Commit | Line | Data |
---|---|---|
5a8921ba DT |
1 | .. contents:: |
2 | .. sectnum:: | |
3 | ||
7d35eb1a DV |
4 | ======================================= |
5 | BPF Instruction Set Specification, v1.0 | |
6 | ======================================= | |
5a8921ba | 7 | |
7d35eb1a | 8 | This document specifies version 1.0 of the BPF instruction set. |
88691e9e | 9 | |
d00d5b82 DT |
10 | Documentation conventions |
11 | ========================= | |
12 | ||
2369e526 WH |
13 | For brevity and consistency, this document refers to families |
14 | of types using a shorthand syntax and refers to several expository, | |
15 | mnemonic functions when describing the semantics of instructions. | |
16 | The range of valid values for those types and the semantics of those | |
17 | functions are defined in the following subsections. | |
18 | ||
19 | Types | |
20 | ----- | |
21 | This document refers to integer types with the notation `SN` to specify | |
22 | a type's signedness (`S`) and bit width (`N`), respectively. | |
23 | ||
24 | .. table:: Meaning of signedness notation. | |
25 | ||
26 | ==== ========= | |
27 | `S` Meaning | |
28 | ==== ========= | |
29 | `u` unsigned | |
30 | `s` signed | |
31 | ==== ========= | |
32 | ||
33 | .. table:: Meaning of bit-width notation. | |
34 | ||
35 | ===== ========= | |
36 | `N` Bit width | |
37 | ===== ========= | |
38 | `8` 8 bits | |
39 | `16` 16 bits | |
40 | `32` 32 bits | |
41 | `64` 64 bits | |
42 | `128` 128 bits | |
43 | ===== ========= | |
44 | ||
45 | For example, `u32` is a type whose valid values are all the 32-bit unsigned | |
46 | numbers and `s16` is a types whose valid values are all the 16-bit signed | |
47 | numbers. | |
48 | ||
49 | Functions | |
50 | --------- | |
51 | * `htobe16`: Takes an unsigned 16-bit number in host-endian format and | |
52 | returns the equivalent number as an unsigned 16-bit number in big-endian | |
53 | format. | |
54 | * `htobe32`: Takes an unsigned 32-bit number in host-endian format and | |
55 | returns the equivalent number as an unsigned 32-bit number in big-endian | |
56 | format. | |
57 | * `htobe64`: Takes an unsigned 64-bit number in host-endian format and | |
58 | returns the equivalent number as an unsigned 64-bit number in big-endian | |
59 | format. | |
60 | * `htole16`: Takes an unsigned 16-bit number in host-endian format and | |
61 | returns the equivalent number as an unsigned 16-bit number in little-endian | |
62 | format. | |
63 | * `htole32`: Takes an unsigned 32-bit number in host-endian format and | |
64 | returns the equivalent number as an unsigned 32-bit number in little-endian | |
65 | format. | |
66 | * `htole64`: Takes an unsigned 64-bit number in host-endian format and | |
67 | returns the equivalent number as an unsigned 64-bit number in little-endian | |
68 | format. | |
69 | * `bswap16`: Takes an unsigned 16-bit number in either big- or little-endian | |
70 | format and returns the equivalent number with the same bit width but | |
71 | opposite endianness. | |
72 | * `bswap32`: Takes an unsigned 32-bit number in either big- or little-endian | |
73 | format and returns the equivalent number with the same bit width but | |
74 | opposite endianness. | |
75 | * `bswap64`: Takes an unsigned 64-bit number in either big- or little-endian | |
76 | format and returns the equivalent number with the same bit width but | |
77 | opposite endianness. | |
88691e9e | 78 | |
e546a119 WH |
79 | |
80 | Definitions | |
81 | ----------- | |
82 | ||
83 | .. glossary:: | |
84 | ||
85 | Sign Extend | |
86 | To `sign extend an` ``X`` `-bit number, A, to a` ``Y`` `-bit number, B ,` means to | |
87 | ||
88 | #. Copy all ``X`` bits from `A` to the lower ``X`` bits of `B`. | |
89 | #. Set the value of the remaining ``Y`` - ``X`` bits of `B` to the value of | |
90 | the most-significant bit of `A`. | |
91 | ||
92 | .. admonition:: Example | |
93 | ||
94 | Sign extend an 8-bit number ``A`` to a 16-bit number ``B`` on a big-endian platform: | |
95 | :: | |
96 | ||
97 | A: 10000110 | |
98 | B: 11111111 10000110 | |
99 | ||
81777efb DT |
100 | Conformance groups |
