686 lines · plain
1.. SPDX-License-Identifier: GPL-2.02 3.. _networking-filter:4 5=======================================================6Linux Socket Filtering aka Berkeley Packet Filter (BPF)7=======================================================8 9Notice10------11 12This file used to document the eBPF format and mechanisms even when not13related to socket filtering. The ../bpf/index.rst has more details14on eBPF.15 16Introduction17------------18 19Linux Socket Filtering (LSF) is derived from the Berkeley Packet Filter.20Though there are some distinct differences between the BSD and Linux21Kernel filtering, but when we speak of BPF or LSF in Linux context, we22mean the very same mechanism of filtering in the Linux kernel.23 24BPF allows a user-space program to attach a filter onto any socket and25allow or disallow certain types of data to come through the socket. LSF26follows exactly the same filter code structure as BSD's BPF, so referring27to the BSD bpf.4 manpage is very helpful in creating filters.28 29On Linux, BPF is much simpler than on BSD. One does not have to worry30about devices or anything like that. You simply create your filter code,31send it to the kernel via the SO_ATTACH_FILTER option and if your filter32code passes the kernel check on it, you then immediately begin filtering33data on that socket.34 35You can also detach filters from your socket via the SO_DETACH_FILTER36option. This will probably not be used much since when you close a socket37that has a filter on it the filter is automagically removed. The other38less common case may be adding a different filter on the same socket where39you had another filter that is still running: the kernel takes care of40removing the old one and placing your new one in its place, assuming your41filter has passed the checks, otherwise if it fails the old filter will42remain on that socket.43 44SO_LOCK_FILTER option allows to lock the filter attached to a socket. Once45set, a filter cannot be removed or changed. This allows one process to46setup a socket, attach a filter, lock it then drop privileges and be47assured that the filter will be kept until the socket is closed.48 49The biggest user of this construct might be libpcap. Issuing a high-level50filter command like `tcpdump -i em1 port 22` passes through the libpcap51internal compiler that generates a structure that can eventually be loaded52via SO_ATTACH_FILTER to the kernel. `tcpdump -i em1 port 22 -ddd`53displays what is being placed into this structure.54 55Although we were only speaking about sockets here, BPF in Linux is used56in many more places. There's xt_bpf for netfilter, cls_bpf in the kernel57qdisc layer, SECCOMP-BPF (SECure COMPuting [1]_), and lots of other places58such as team driver, PTP code, etc where BPF is being used.59 60.. [1] Documentation/userspace-api/seccomp_filter.rst61 62Original BPF paper:63 64Steven McCanne and Van Jacobson. 1993. The BSD packet filter: a new65architecture for user-level packet capture. In Proceedings of the66USENIX Winter 1993 Conference Proceedings on USENIX Winter 199367Conference Proceedings (USENIX'93). USENIX Association, Berkeley,68CA, USA, 2-2. [http://www.tcpdump.org/papers/bpf-usenix93.pdf]69 70Structure71---------72 73User space applications include <linux/filter.h> which contains the74following relevant structures::75 76 struct sock_filter { /* Filter block */77 __u16 code; /* Actual filter code */78 __u8 jt; /* Jump true */79 __u8 jf; /* Jump false */80 __u32 k; /* Generic multiuse field */81 };82 83Such a structure is assembled as an array of 4-tuples, that contains84a code, jt, jf and k value. jt and jf are jump offsets and k a generic85value to be used for a provided code::86 87 struct sock_fprog { /* Required for SO_ATTACH_FILTER. */88 unsigned short len; /* Number of filter blocks */89 struct sock_filter __user *filter;90 };91 92For socket filtering, a pointer to this structure (as shown in93follow-up example) is being passed to the kernel through setsockopt(2).94 95Example96-------97 98::99 100 #include <sys/socket.h>101 #include <sys/types.h>102 #include <arpa/inet.h>103 #include <linux/if_ether.h>104 /* ... */105 106 /* From the example above: tcpdump -i em1 port 22 -dd */107 struct sock_filter code[] = {108 { 0x28, 0, 0, 0x0000000c },109 { 0x15, 0, 8, 0x000086dd },110 { 0x30, 0, 0, 0x00000014 },111 { 0x15, 2, 0, 0x00000084 },112 { 0x15, 1, 0, 0x00000006 },113 { 0x15, 0, 17, 0x00000011 },114 { 0x28, 0, 0, 0x00000036 },115 { 0x15, 14, 0, 0x00000016 },116 { 0x28, 0, 0, 0x00000038 },117 { 0x15, 12, 13, 0x00000016 },118 { 0x15, 0, 12, 0x00000800 },119 { 0x30, 0, 0, 0x00000017 },120 { 0x15, 2, 0, 0x00000084 },121 { 0x15, 1, 0, 0x00000006 },122 { 0x15, 0, 8, 0x00000011 },123 { 0x28, 0, 0, 0x00000014 },124 { 0x45, 6, 0, 0x00001fff },125 { 0xb1, 0, 0, 0x0000000e },126 { 0x48, 0, 0, 0x0000000e },127 { 0x15, 2, 0, 0x00000016 },128 { 0x48, 0, 0, 0x00000010 },129 { 0x15, 0, 1, 0x00000016 },130 { 0x06, 0, 0, 0x0000ffff },131 { 0x06, 0, 0, 0x00000000 },132 };133 134 struct sock_fprog bpf = {135 .len = ARRAY_SIZE(code),136 .filter = code,137 };138 139 sock = socket(PF_PACKET, SOCK_RAW, htons(ETH_P_ALL));140 if (sock < 0)141 /* ... bail out ... */142 143 ret = setsockopt(sock, SOL_SOCKET, SO_ATTACH_FILTER, &bpf, sizeof(bpf));144 if (ret < 0)145 /* ... bail out ... */146 147 /* ... */148 close(sock);149 150The above example code attaches a socket filter for a PF_PACKET socket151in order to let all IPv4/IPv6 packets with port 22 pass. The rest will152be dropped for this socket.153 154The setsockopt(2) call to SO_DETACH_FILTER doesn't need any arguments155and SO_LOCK_FILTER for preventing the filter to be detached, takes an156integer value with 0 or 1.157 158Note that socket filters are not restricted to PF_PACKET sockets only,159but can also be used on other socket families.160 161Summary of system calls:162 163 * setsockopt(sockfd, SOL_SOCKET, SO_ATTACH_FILTER, &val, sizeof(val));164 * setsockopt(sockfd, SOL_SOCKET, SO_DETACH_FILTER, &val, sizeof(val));165 * setsockopt(sockfd, SOL_SOCKET, SO_LOCK_FILTER, &val, sizeof(val));166 167Normally, most use cases for socket filtering on packet sockets will be168covered by libpcap in high-level syntax, so as an application developer169you should stick to that. libpcap wraps its own layer around all that.170 171Unless i) using/linking to libpcap is not an option, ii) the required BPF172filters use Linux extensions that are not supported by libpcap's compiler,173iii) a filter might be more complex and not cleanly implementable with174libpcap's compiler, or iv) particular filter codes should be optimized175differently than libpcap's internal compiler does; then in such cases176writing such a filter "by hand" can be of an alternative. For example,177xt_bpf and cls_bpf users might have requirements that could result in178more complex filter code, or one that cannot be expressed with libpcap179(e.g. different return codes for various code paths). Moreover, BPF JIT180implementors may wish to manually write test cases and thus need low-level181access to BPF code as well.182 183BPF engine and instruction set184------------------------------185 186Under tools/bpf/ there's a small helper tool called bpf_asm which can187be used to write low-level filters for example scenarios mentioned in the188previous section. Asm-like syntax mentioned here has been implemented in189bpf_asm and will be used for further explanations (instead of dealing with190less readable opcodes directly, principles are the same). The syntax is191closely modelled after Steven McCanne's and Van Jacobson's BPF paper.192 193The BPF architecture consists of the following basic elements:194 195 ======= ====================================================196 Element Description197 ======= ====================================================198 A 32 bit wide accumulator199 X 32 bit wide X register200 M[] 16 x 32 bit wide misc registers aka "scratch memory201 store", addressable from 0 to 15202 ======= ====================================================203 204A program, that is translated by bpf_asm into "opcodes" is an array