1 =====================
2 BPF Type Format (BTF)
3 =====================
4
5 1. Introduction
6 ===============
7
8 BTF (BPF Type Format) is the metadata format which encodes the debug info
9 related to BPF program/map. The name BTF was used initially to describe data
10 types. The BTF was later extended to include function info for defined
11 subroutines, and line info for source/line information.
12
13 The debug info is used for map pretty print, function signature, etc. The
14 function signature enables better bpf program/function kernel symbol. The line
15 info helps generate source annotated translated byte code, jited code and
16 verifier log.
17
18 The BTF specification contains two parts,
19 * BTF kernel API
20 * BTF ELF file format
21
22 The kernel API is the contract between user space and kernel. The kernel
23 verifies the BTF info before using it. The ELF file format is a user space
24 contract between ELF file and libbpf loader.
25
26 The type and string sections are part of the BTF kernel API, describing the
27 debug info (mostly types related) referenced by the bpf program. These two
28 sections are discussed in details in :ref:`BTF_Type_String`.
29
30 .. _BTF_Type_String:
31
32 2. BTF Type and String Encoding
33 ===============================
34
35 The file ``include/uapi/linux/btf.h`` provides high-level definition of how
36 types/strings are encoded.
37
38 The beginning of data blob must be::
39
40 struct btf_header {
41 __u16 magic;
42 __u8 version;
43 __u8 flags;
44 __u32 hdr_len;
45
46 /* All offsets are in bytes relative to the end of this header */
47 __u32 type_off; /* offset of type section */
48 __u32 type_len; /* length of type section */
49 __u32 str_off; /* offset of string section */
50 __u32 str_len; /* length of string section */
51 };
52
53 The magic is ``0xeB9F``, which has different encoding for big and little
54 endian systems, and can be used to test whether BTF is generated for big- or
55 little-endian target. The ``btf_header`` is designed to be extensible with
56 ``hdr_len`` equal to ``sizeof(struct btf_header)`` when a data blob is
57 generated.
58
59 2.1 String Encoding
60 -------------------
61
62 The first string in the string section must be a null string. The rest of
63 string table is a concatenation of other null-terminated strings.
64
65 2.2 Type Encoding
66 -----------------
67
68 The type id ``0`` is reserved for ``void`` type. The type section is parsed
69 sequentially and type id is assigned to each recognized type starting from id
70 ``1``. Currently, the following types are supported::
71
72 #define BTF_KIND_INT 1 /* Integer */
73 #define BTF_KIND_PTR 2 /* Pointer */
74 #define BTF_KIND_ARRAY 3 /* Array */
75 #define BTF_KIND_STRUCT 4 /* Struct */
76 #define BTF_KIND_UNION 5 /* Union */
77 #define BTF_KIND_ENUM 6 /* Enumeration up to 32-bit values */
78 #define BTF_KIND_FWD 7 /* Forward */
79 #define BTF_KIND_TYPEDEF 8 /* Typedef */
80 #define BTF_KIND_VOLATILE 9 /* Volatile */
81 #define BTF_KIND_CONST 10 /* Const */
82 #define BTF_KIND_RESTRICT 11 /* Restrict */
83 #define BTF_KIND_FUNC 12 /* Function */
84 #define BTF_KIND_FUNC_PROTO 13 /* Function Proto */
85 #define BTF_KIND_VAR 14 /* Variable */
86 #define BTF_KIND_DATASEC 15 /* Section */
87 #define BTF_KIND_FLOAT 16 /* Floating point */
88 #define BTF_KIND_DECL_TAG 17 /* Decl Tag */
89 #define BTF_KIND_TYPE_TAG 18 /* Type Tag */
90 #define BTF_KIND_ENUM64 19 /* Enumeration up to 64-bit values */
91
92 Note that the type section encodes debug info, not just pure types.
93 ``BTF_KIND_FUNC`` is not a type, and it represents a defined subprogram.
94
95 Each type contains the following common data::
96
97 struct btf_type {
98 __u32 name_off;
99 /* "info" bits arrangement
100 * bits 0-15: vlen (e.g. # of struct's members)
101 * bits 16-23: unused
102 * bits 24-28: kind (e.g. int, ptr, array...etc)
103 * bits 29-30: unused
104 * bit 31: kind_flag, currently used by
105 * struct, union, fwd, enum and enum64.
106 */
107 __u32 info;
108 /* "size" is used by INT, ENUM, STRUCT, UNION and ENUM64.