101 | ------------------ | |
102 | ||
103 | An implementation does not need to support all instructions specified in this | |
104 | document (e.g., deprecated instructions). Instead, a number of conformance | |
105 | groups are specified. An implementation must support the "basic" conformance | |
106 | group and may support additional conformance groups, where supporting a | |
107 | conformance group means it must support all instructions in that conformance | |
108 | group. | |
109 | ||
110 | The use of named conformance groups enables interoperability between a runtime | |
111 | that executes instructions, and tools as such compilers that generate | |
112 | instructions for the runtime. Thus, capability discovery in terms of | |
113 | conformance groups might be done manually by users or automatically by tools. | |
114 | ||
115 | Each conformance group has a short ASCII label (e.g., "basic") that | |
116 | corresponds to a set of instructions that are mandatory. That is, each | |
117 | instruction has one or more conformance groups of which it is a member. | |
118 | ||
119 | The "basic" conformance group includes all instructions defined in this | |
120 | specification unless otherwise noted. | |
121 | ||
62e46838 CH |
122 | Instruction encoding |
123 | ==================== | |
124 | ||
7d35eb1a | 125 | BPF has two instruction encodings: |
5ca15b8a | 126 | |
5a8921ba | 127 | * the basic instruction encoding, which uses 64 bits to encode an instruction |
a92adde8 DT |
128 | * the wide instruction encoding, which appends a second 64-bit immediate (i.e., |
129 | constant) value after the basic instruction for a total of 128 bits. | |
5ca15b8a | 130 | |
ae256f95 JM |
131 | The fields conforming an encoded basic instruction are stored in the |
132 | following order:: | |
62e46838 | 133 | |
ae256f95 JM |
134 | opcode:8 src_reg:4 dst_reg:4 offset:16 imm:32 // In little-endian BPF. |
135 | opcode:8 dst_reg:4 src_reg:4 offset:16 imm:32 // In big-endian BPF. | |
a92adde8 DT |
136 | |
137 | **imm** | |
138 | signed integer immediate value | |
139 | ||
140 | **offset** | |
141 | signed integer offset used with pointer arithmetic | |
142 | ||
143 | **src_reg** | |
144 | the source register number (0-10), except where otherwise specified | |
145 | (`64-bit immediate instructions`_ reuse this field for other purposes) | |
146 | ||
147 | **dst_reg** | |
148 | destination register number (0-10) | |
149 | ||
150 | **opcode** | |
151 | operation to perform | |
62e46838 | 152 | |
ae256f95 JM |
153 | Note that the contents of multi-byte fields ('imm' and 'offset') are |
154 | stored using big-endian byte ordering in big-endian BPF and | |
155 | little-endian byte ordering in little-endian BPF. | |
746ce767 | 156 | |
ae256f95 | 157 | For example:: |
746ce767 | 158 | |
ae256f95 JM |
159 | opcode offset imm assembly |
160 | src_reg dst_reg | |
161 | 07 0 1 00 00 44 33 22 11 r1 += 0x11223344 // little | |
162 | dst_reg src_reg | |
163 | 07 1 0 00 00 11 22 33 44 r1 += 0x11223344 // big | |
746ce767 | 164 | |
62e46838 CH |
165 | Note that most instructions do not use all of the fields. |
166 | Unused fields shall be cleared to zero. | |
167 | ||
a92adde8 DT |
168 | As discussed below in `64-bit immediate instructions`_, a 64-bit immediate |
169 | instruction uses a 64-bit immediate value that is constructed as follows. | |
170 | The 64 bits following the basic instruction contain a pseudo instruction | |
171 | using the same format but with opcode, dst_reg, src_reg, and offset all set to zero, | |
172 | and imm containing the high 32 bits of the immediate value. | |
173 | ||
ae256f95 JM |
174 | This is depicted in the following figure:: |
175 | ||
176 | basic_instruction | |
e48f0f4a DT |
177 | .------------------------------. |
178 | | | | |
179 | opcode:8 regs:8 offset:16 imm:32 unused:32 imm:32 | |
180 | | | | |
181 | '--------------' | |
182 | pseudo instruction | |
a92adde8 DT |
183 | |
184 | Thus the 64-bit immediate value is constructed as follows: | |
185 | ||
186 | imm64 = (next_imm << 32) | imm | |
187 | ||
188 | where 'next_imm' refers to the imm value of the pseudo instruction | |