that205consists of the following elements (as already mentioned)::206 207 op:16, jt:8, jf:8, k:32208 209The element op is a 16 bit wide opcode that has a particular instruction210encoded. jt and jf are two 8 bit wide jump targets, one for condition211"jump if true", the other one "jump if false". Eventually, element k212contains a miscellaneous argument that can be interpreted in different213ways depending on the given instruction in op.214 215The instruction set consists of load, store, branch, alu, miscellaneous216and return instructions that are also represented in bpf_asm syntax. This217table lists all bpf_asm instructions available resp. what their underlying218opcodes as defined in linux/filter.h stand for:219 220 =========== =================== =====================221 Instruction Addressing mode Description222 =========== =================== =====================223 ld 1, 2, 3, 4, 12 Load word into A224 ldi 4 Load word into A225 ldh 1, 2 Load half-word into A226 ldb 1, 2 Load byte into A227 ldx 3, 4, 5, 12 Load word into X228 ldxi 4 Load word into X229 ldxb 5 Load byte into X230 231 st 3 Store A into M[]232 stx 3 Store X into M[]233 234 jmp 6 Jump to label235 ja 6 Jump to label236 jeq 7, 8, 9, 10 Jump on A == <x>237 jneq 9, 10 Jump on A != <x>238 jne 9, 10 Jump on A != <x>239 jlt 9, 10 Jump on A < <x>240 jle 9, 10 Jump on A <= <x>241 jgt 7, 8, 9, 10 Jump on A > <x>242 jge 7, 8, 9, 10 Jump on A >= <x>243 jset 7, 8, 9, 10 Jump on A & <x>244 245 add 0, 4 A + <x>246 sub 0, 4 A - <x>247 mul 0, 4 A * <x>248 div 0, 4 A / <x>249 mod 0, 4 A % <x>250 neg !A251 and 0, 4 A & <x>252 or 0, 4 A | <x>253 xor 0, 4 A ^ <x>254 lsh 0, 4 A << <x>255 rsh 0, 4 A >> <x>256 257 tax Copy A into X258 txa Copy X into A259 260 ret 4, 11 Return261 =========== =================== =====================262 263The next table shows addressing formats from the 2nd column:264 265 =============== =================== ===============================================266 Addressing mode Syntax Description267 =============== =================== ===============================================268 0 x/%x Register X269 1 [k] BHW at byte offset k in the packet270 2 [x + k] BHW at the offset X + k in the packet271 3 M[k] Word at offset k in M[]272 4 #k Literal value stored in k273 5 4*([k]&0xf) Lower nibble * 4 at byte offset k in the packet274 6 L Jump label L275 7 #k,Lt,Lf Jump to Lt if true, otherwise jump to Lf276 8 x/%x,Lt,Lf Jump to Lt if true, otherwise jump to Lf277 9 #k,Lt Jump to Lt if predicate is true278 10 x/%x,Lt Jump to Lt if predicate is true279 11 a/%a Accumulator A280 12 extension BPF extension281 =============== =================== ===============================================282 283The Linux kernel also has a couple of BPF extensions that are used along284with the class of load instructions by "overloading" the k argument with285a negative offset + a particular extension offset. The result of such BPF286extensions are loaded into A.287 288Possible BPF extensions are shown in the following table:289 290 =================================== =================================================291 Extension Description292 =================================== =================================================293 len skb->len294 proto skb->protocol295 type skb->pkt_type296 poff Payload start offset297 ifidx skb->dev->ifindex298 nla Netlink attribute of type X with offset A299 nlan Nested Netlink attribute of type X with offset A300 mark skb->mark301 queue skb->queue_mapping302 hatype skb->dev->type303 rxhash skb->hash304 cpu raw_smp_processor_id()305 vlan_tci skb_vlan_tag_get(skb)306 vlan_avail skb_vlan_tag_present(skb)307 vlan_tpid skb->vlan_proto308 rand get_random_u32()309 =================================== =================================================310 311These extensions can also be prefixed with '#'.312Examples for low-level BPF:313 314**ARP packets**::315 316 ldh [12]317 jne #0x806, drop318 ret #-1319 drop: ret #0320 321**IPv4 TCP packets**::322 323 ldh [12]324 jne #0x800, drop325 ldb [23]326 jneq #6, drop327 ret #-1328 drop: ret #0329 330**icmp random packet sampling, 1 in 4**::331 332 ldh [12]333 jne #0x800, drop334 ldb [23]335 jneq #1, drop336 # get a random uint32 number337 ld rand338 mod #4339 jneq #1, drop340 ret #-1341 drop: ret #0342 343**SECCOMP filter example**::344 345 ld [4] /* offsetof(struct seccomp_data, arch) */346 jne #0xc000003e, bad /* AUDIT_ARCH_X86_64 */347 ld [0] /* offsetof(struct seccomp_data, nr) */348 jeq #15, good /* __NR_rt_sigreturn */349 jeq #231, good /* __NR_exit_group */350 jeq #60, good /* __NR_exit */351 jeq #0, good /* __NR_read */352 jeq #1, good /* __NR_write */353 jeq #5, good /* __NR_fstat */354 jeq #9, good /* __NR_mmap */355 jeq #14, good /* __NR_rt_sigprocmask */356 jeq #13, good /* __NR_rt_sigaction */357 jeq #35, good /* __NR_nanosleep */358 bad: ret #0 /* SECCOMP_RET_KILL_THREAD */359 good: ret #0x7fff0000 /* SECCOMP_RET_ALLOW */360 361Examples for low-level BPF extension:362 363**Packet for interface index 13**::364 365 ld ifidx366 jneq #13, drop367 ret #-1368 drop: ret #0369 370**(Accelerated) VLAN w/ id 10**::371 372 ld vlan_tci373 jneq #10, drop374 ret #-1375 drop: ret #0376 377The above example code can be placed into a file (here called "foo"), and378then be passed to the bpf_asm tool for generating opcodes, output that xt_bpf379and cls_bpf understands and can directly be loaded with. Example with above380ARP code::381 382 $ ./bpf_asm foo383 4,40 0 0 12,21 0 1 2054,6 0 0 4294967295,6 0 0 0,384 385In copy and paste C-like output::386 387 $ ./bpf_asm -c foo388 { 0x28, 0, 0, 0x0000000c },389 { 0x15, 0, 1, 0x00000806 },390 { 0x06, 0, 0, 0xffffffff },391 { 0x06, 0, 0, 0000000000 },392 393In particular, as usage with xt_bpf or cls_bpf can result in more complex BPF394filters that might not be obvious at first, it's good to test filters before395attaching to a live system. For that purpose, there's a small tool called396bpf_dbg under tools/bpf/ in the kernel source directory. This debugger allows397for testing BPF filters against given pcap files, single stepping through the398BPF code on the pcap's packets and to do BPF machine register dumps.399 400Starting bpf_dbg is trivial and just requires issuing::401 402 # ./bpf_dbg403 404In case input and output do not equal stdin/stdout, bpf_dbg takes an405alternative stdin source as a first argument, and an alternative stdout406sink as a second one, e.g. `./bpf_dbg test_in.txt test_out.txt`.407 408Other than that, a particular libreadline configuration can be set via409file "~/.bpf_dbg_init" and the command history is stored in the file410"~/.bpf_dbg_history".411 412Interaction in bpf_dbg happens through a shell that also has auto-completion413support (follow-up example commands starting with '>' denote bpf_dbg shell).414The usual workflow would be to ...415 416* load bpf 6,40 0 0 12,21 0 3 2048,48 0 0 23,21 0 1 1,6 0 0 65535,6 0 0 0417 Loads a BPF filter from standard output of bpf_asm, or transformed via418 e.g. ``tcpdump -iem1 -ddd port 22 | tr '\n' ','``. Note that for JIT419 debugging (next section), this command creates a temporary socket and420 loads the BPF code into the kernel. Thus, this will also be useful for421 JIT developers.422 423* load pcap foo.pcap424 425 Loads standard tcpdump pcap file.426 427* run [<n>]428 429bpf passes:1 fails:9430 Runs through all packets from a pcap to account how many passes and fails431 the filter will generate. A limit of packets to traverse can be given.432 433* disassemble::434 435 l0: ldh [12]436 l1: jeq #0x800, l2, l5437 l2: ldb [23]438 l3: jeq #0x1, l4, l5439 l4: ret #0xffff440 l5: ret #0441 442 Prints out BPF code disassembly.443 444* dump::445 446 /* { op, jt, jf, k }, */447 { 0x28, 0, 0, 0x0000000c },448 { 0x15, 0, 3, 0x00000800 },449 { 0x30, 0, 0, 0x00000017 },450 { 0x15, 0, 1, 0x00000001 },451 { 0x06, 0, 0, 0x0000ffff },452 { 0x06, 0, 0, 0000000000 },453 454 Prints out C-style BPF code dump.455 456* breakpoint 0::457 458 breakpoint at: l0: ldh [12]459 460* breakpoint 1::461 462 breakpoint at: l1: jeq #0x800, l2, l5463 464 ...465 466 Sets breakpoints at particular