109 * "size" tells the size of the type it is describing.
110 *
111 * "type" is used by PTR, TYPEDEF, VOLATILE, CONST, RESTRICT,
112 * FUNC, FUNC_PROTO, DECL_TAG and TYPE_TAG.
113 * "type" is a type_id referring to another type.
114 */
115 union {
116 __u32 size;
117 __u32 type;
118 };
119 };
120
121 For certain kinds, the common data are followed by kind-specific data. The
122 ``name_off`` in ``struct btf_type`` specifies the offset in the string table.
123 The following sections detail encoding of each kind.
124
125 2.2.1 BTF_KIND_INT
126 ~~~~~~~~~~~~~~~~~~
127
128 ``struct btf_type`` encoding requirement:
129 * ``name_off``: any valid offset
130 * ``info.kind_flag``: 0
131 * ``info.kind``: BTF_KIND_INT
132 * ``info.vlen``: 0
133 * ``size``: the size of the int type in bytes.
134
135 ``btf_type`` is followed by a ``u32`` with the following bits arrangement::
136
137 #define BTF_INT_ENCODING(VAL) (((VAL) & 0x0f000000) >> 24)
138 #define BTF_INT_OFFSET(VAL) (((VAL) & 0x00ff0000) >> 16)
139 #define BTF_INT_BITS(VAL) ((VAL) & 0x000000ff)
140
141 The ``BTF_INT_ENCODING`` has the following attributes::
142
143 #define BTF_INT_SIGNED (1 << 0)
144 #define BTF_INT_CHAR (1 << 1)
145 #define BTF_INT_BOOL (1 << 2)
146
147 The ``BTF_INT_ENCODING()`` provides extra information: signedness, char, or
148 bool, for the int type. The char and bool encoding are mostly useful for
149 pretty print. At most one encoding can be specified for the int type.
150
151 The ``BTF_INT_BITS()`` specifies the number of actual bits held by this int
152 type. For example, a 4-bit bitfield encodes ``BTF_INT_BITS()`` equals to 4.
153 The ``btf_type.size * 8`` must be equal to or greater than ``BTF_INT_BITS()``
154 for the type. The maximum value of ``BTF_INT_BITS()`` is 128.
155
156 The ``BTF_INT_OFFSET()`` specifies the starting bit offset to calculate values
157 for this int. For example, a bitfield struct member has:
158
159 * btf member bit offset 100 from the start of the structure,
160 * btf member pointing to an int type,
161 * the int type has ``BTF_INT_OFFSET() = 2`` and ``BTF_INT_BITS() = 4``
162
163 Then in the struct memory layout, this member will occupy ``4`` bits starting
164 from bits ``100 + 2 = 102``.
165
166 Alternatively, the bitfield struct member can be the following to access the
167 same bits as the above:
168
169 * btf member bit offset 102,
170 * btf member pointing to an int type,
171 * the int type has ``BTF_INT_OFFSET() = 0`` and ``BTF_INT_BITS() = 4``
172
173 The original intention of ``BTF_INT_OFFSET()`` is to provide flexibility of
174 bitfield encoding. Currently, both llvm and pahole generate
175 ``BTF_INT_OFFSET() = 0`` for all int types.
176
177 2.2.2 BTF_KIND_PTR
178 ~~~~~~~~~~~~~~~~~~
179
180 ``struct btf_type`` encoding requirement:
181 * ``name_off``: 0
182 * ``info.kind_flag``: 0
183 * ``info.kind``: BTF_KIND_PTR
184 * ``info.vlen``: 0
185 * ``type``: the pointee type of the pointer
186
187 No additional type data follow ``btf_type``.
188
189 2.2.3 BTF_KIND_ARRAY
190 ~~~~~~~~~~~~~~~~~~~~
191
192 ``struct btf_type`` encoding requirement:
193 * ``name_off``: 0
194 * ``info.kind_flag``: 0
195 * ``info.kind``: BTF_KIND_ARRAY
196 * ``info.vlen``: 0
197 * ``size/type``: 0, not used
198
199 ``btf_type`` is followed by one ``struct btf_array``::
200
201 struct btf_array {
202 __u32 type;
203 __u32 index_type;
204 __u32 nelems;
205 };
206
207 The ``struct btf_array`` encoding:
208 * ``type``: the element type
209 * ``index_type``: the index type
210 * ``nelems``: the number of elements for this array (``0`` is also allowed).