ae256f95 JM |
189 | following the basic instruction. The unused bytes in the pseudo |
190 | instruction are reserved and shall be cleared to zero. | |
a92adde8 | 191 | |
5e4dd19f | 192 | Instruction classes |
62e46838 | 193 | ------------------- |
88691e9e | 194 | |
5e4dd19f | 195 | The three LSB bits of the 'opcode' field store the instruction class: |
88691e9e | 196 | |
5a8921ba DT |
197 | ========= ===== =============================== =================================== |
198 | class value description reference | |
199 | ========= ===== =============================== =================================== | |
200 | BPF_LD 0x00 non-standard load operations `Load and store instructions`_ | |
201 | BPF_LDX 0x01 load into register operations `Load and store instructions`_ | |
202 | BPF_ST 0x02 store from immediate operations `Load and store instructions`_ | |
203 | BPF_STX 0x03 store from register operations `Load and store instructions`_ | |
204 | BPF_ALU 0x04 32-bit arithmetic operations `Arithmetic and jump instructions`_ | |
205 | BPF_JMP 0x05 64-bit jump operations `Arithmetic and jump instructions`_ | |
206 | BPF_JMP32 0x06 32-bit jump operations `Arithmetic and jump instructions`_ | |
207 | BPF_ALU64 0x07 64-bit arithmetic operations `Arithmetic and jump instructions`_ | |
208 | ========= ===== =============================== =================================== | |
88691e9e | 209 | |
5e4dd19f CH |
210 | Arithmetic and jump instructions |
211 | ================================ | |
212 | ||
5a8921ba DT |
213 | For arithmetic and jump instructions (``BPF_ALU``, ``BPF_ALU64``, ``BPF_JMP`` and |
214 | ``BPF_JMP32``), the 8-bit 'opcode' field is divided into three parts: | |
88691e9e | 215 | |
5a8921ba DT |
216 | ============== ====== ================= |
217 | 4 bits (MSB) 1 bit 3 bits (LSB) | |
218 | ============== ====== ================= | |
a92adde8 | 219 | code source instruction class |
5a8921ba | 220 | ============== ====== ================= |
88691e9e | 221 | |
a92adde8 DT |
222 | **code** |
223 | the operation code, whose meaning varies by instruction class | |
88691e9e | 224 | |
a92adde8 DT |
225 | **source** |
226 | the source operand location, which unless otherwise specified is one of: | |
88691e9e | 227 | |
a92adde8 DT |
228 | ====== ===== ============================================== |
229 | source value description | |
230 | ====== ===== ============================================== | |
231 | BPF_K 0x00 use 32-bit 'imm' value as source operand | |
232 | BPF_X 0x08 use 'src_reg' register value as source operand | |
233 | ====== ===== ============================================== | |
88691e9e | 234 | |
a92adde8 DT |
235 | **instruction class** |
236 | the instruction class (see `Instruction classes`_) | |
be3193cd CH |
237 | |
238 | Arithmetic instructions | |
239 | ----------------------- | |
240 | ||
5a8921ba | 241 | ``BPF_ALU`` uses 32-bit wide operands while ``BPF_ALU64`` uses 64-bit wide operands for |
be3193cd | 242 | otherwise identical operations. |
a92adde8 DT |
243 | The 'code' field encodes the operation as below, where 'src' and 'dst' refer |
244 | to the values of the source and destination registers, respectively. | |
5a8921ba | 245 | |
fb213ecb YS |
246 | ========= ===== ======= ========================================================== |
247 | code value offset description | |
248 | ========= ===== ======= ========================================================== | |
249 | BPF_ADD 0x00 0 dst += src | |
250 | BPF_SUB 0x10 0 dst -= src | |
251 | BPF_MUL 0x20 0 dst \*= src | |
252 | BPF_DIV 0x30 0 dst = (src != 0) ? (dst / src) : 0 | |
253 | BPF_SDIV 0x30 1 dst = (src != 0) ? (dst s/ src) : 0 | |
254 | BPF_OR 0x40 0 dst \|= src | |
255 | BPF_AND 0x50 0 dst &= src | |
256 | BPF_LSH 0x60 0 dst <<= (src & mask) | |
257 | BPF_RSH 0x70 0 dst >>= (src & mask) | |
258 | BPF_NEG 0x80 0 dst = -dst | |
259 | BPF_MOD 0x90 0 dst = (src != 0) ? (dst % src) : dst | |
260 | BPF_SMOD 0x90 1 dst = (src != 0) ? (dst s% src) : dst | |
261 | BPF_XOR 0xa0 0 dst ^= src | |
262 | BPF_MOV 0xb0 0 dst = src | |
