei Starovoitov * __builtin_preserve_access_index(&src->a.b.c)); 6920fbd94c7aSAlexei Starovoitov * 6921fbd94c7aSAlexei Starovoitov * In this case Clang will emit field relocation recording necessary data to 6922fbd94c7aSAlexei Starovoitov * be able to find offset of embedded `a.b.c` field within `src` struct. 6923fbd94c7aSAlexei Starovoitov * 6924fbd94c7aSAlexei Starovoitov * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 6925fbd94c7aSAlexei Starovoitov */ 6926fbd94c7aSAlexei Starovoitov struct bpf_core_relo { 6927fbd94c7aSAlexei Starovoitov __u32 insn_off; 6928fbd94c7aSAlexei Starovoitov __u32 type_id; 6929fbd94c7aSAlexei Starovoitov __u32 access_str_off; 6930fbd94c7aSAlexei Starovoitov enum bpf_core_relo_kind kind; 6931fbd94c7aSAlexei Starovoitov }; 6932fbd94c7aSAlexei Starovoitov 6933daedfb22SAlexei Starovoitov #endif /* _UAPI__LINUX_BPF_H__ */ 6934