211
212 The ``index_type`` can be any regular int type (``u8``, ``u16``, ``u32``,
213 ``u64``, ``unsigned __int128``). The original design of including
214 ``index_type`` follows DWARF, which has an ``index_type`` for its array type.
215 Currently in BTF, beyond type verification, the ``index_type`` is not used.
216
217 The ``struct btf_array`` allows chaining through element type to represent
218 multidimensional arrays. For example, for ``int a[5][6]``, the following type
219 information illustrates the chaining:
220
221 * [1]: int
222 * [2]: array, ``btf_array.type = [1]``, ``btf_array.nelems = 6``
223 * [3]: array, ``btf_array.type = [2]``, ``btf_array.nelems = 5``
224
225 Currently, both pahole and llvm collapse multidimensional array into
226 one-dimensional array, e.g., for ``a[5][6]``, the ``btf_array.nelems`` is
227 equal to ``30``. This is because the original use case is map pretty print
228 where the whole array is dumped out so one-dimensional array is enough. As
229 more BTF usage is explored, pahole and llvm can be changed to generate proper
230 chained representation for multidimensional arrays.
231
232 2.2.4 BTF_KIND_STRUCT
233 ~~~~~~~~~~~~~~~~~~~~~
234 2.2.5 BTF_KIND_UNION
235 ~~~~~~~~~~~~~~~~~~~~
236
237 ``struct btf_type`` encoding requirement:
238 * ``name_off``: 0 or offset to a valid C identifier
239 * ``info.kind_flag``: 0 or 1
240 * ``info.kind``: BTF_KIND_STRUCT or BTF_KIND_UNION
241 * ``info.vlen``: the number of struct/union members
242 * ``info.size``: the size of the struct/union in bytes
243
244 ``btf_type`` is followed by ``info.vlen`` number of ``struct btf_member``.::
245
246 struct btf_member {
247 __u32 name_off;
248 __u32 type;
249 __u32 offset;
250 };
251
252 ``struct btf_member`` encoding:
253 * ``name_off``: offset to a valid C identifier
254 * ``type``: the member type
255 * ``offset``: <see below>
256
257 If the type info ``kind_flag`` is not set, the offset contains only bit offset
258 of the member. Note that the base type of the bitfield can only be int or enum
259 type. If the bitfield size is 32, the base type can be either int or enum
260 type. If the bitfield size is not 32, the base type must be int, and int type
261 ``BTF_INT_BITS()`` encodes the bitfield size.
262
263 If the ``kind_flag`` is set, the ``btf_member.offset`` contains both member
264 bitfield size and bit offset. The bitfield size and bit offset are calculated
265 as below.::
266
267 #define BTF_MEMBER_BITFIELD_SIZE(val) ((val) >> 24)
268 #define BTF_MEMBER_BIT_OFFSET(val) ((val) & 0xffffff)
269
270 In this case, if the base type is an int type, it must be a regular int type:
271
272 * ``BTF_INT_OFFSET()`` must be 0.
273 * ``BTF_INT_BITS()`` must be equal to ``{1,2,4,8,16} * 8``.
274
275 Commit 9d5f9f701b18 introduced ``kind_flag`` and explains why both modes
276 exist.
277
278 2.2.6 BTF_KIND_ENUM
279 ~~~~~~~~~~~~~~~~~~~
280
281 ``struct btf_type`` encoding requirement:
282 * ``name_off``: 0 or offset to a valid C identifier
283 * ``info.kind_flag``: 0 for unsigned, 1 for signed
284 * ``info.kind``: BTF_KIND_ENUM
285 * ``info.vlen``: number of enum values
286 * ``size``: 1/2/4/8
287
288 ``btf_type`` is followed by ``info.vlen`` number of ``struct btf_enum``.::
289
290 struct btf_enum {
291 __u32 name_off;
292 __s32 val;
293 };
294
295 The ``btf_enum`` encoding:
296 * ``name_off``: offset to a valid C identifier
297 * ``val``: any value
298
299 If the original enum value is signed and the size is less than 4,
300 that value will be sign extended into 4 bytes. If the size is 8,
301 the value will be truncated into 4 bytes.