263 | BPF_MOVSX 0xb0 8/16/32 dst = (s8,s16,s32)src | |
e546a119 | 264 | BPF_ARSH 0xc0 0 :term:`sign extending<Sign Extend>` dst >>= (src & mask) |
fb213ecb YS |
265 | BPF_END 0xd0 0 byte swap operations (see `Byte swap instructions`_ below) |
266 | ========= ===== ======= ========================================================== | |
5a8921ba | 267 | |
0eb9d19e | 268 | Underflow and overflow are allowed during arithmetic operations, meaning |
7d35eb1a | 269 | the 64-bit or 32-bit value will wrap. If BPF program execution would |
0eb9d19e DT |
270 | result in division by zero, the destination register is instead set to zero. |
271 | If execution would result in modulo by zero, for ``BPF_ALU64`` the value of | |
272 | the destination register is unchanged whereas for ``BPF_ALU`` the upper | |
273 | 32 bits of the destination register are zeroed. | |
274 | ||
5a8921ba | 275 | ``BPF_ADD | BPF_X | BPF_ALU`` means:: |
be3193cd | 276 | |
a92adde8 | 277 | dst = (u32) ((u32) dst + (u32) src) |
be3193cd | 278 | |
d00d5b82 DT |
279 | where '(u32)' indicates that the upper 32 bits are zeroed. |
280 | ||
5a8921ba | 281 | ``BPF_ADD | BPF_X | BPF_ALU64`` means:: |
be3193cd | 282 | |
a92adde8 | 283 | dst = dst + src |
be3193cd | 284 | |
5a8921ba | 285 | ``BPF_XOR | BPF_K | BPF_ALU`` means:: |
be3193cd | 286 | |
a92adde8 | 287 | dst = (u32) dst ^ (u32) imm32 |
be3193cd | 288 | |
5a8921ba | 289 | ``BPF_XOR | BPF_K | BPF_ALU64`` means:: |
be3193cd | 290 | |
a92adde8 | 291 | dst = dst ^ imm32 |
be3193cd | 292 | |
ee932bf9 YS |
293 | Note that most instructions have instruction offset of 0. Only three instructions |
294 | (``BPF_SDIV``, ``BPF_SMOD``, ``BPF_MOVSX``) have a non-zero offset. | |
245d4c40 | 295 | |
e546a119 | 296 | The division and modulo operations support both unsigned and signed flavors. |
245d4c40 | 297 | |
ee932bf9 YS |
298 | For unsigned operations (``BPF_DIV`` and ``BPF_MOD``), for ``BPF_ALU``, |
299 | 'imm' is interpreted as a 32-bit unsigned value. For ``BPF_ALU64``, | |
e546a119 WH |
300 | 'imm' is first :term:`sign extended<Sign Extend>` from 32 to 64 bits, and then |
301 | interpreted as a 64-bit unsigned value. | |
ee932bf9 YS |
302 | |
303 | For signed operations (``BPF_SDIV`` and ``BPF_SMOD``), for ``BPF_ALU``, | |
304 | 'imm' is interpreted as a 32-bit signed value. For ``BPF_ALU64``, 'imm' | |
e546a119 WH |
305 | is first :term:`sign extended<Sign Extend>` from 32 to 64 bits, and then |
306 | interpreted as a 64-bit signed value. | |
ee932bf9 | 307 | |
0e133a13 DT |
308 | Note that there are varying definitions of the signed modulo operation |
309 | when the dividend or divisor are negative, where implementations often | |
310 | vary by language such that Python, Ruby, etc. differ from C, Go, Java, | |
311 | etc. This specification requires that signed modulo use truncated division | |
312 | (where -13 % 3 == -1) as implemented in C, Go, etc.: | |
313 | ||
314 | a % n = a - n * trunc(a / n) | |
315 | ||
ee932bf9 | 316 | The ``BPF_MOVSX`` instruction does a move operation with sign extension. |
e546a119 | 317 | ``BPF_ALU | BPF_MOVSX`` :term:`sign extends<Sign Extend>` 8-bit and 16-bit operands into 32 |
ee932bf9 | 318 | bit operands, and zeroes the remaining upper 32 bits. |
e546a119 | 319 | ``BPF_ALU64 | BPF_MOVSX`` :term:`sign extends<Sign Extend>` 8-bit, 16-bit, and 32-bit |
20e109ea DT |
320 | operands into 64 bit operands. Unlike other arithmetic instructions, |
321 | ``BPF_MOVSX`` is only defined for register source operands (``BPF_X``). | |
be3193cd | 322 | |
e48f0f4a DT |
323 | The ``BPF_NEG`` instruction is only defined when the source bit is clear |
324 | (``BPF_K``). | |
325 | ||
8819495a DT |
326 | Shift operations use a mask of 0x3F (63) for 64-bit operations and 0x1F (31) |
327 | for 32-bit operations. | |
328 | ||
dd33fb57 | 329 | Byte swap instructions |
ee932bf9 | 330 | ---------------------- |
dd33fb57 | 331 | |
245d4c40 YS |
332 | The byte swap instructions use instruction classes of ``BPF_ALU`` and ``BPF_ALU64`` |
333 | and a 4-bit 'code' field of ``BPF_END``. | |
dd33fb57 | 334 | |
67b97e58 | 335 | The byte swap instructions operate on the destination register |