BPF instructions. Issuing a `run` command467 will walk through the pcap file continuing from the current packet and468 break when a breakpoint is being hit (another `run` will continue from469 the currently active breakpoint executing next instructions):470 471 * run::472 473 -- register dump --474 pc: [0] <-- program counter475 code: [40] jt[0] jf[0] k[12] <-- plain BPF code of current instruction476 curr: l0: ldh [12] <-- disassembly of current instruction477 A: [00000000][0] <-- content of A (hex, decimal)478 X: [00000000][0] <-- content of X (hex, decimal)479 M[0,15]: [00000000][0] <-- folded content of M (hex, decimal)480 -- packet dump -- <-- Current packet from pcap (hex)481 len: 42482 0: 00 19 cb 55 55 a4 00 14 a4 43 78 69 08 06 00 01483 16: 08 00 06 04 00 01 00 14 a4 43 78 69 0a 3b 01 26484 32: 00 00 00 00 00 00 0a 3b 01 01485 (breakpoint)486 >487 488 * breakpoint::489 490 breakpoints: 0 1491 492 Prints currently set breakpoints.493 494* step [-<n>, +<n>]495 496 Performs single stepping through the BPF program from the current pc497 offset. Thus, on each step invocation, above register dump is issued.498 This can go forwards and backwards in time, a plain `step` will break499 on the next BPF instruction, thus +1. (No `run` needs to be issued here.)500 501* select <n>502 503 Selects a given packet from the pcap file to continue from. Thus, on504 the next `run` or `step`, the BPF program is being evaluated against505 the user pre-selected packet. Numbering starts just as in Wireshark506 with index 1.507 508* quit509 510 Exits bpf_dbg.511 512JIT compiler513------------514 515The Linux kernel has a built-in BPF JIT compiler for x86_64, SPARC,516PowerPC, ARM, ARM64, MIPS, RISC-V, s390, and ARC and can be enabled through517CONFIG_BPF_JIT. The JIT compiler is transparently invoked for each518attached filter from user space or for internal kernel users if it has519been previously enabled by root::520 521 echo 1 > /proc/sys/net/core/bpf_jit_enable522 523For JIT developers, doing audits etc, each compile run can output the generated524opcode image into the kernel log via::525 526 echo 2 > /proc/sys/net/core/bpf_jit_enable527 528Example output from dmesg::529 530 [ 3389.935842] flen=6 proglen=70 pass=3 image=ffffffffa0069c8f531 [ 3389.935847] JIT code: 00000000: 55 48 89 e5 48 83 ec 60 48 89 5d f8 44 8b 4f 68532 [ 3389.935849] JIT code: 00000010: 44 2b 4f 6c 4c 8b 87 d8 00 00 00 be 0c 00 00 00533 [ 3389.935850] JIT code: 00000020: e8 1d 94 ff e0 3d 00 08 00 00 75 16 be 17 00 00534 [ 3389.935851] JIT code: 00000030: 00 e8 28 94 ff e0 83 f8 01 75 07 b8 ff ff 00 00535 [ 3389.935852] JIT code: 00000040: eb 02 31 c0 c9 c3536 537When CONFIG_BPF_JIT_ALWAYS_ON is enabled, bpf_jit_enable is permanently set to 1 and538setting any other value than that will return in failure. This is even the case for539setting bpf_jit_enable to 2, since dumping the final JIT image into the kernel log540is discouraged and introspection through bpftool (under tools/bpf/bpftool/) is the541generally recommended approach instead.542 543In the kernel source tree under tools/bpf/, there's bpf_jit_disasm for544generating disassembly out of the kernel log's hexdump::545 546 # ./bpf_jit_disasm547 70 bytes emitted from JIT compiler (pass:3, flen:6)548 ffffffffa0069c8f + <x>:549 0: push %rbp550 1: mov %rsp,%rbp551 4: sub $0x60,%rsp552 8: mov %rbx,-0x8(%rbp)553 c: mov 0x68(%rdi),%r9d554 10: sub 0x6c(%rdi),%r9d555 14: mov 0xd8(%rdi),%r8556 1b: mov $0xc,%esi557 20: callq 0xffffffffe0ff9442558 25: cmp $0x800,%eax559 2a: jne 0x0000000000000042560 2c: mov $0x17,%esi561 31: callq 0xffffffffe0ff945e562 36: cmp $0x1,%eax563 39: jne 0x0000000000000042564 3b: mov $0xffff,%eax565 40: jmp 0x0000000000000044566 42: xor %eax,%eax567 44: leaveq568 45: retq569 570 Issuing option `-o` will "annotate" opcodes to resulting assembler571 instructions, which can be very useful for JIT developers:572 573 # ./bpf_jit_disasm -o574 70 bytes emitted from JIT compiler (pass:3, flen:6)575 ffffffffa0069c8f + <x>:576 0: push %rbp577 55578 1: mov %rsp,%rbp579 48 