302
303 2.2.7 BTF_KIND_FWD
304 ~~~~~~~~~~~~~~~~~~
305
306 ``struct btf_type`` encoding requirement:
307 * ``name_off``: offset to a valid C identifier
308 * ``info.kind_flag``: 0 for struct, 1 for union
309 * ``info.kind``: BTF_KIND_FWD
310 * ``info.vlen``: 0
311 * ``type``: 0
312
313 No additional type data follow ``btf_type``.
314
315 2.2.8 BTF_KIND_TYPEDEF
316 ~~~~~~~~~~~~~~~~~~~~~~
317
318 ``struct btf_type`` encoding requirement:
319 * ``name_off``: offset to a valid C identifier
320 * ``info.kind_flag``: 0
321 * ``info.kind``: BTF_KIND_TYPEDEF
322 * ``info.vlen``: 0
323 * ``type``: the type which can be referred by name at ``name_off``
324
325 No additional type data follow ``btf_type``.
326
327 2.2.9 BTF_KIND_VOLATILE
328 ~~~~~~~~~~~~~~~~~~~~~~~
329
330 ``struct btf_type`` encoding requirement:
331 * ``name_off``: 0
332 * ``info.kind_flag``: 0
333 * ``info.kind``: BTF_KIND_VOLATILE
334 * ``info.vlen``: 0
335 * ``type``: the type with ``volatile`` qualifier
336
337 No additional type data follow ``btf_type``.
338
339 2.2.10 BTF_KIND_CONST
340 ~~~~~~~~~~~~~~~~~~~~~
341
342 ``struct btf_type`` encoding requirement:
343 * ``name_off``: 0
344 * ``info.kind_flag``: 0
345 * ``info.kind``: BTF_KIND_CONST
346 * ``info.vlen``: 0
347 * ``type``: the type with ``const`` qualifier
348
349 No additional type data follow ``btf_type``.
350
351 2.2.11 BTF_KIND_RESTRICT
352 ~~~~~~~~~~~~~~~~~~~~~~~~
353
354 ``struct btf_type`` encoding requirement:
355 * ``name_off``: 0
356 * ``info.kind_flag``: 0
357 * ``info.kind``: BTF_KIND_RESTRICT
358 * ``info.vlen``: 0
359 * ``type``: the type with ``restrict`` qualifier
360
361 No additional type data follow ``btf_type``.
362
363 2.2.12 BTF_KIND_FUNC
364 ~~~~~~~~~~~~~~~~~~~~
365
366 ``struct btf_type`` encoding requirement:
367 * ``name_off``: offset to a valid C identifier
368 * ``info.kind_flag``: 0
369 * ``info.kind``: BTF_KIND_FUNC
370 * ``info.vlen``: linkage information (BTF_FUNC_STATIC, BTF_FUNC_GLOBAL
371 or BTF_FUNC_EXTERN - see :ref:`BTF_Function_Linkage_Constants`)
372 * ``type``: a BTF_KIND_FUNC_PROTO type
373
374 No additional type data follow ``btf_type``.
375
376 A BTF_KIND_FUNC defines not a type, but a subprogram (function) whose
377 signature is defined by ``type``. The subprogram is thus an instance of that
378 type. The BTF_KIND_FUNC may in turn be referenced by a func_info in the
379 :ref:`BTF_Ext_Section` (ELF) or in the arguments to :ref:`BPF_Prog_Load`
380 (ABI).
381
382 Currently, only linkage values of BTF_FUNC_STATIC and BTF_FUNC_GLOBAL are
383 supported in the kernel.
384
385 2.2.13 BTF_KIND_FUNC_PROTO
386 ~~~~~~~~~~~~~~~~~~~~~~~~~~
387
388 ``struct btf_type`` encoding requirement:
389 * ``name_off``: 0
390 * ``info.kind_flag``: 0
391 * ``info.kind``: BTF_KIND_FUNC_PROTO
392 * ``info.vlen``: # of parameters
393 * ``type``: the return type
394
395 ``btf_type`` is followed by ``info.vlen`` number of ``struct btf_param``.::
396
397 struct btf_param {
398 __u32 name_off;
399 __u32 type;
400 };
401
402 If a BTF_KIND_FUNC_PROTO type is referred by a BTF_KIND_FUNC type, then
403 ``btf_param.name_off`` must point to a valid C identifier except for the
404 possible last argument representing the variable argument. The btf_param.type
405 refers to parameter type.
406
407 If the function has variable arguments, the last parameter is encoded with
408 ``name_off = 0`` and ``type = 0``.