dd33fb57 CH |
336 | only and do not use a separate source register or immediate value. |
337 | ||
ee932bf9 YS |
338 | For ``BPF_ALU``, the 1-bit source operand field in the opcode is used to |
339 | select what byte order the operation converts from or to. For | |
340 | ``BPF_ALU64``, the 1-bit source operand field in the opcode is reserved | |
341 | and must be set to 0. | |
dd33fb57 | 342 | |
245d4c40 YS |
343 | ========= ========= ===== ================================================= |
344 | class source value description | |
345 | ========= ========= ===== ================================================= | |
346 | BPF_ALU BPF_TO_LE 0x00 convert between host byte order and little endian | |
347 | BPF_ALU BPF_TO_BE 0x08 convert between host byte order and big endian | |
ee932bf9 | 348 | BPF_ALU64 Reserved 0x00 do byte swap unconditionally |
245d4c40 | 349 | ========= ========= ===== ================================================= |
dd33fb57 | 350 | |
5a8921ba | 351 | The 'imm' field encodes the width of the swap operations. The following widths |
dd33fb57 CH |
352 | are supported: 16, 32 and 64. |
353 | ||
354 | Examples: | |
355 | ||
2369e526 | 356 | ``BPF_ALU | BPF_TO_LE | BPF_END`` with imm = 16/32/64 means:: |
dd33fb57 | 357 | |
a92adde8 | 358 | dst = htole16(dst) |
2369e526 WH |
359 | dst = htole32(dst) |
360 | dst = htole64(dst) | |
dd33fb57 | 361 | |
2369e526 | 362 | ``BPF_ALU | BPF_TO_BE | BPF_END`` with imm = 16/32/64 means:: |
dd33fb57 | 363 | |
2369e526 WH |
364 | dst = htobe16(dst) |
365 | dst = htobe32(dst) | |
a92adde8 | 366 | dst = htobe64(dst) |
dd33fb57 | 367 | |
245d4c40 YS |
368 | ``BPF_ALU64 | BPF_TO_LE | BPF_END`` with imm = 16/32/64 means:: |
369 | ||
2369e526 WH |
370 | dst = bswap16(dst) |
371 | dst = bswap32(dst) | |
372 | dst = bswap64(dst) | |
245d4c40 | 373 | |
be3193cd CH |
374 | Jump instructions |
375 | ----------------- | |
376 | ||
5a8921ba | 377 | ``BPF_JMP32`` uses 32-bit wide operands while ``BPF_JMP`` uses 64-bit wide operands for |
be3193cd | 378 | otherwise identical operations. |
5a8921ba DT |
379 | The 'code' field encodes the operation as below: |
380 | ||
e48f0f4a DT |
381 | ======== ===== === =============================== ============================================= |
382 | code value src description notes | |
383 | ======== ===== === =============================== ============================================= | |
384 | BPF_JA 0x0 0x0 PC += offset BPF_JMP | BPF_K only | |
385 | BPF_JA 0x0 0x0 PC += imm BPF_JMP32 | BPF_K only | |
8cfee110 | 386 | BPF_JEQ 0x1 any PC += offset if dst == src |
e48f0f4a DT |
387 | BPF_JGT 0x2 any PC += offset if dst > src unsigned |
388 | BPF_JGE 0x3 any PC += offset if dst >= src unsigned | |
8cfee110 DT |
389 | BPF_JSET 0x4 any PC += offset if dst & src |
390 | BPF_JNE 0x5 any PC += offset if dst != src | |
e48f0f4a DT |
391 | BPF_JSGT 0x6 any PC += offset if dst > src signed |
392 | BPF_JSGE 0x7 any PC += offset if dst >= src signed | |
393 | BPF_CALL 0x8 0x0 call helper function by address BPF_JMP | BPF_K only, see `Helper functions`_ | |
394 | BPF_CALL 0x8 0x1 call PC += imm BPF_JMP | BPF_K only, see `Program-local functions`_ | |
395 | BPF_CALL 0x8 0x2 call helper function by BTF ID BPF_JMP | BPF_K only, see `Helper functions`_ | |
396 | BPF_EXIT 0x9 0x0 return BPF_JMP | BPF_K only | |
397 | BPF_JLT 0xa any PC += offset if dst < src unsigned | |
398 | BPF_JLE 0xb any PC += offset if dst <= src unsigned | |
399 | BPF_JSLT 0xc any PC += offset if dst < src signed | |
400 | BPF_JSLE 0xd any PC += offset if dst <= src signed | |
401 | ======== ===== === =============================== ============================================= | |
41db511a | 402 | |
7d35eb1a | 403 | The BPF program needs to store the return value into register R0 before doing a |
8cfee110 | 404 | ``BPF_EXIT``. |
88691e9e | 405 | |
b9fe8e8d DT |
406 | Example: |
407 | ||
408 | ``BPF_JSGE | BPF_X | BPF_JMP32`` (0x7e) means:: | |
409 | ||
410 | if (s32)dst s>= (s32)src goto +offset | |
411 | ||
412 | where 's>=' indicates a signed '>=' comparison. | |
413 | ||