89 e5580 4: sub $0x60,%rsp581 48 83 ec 60582 8: mov %rbx,-0x8(%rbp)583 48 89 5d f8584 c: mov 0x68(%rdi),%r9d585 44 8b 4f 68586 10: sub 0x6c(%rdi),%r9d587 44 2b 4f 6c588 14: mov 0xd8(%rdi),%r8589 4c 8b 87 d8 00 00 00590 1b: mov $0xc,%esi591 be 0c 00 00 00592 20: callq 0xffffffffe0ff9442593 e8 1d 94 ff e0594 25: cmp $0x800,%eax595 3d 00 08 00 00596 2a: jne 0x0000000000000042597 75 16598 2c: mov $0x17,%esi599 be 17 00 00 00600 31: callq 0xffffffffe0ff945e601 e8 28 94 ff e0602 36: cmp $0x1,%eax603 83 f8 01604 39: jne 0x0000000000000042605 75 07606 3b: mov $0xffff,%eax607 b8 ff ff 00 00608 40: jmp 0x0000000000000044609 eb 02610 42: xor %eax,%eax611 31 c0612 44: leaveq613 c9614 45: retq615 c3616 617For BPF JIT developers, bpf_jit_disasm, bpf_asm and bpf_dbg provides a useful618toolchain for developing and testing the kernel's JIT compiler.619 620BPF kernel internals621--------------------622Internally, for the kernel interpreter, a different instruction set623format with similar underlying principles from BPF described in previous624paragraphs is being used. However, the instruction set format is modelled625closer to the underlying architecture to mimic native instruction sets, so626that a better performance can be achieved (more details later). This new627ISA is called eBPF. See the ../bpf/index.rst for details. (Note: eBPF which628originates from [e]xtended BPF is not the same as BPF extensions! While629eBPF is an ISA, BPF extensions date back to classic BPF's 'overloading'630of BPF_LD | BPF_{B,H,W} | BPF_ABS instruction.)631 632The new instruction set was originally designed with the possible goal in633mind to write programs in "restricted C" and compile into eBPF with a optional634GCC/LLVM backend, so that it can just-in-time map to modern 64-bit CPUs with635minimal performance overhead over two steps, that is, C -> eBPF -> native code.636 637Currently, the new format is being used for running user BPF programs, which638includes seccomp BPF, classic socket filters, cls_bpf traffic classifier,639team driver's classifier for its load-balancing mode, netfilter's xt_bpf640extension, PTP dissector/classifier, and much more. They are all internally641converted by the kernel into the new instruction set representation and run642in the eBPF interpreter. For in-kernel handlers, this all works transparently643by using bpf_prog_create() for setting up the filter, resp.644bpf_prog_destroy() for destroying it. The function645bpf_prog_run(filter, ctx) transparently invokes eBPF interpreter or JITed646code to run the filter. 'filter' is a pointer to struct bpf_prog that we647got from bpf_prog_create(), and 'ctx' the given context (e.g.648skb pointer). All constraints and restrictions from bpf_check_classic() apply649before a conversion to the new layout is being done behind the scenes!650 651Currently, the classic BPF format is being used for JITing on most65232-bit architectures, whereas x86-64, aarch64, s390x, powerpc64,653sparc64, arm32, riscv64, riscv32, loongarch64, arc perform JIT compilation654from eBPF instruction set.655 656Testing657-------658 659Next to the BPF toolchain, the kernel also ships a test module that contains660various test cases for classic and eBPF that can be executed against661the BPF interpreter and JIT compiler. It can be found in lib/test_bpf.c and662enabled via Kconfig::663 664 CONFIG_TEST_BPF=m665 666After the module has been built and installed, the test suite can be executed667via insmod or modprobe against 'test_bpf' module. Results of the test cases668including timings in nsec can be found in the kernel log (dmesg).669 670Misc671----672 673Also trinity, the Linux syscall fuzzer, has built-in support for BPF and674SECCOMP-BPF kernel fuzzing.675 676Written by677----------678 679The document was written in the hope that it is found useful and in order680to give potential BPF hackers or security auditors a better overview of681the underlying architecture.682 683- Jay Schulist <jschlst@samba.org>684- Daniel Borkmann <daniel@iogearbox.net>685- Alexei Starovoitov <ast@kernel.org>686