245d4c40 YS |
414 | ``BPF_JA | BPF_K | BPF_JMP32`` (0x06) means:: |
415 | ||
416 | gotol +imm | |
417 | ||
418 | where 'imm' means the branch offset comes from insn 'imm' field. | |
419 | ||
ee932bf9 YS |
420 | Note that there are two flavors of ``BPF_JA`` instructions. The |
421 | ``BPF_JMP`` class permits a 16-bit jump offset specified by the 'offset' | |
422 | field, whereas the ``BPF_JMP32`` class permits a 32-bit jump offset | |
423 | specified by the 'imm' field. A > 16-bit conditional jump may be | |
424 | converted to a < 16-bit conditional jump plus a 32-bit unconditional | |
425 | jump. | |
245d4c40 | 426 | |
c1f9e14e DT |
427 | Helper functions |
428 | ~~~~~~~~~~~~~~~~ | |
429 | ||
430 | Helper functions are a concept whereby BPF programs can call into a | |
8cfee110 DT |
431 | set of function calls exposed by the underlying platform. |
432 | ||
433 | Historically, each helper function was identified by an address | |
434 | encoded in the imm field. The available helper functions may differ | |
435 | for each program type, but address values are unique across all program types. | |
436 | ||
437 | Platforms that support the BPF Type Format (BTF) support identifying | |
438 | a helper function by a BTF ID encoded in the imm field, where the BTF ID | |
439 | identifies the helper name and type. | |
440 | ||
441 | Program-local functions | |
442 | ~~~~~~~~~~~~~~~~~~~~~~~ | |
443 | Program-local functions are functions exposed by the same BPF program as the | |
444 | caller, and are referenced by offset from the call instruction, similar to | |
2d71a90f WH |
445 | ``BPF_JA``. The offset is encoded in the imm field of the call instruction. |
446 | A ``BPF_EXIT`` within the program-local function will return to the caller. | |
88691e9e | 447 | |
5e4dd19f CH |
448 | Load and store instructions |
449 | =========================== | |
450 | ||
5a8921ba | 451 | For load and store instructions (``BPF_LD``, ``BPF_LDX``, ``BPF_ST``, and ``BPF_STX``), the |
5e4dd19f CH |
452 | 8-bit 'opcode' field is divided as: |
453 | ||
5a8921ba DT |
454 | ============ ====== ================= |
455 | 3 bits (MSB) 2 bits 3 bits (LSB) | |
456 | ============ ====== ================= | |
457 | mode size instruction class | |
458 | ============ ====== ================= | |
459 | ||
460 | The mode modifier is one of: | |
461 | ||
462 | ============= ===== ==================================== ============= | |
463 | mode modifier value description reference | |
464 | ============= ===== ==================================== ============= | |
465 | BPF_IMM 0x00 64-bit immediate instructions `64-bit immediate instructions`_ | |
466 | BPF_ABS 0x20 legacy BPF packet access (absolute) `Legacy BPF Packet access instructions`_ | |
467 | BPF_IND 0x40 legacy BPF packet access (indirect) `Legacy BPF Packet access instructions`_ | |
468 | BPF_MEM 0x60 regular load and store operations `Regular load and store operations`_ | |
245d4c40 | 469 | BPF_MEMSX 0x80 sign-extension load operations `Sign-extension load operations`_ |
5a8921ba DT |
470 | BPF_ATOMIC 0xc0 atomic operations `Atomic operations`_ |
471 | ============= ===== ==================================== ============= | |
5e4dd19f CH |
472 | |
473 | The size modifier is one of: | |
88691e9e | 474 | |
5e4dd19f CH |
475 | ============= ===== ===================== |
476 | size modifier value description | |
477 | ============= ===== ===================== | |
478 | BPF_W 0x00 word (4 bytes) | |
479 | BPF_H 0x08 half word (2 bytes) | |
480 | BPF_B 0x10 byte | |
481 | BPF_DW 0x18 double word (8 bytes) | |
482 | ============= ===== ===================== | |
88691e9e | 483 | |
63d8c242 CH |
484 | Regular load and store operations |
485 | --------------------------------- | |
486 | ||
487 | The ``BPF_MEM`` mode modifier is used to encode regular load and store | |
488 | instructions that transfer data between a register and memory. | |
489 | ||
490 | ``BPF_MEM | <size> | BPF_STX`` means:: | |
88691e9e | 491 | |
a92adde8 | 492 | *(size *) (dst + offset) = src |
88691e9e | 493 | |
63d8c242 | 494 | ``BPF_MEM | <size> | BPF_ST`` means:: |
88691e9e | 495 | |
a92adde8 | 496 | *(size *) (dst + offset) = imm32 |
5e4dd19f | 497 | |
63d8c242 | 498 | ``BPF_MEM | <size> | BPF_LDX`` means:: |
5e4dd19f | 499 | |
245d4c40 YS |
500 | dst = *(unsigned size *) (src + offset) |
501 | ||
502 | Where size is one of: ``BPF_B``, ``BPF_H``, ``BPF_W``, or ``BPF_DW`` and | |
ee932bf9 | 503 | 'unsigned size' is one of u8, u16, u32 or u64. |
245d4c40 | 504 | |
fb213ecb YS |
505 | Sign-extension load operations |
506 | ------------------------------ | |
507 | ||
e546a119 | 508 | The ``BPF_MEMSX`` mode modifier is used to encode :term:`sign-extension<Sign Extend>` load |
245d4c40 YS |
509 | instructions that transfer data between a register and memory. |
510 | ||
511 | ``BPF_MEMSX | <size> | BPF_LDX`` means:: | |
512 | ||
513 | dst = *(signed size *) (src + offset) | |
5e4dd19f | 514 | |
245d4c40 | 515 | Where size is one of: ``BPF_B``, ``BPF_H`` or ``BPF_W``, and |
ee932bf9 | 516 | 'signed size' is one of s8, s16 or s32. |
5e4dd19f | 517 | |
5e4dd19f CH |
518 | Atomic operations |
519 | ----------------- | |
88691e9e | 520 | |
594d3234 CH |
521 | Atomic operations are operations that operate on memory and can not be |
522 | interrupted or corrupted by other access to the same memory region | |
7d35eb1a | 523 | by other BPF programs or means outside of this specification. |
88691e9e | 524 | |
7d35eb1a | 525 | All atomic operations supported by BPF are encoded as store operations |
594d3234 | 526 | that use the ``BPF_ATOMIC`` mode modifier as follows: |
88691e9e | 527 | |
5a8921ba DT |
528 | * ``BPF_ATOMIC | BPF_W | BPF_STX`` for 32-bit operations |
529 | * ``BPF_ATOMIC | BPF_DW | BPF_STX`` for 64-bit operations | |
530 | * 8-bit and 16-bit wide atomic operations are not supported. | |
88691e9e | 531 | |
5a8921ba | 532 | The 'imm' field is used to encode the actual atomic operation. |
594d3234 | 533 | Simple atomic operation use a subset of the values defined to encode |
5a8921ba | 534 | arithmetic operations in the 'imm' field to encode the atomic operation: |
88691e9e | 535 | |
5a8921ba DT |
536 | ======== ===== =========== |
537 | imm value description | |
538 | ======== ===== =========== | |
539 | BPF_ADD 0x00 atomic add | |
540 | BPF_OR 0x40 atomic or | |
541 | BPF_AND 0x50 atomic and | |
542 | BPF_XOR 0xa0 atomic xor | |
543 | ======== ===== =========== | |
88691e9e | 544 | |
88691e9e | 545 | |
5a8921ba | 546 | ``BPF_ATOMIC | BPF_W | BPF_STX`` with 'imm' = BPF_ADD means:: |
88691e9e | 547 | |
a92adde8 | 548 | *(u32 *)(dst + offset) += src |
88691e9e | 549 | |
5a8921ba | 550 | ``BPF_ATOMIC | BPF_DW | BPF_STX`` with 'imm' = BPF ADD means:: |
88691e9e | 551 | |
a92adde8 | 552 | *(u64 *)(dst + offset) += src |
88691e9e | 553 | |
594d3234 CH |
554 | In addition to the simple atomic operations, there also is a modifier and |
555 | two complex atomic operations: | |
556 | ||
5a8921ba DT |
557 | =========== ================ =========================== |
558 | imm value description | |
559 | =========== ================ =========================== | |
560 | BPF_FETCH 0x01 modifier: return old value | |
561 | BPF_XCHG 0xe0 | BPF_FETCH atomic exchange | |
562 | BPF_CMPXCHG 0xf0 | BPF_FETCH atomic compare and exchange | |
563 | =========== ================ =========================== | |
594d3234 CH |
564 | |
565 | The ``BPF_FETCH`` modifier is optional for simple atomic operations, and | |
566 | always set for the complex atomic operations. If the ``BPF_FETCH`` flag | |
a92adde8 | 567 | is set, then the operation also overwrites ``src`` with the value that |
594d3234 CH |
568 | was in memory before it was modified. |
569 | ||
a92adde8 DT |
570 | The ``BPF_XCHG`` operation atomically exchanges ``src`` with the value |
571 | addressed by ``dst + offset``. | |
594d3234 CH |
572 | |
573 | The ``BPF_CMPXCHG`` operation atomically compares the value addressed by | |
a92adde8 DT |
574 | ``dst + offset`` with ``R0``. If they match, the value addressed by |
575 | ``dst + offset`` is replaced with ``src``. In either case, the | |
576 | value that was at ``dst + offset`` before the operation is zero-extended | |
594d3234 | 577 | and loaded back to ``R0``. |
88691e9e | 578 | |
5ca15b8a CH |
579 | 64-bit immediate instructions |
580 | ----------------------------- | |
581 | ||
5a8921ba | 582 | Instructions with the ``BPF_IMM`` 'mode' modifier use the wide instruction |
16b7c970 DT |
583 | encoding defined in `Instruction encoding`_, and use the 'src' field of the |
584 | basic instruction to hold an opcode subtype. | |
585 | ||
586 | The following table defines a set of ``BPF_IMM | BPF_DW | BPF_LD`` instructions | |
587 | with opcode subtypes in the 'src' field, using new terms such as "map" | |
588 | defined further below: | |
589 | ||
590 | ========================= ====== === ========================================= =========== ============== | |
591 | opcode construction opcode src pseudocode imm type dst type | |
592 | ========================= ====== === ========================================= =========== ============== | |
593 | BPF_IMM | BPF_DW | BPF_LD 0x18 0x0 dst = imm64 integer integer | |
594 | BPF_IMM | BPF_DW | BPF_LD 0x18 0x1 dst = map_by_fd(imm) map fd map | |
595 | BPF_IMM | BPF_DW | BPF_LD 0x18 0x2 dst = map_val(map_by_fd(imm)) + next_imm map fd data pointer | |
596 | BPF_IMM | BPF_DW | BPF_LD 0x18 0x3 dst = var_addr(imm) variable id data pointer | |
597 | BPF_IMM | BPF_DW | BPF_LD 0x18 0x4 dst = code_addr(imm) integer code pointer | |
598 | BPF_IMM | BPF_DW | BPF_LD 0x18 0x5 dst = map_by_idx(imm) map index map | |
599 | BPF_IMM | BPF_DW | BPF_LD 0x18 0x6 dst = map_val(map_by_idx(imm)) + next_imm map index data pointer | |
600 | ========================= ====== === ========================================= =========== ============== | |
601 | ||
602 | where | |
603 | ||
604 | * map_by_fd(imm) means to convert a 32-bit file descriptor into an address of a map (see `Maps`_) | |
605 | * map_by_idx(imm) means to convert a 32-bit index into an address of a map | |
606 | * map_val(map) gets the address of the first value in a given map | |
607 | * var_addr(imm) gets the address of a platform variable (see `Platform Variables`_) with a given id | |
608 | * code_addr(imm) gets the address of the instruction at a specified relative offset in number of (64-bit) instructions | |
609 | * the 'imm type' can be used by disassemblers for display | |
610 | * the 'dst type' can be used for verification and JIT compilation purposes | |
611 | ||
612 | Maps | |
613 | ~~~~ | |
614 | ||
7d35eb1a | 615 | Maps are shared memory regions accessible by BPF programs on some platforms. |
16b7c970 DT |
616 | A map can have various semantics as defined in a separate document, and may or |
617 | may not have a single contiguous memory region, but the 'map_val(map)' is | |
618 | currently only defined for maps that do have a single contiguous memory region. | |
619 | ||
620 | Each map can have a file descriptor (fd) if supported by the platform, where | |
621 | 'map_by_fd(imm)' means to get the map with the specified file descriptor. Each | |
622 | BPF program can also be defined to use a set of maps associated with the | |
623 | program at load time, and 'map_by_idx(imm)' means to get the map with the given | |
624 | index in the set associated with the BPF program containing the instruction. | |
625 | ||
626 | Platform Variables | |
627 | ~~~~~~~~~~~~~~~~~~ | |
628 | ||
629 | Platform variables are memory regions, identified by integer ids, exposed by | |
630 | the runtime and accessible by BPF programs on some platforms. The | |
631 | 'var_addr(imm)' operation means to get the address of the memory region | |
632 | identified by the given id. | |
63d000c3 | 633 | |
15175336 CH |
634 | Legacy BPF Packet access instructions |
635 | ------------------------------------- | |
63d000c3 | 636 | |
7d35eb1a | 637 | BPF previously introduced special instructions for access to packet data that were |
6166da0a | 638 | carried over from classic BPF. However, these instructions are |
81777efb DT |
639 | deprecated and should no longer be used. All legacy packet access |
640 | instructions belong to the "legacy" conformance group instead of the "basic" | |
641 | conformance group. |