409
410 2.2.14 BTF_KIND_VAR
411 ~~~~~~~~~~~~~~~~~~~
412
413 ``struct btf_type`` encoding requirement:
414 * ``name_off``: offset to a valid C identifier
415 * ``info.kind_flag``: 0
416 * ``info.kind``: BTF_KIND_VAR
417 * ``info.vlen``: 0
418 * ``type``: the type of the variable
419
420 ``btf_type`` is followed by a single ``struct btf_variable`` with the
421 following data::
422
423 struct btf_var {
424 __u32 linkage;
425 };
426
427 ``btf_var.linkage`` may take the values: BTF_VAR_STATIC, BTF_VAR_GLOBAL_ALLOCATED or BTF_VAR_GLOBAL_EXTERN -
428 see :ref:`BTF_Var_Linkage_Constants`.
429
430 Not all type of global variables are supported by LLVM at this point.
431 The following is currently available:
432
433 * static variables with or without section attributes
434 * global variables with section attributes
435
436 The latter is for future extraction of map key/value type id's from a
437 map definition.
438
439 2.2.15 BTF_KIND_DATASEC
440 ~~~~~~~~~~~~~~~~~~~~~~~
441
442 ``struct btf_type`` encoding requirement:
443 * ``name_off``: offset to a valid name associated with a variable or
444 one of .data/.bss/.rodata
445 * ``info.kind_flag``: 0
446 * ``info.kind``: BTF_KIND_DATASEC
447 * ``info.vlen``: # of variables
448 * ``size``: total section size in bytes (0 at compilation time, patched
449 to actual size by BPF loaders such as libbpf)
450
451 ``btf_type`` is followed by ``info.vlen`` number of ``struct btf_var_secinfo``.::
452
453 struct btf_var_secinfo {
454 __u32 type;
455 __u32 offset;
456 __u32 size;
457 };
458
459 ``struct btf_var_secinfo`` encoding:
460 * ``type``: the type of the BTF_KIND_VAR variable
461 * ``offset``: the in-section offset of the variable
462 * ``size``: the size of the variable in bytes
463
464 2.2.16 BTF_KIND_FLOAT
465 ~~~~~~~~~~~~~~~~~~~~~
466
467 ``struct btf_type`` encoding requirement:
468 * ``name_off``: any valid offset
469 * ``info.kind_flag``: 0
470 * ``info.kind``: BTF_KIND_FLOAT
471 * ``info.vlen``: 0
472 * ``size``: the size of the float type in bytes: 2, 4, 8, 12 or 16.
473
474 No additional type data follow ``btf_type``.
475
476 2.2.17 BTF_KIND_DECL_TAG
477 ~~~~~~~~~~~~~~~~~~~~~~~~
478
479 ``struct btf_type`` encoding requirement:
480 * ``name_off``: offset to a non-empty string
481 * ``info.kind_flag``: 0
482 * ``info.kind``: BTF_KIND_DECL_TAG
483 * ``info.vlen``: 0
484 * ``type``: ``struct``, ``union``, ``func``, ``var`` or ``typedef``
485
486 ``btf_type`` is followed by ``struct btf_decl_tag``.::
487
488 struct btf_decl_tag {
489 __u32 component_idx;
490 };
491
492 The ``name_off`` encodes btf_decl_tag attribute string.
493 The ``type`` should be ``struct``, ``union``, ``func``, ``var`` or ``typedef``.
494 For ``var`` or ``typedef`` type, ``btf_decl_tag.component_idx`` must be ``-1``.
495 For the other three types, if the btf_decl_tag attribute is
496 applied to the ``struct``, ``union`` or ``func`` itself,
497 ``btf_decl_tag.component_idx`` must be ``-1``. Otherwise,
498 the attribute is applied to a ``struct``/``union`` member or
499 a ``func`` argument, and ``btf_decl_tag.component_idx`` should be a
500 valid index (starting from 0) pointing to a member or an argument.
501
502 2.2.18 BTF_KIND_TYPE_TAG
503 ~~~~~~~~~~~~~~~~~~~~~~~~
504
505 ``struct btf_type`` encoding requirement:
506 * ``name_off``: offset to a non-empty string
507 * ``info.kind_flag``: 0
508 * ``info.kind``: BTF_KIND_TYPE_TAG
509 * ``info.vlen``: 0
510 * ``type``: the type with ``btf_type_tag`` attribute
511
512 Currently, ``BTF_KIND_TYPE_TAG`` is only emitted for pointer types.
513 It has the following btf type chain:
514 ::
515
516 ptr -> [type_tag]*
517 -> [const | volatile | restrict | typedef]*
518 -> base_type
519
520 Basically, a pointer type points to zero or more
521 type_tag, then zero or more const/volatile/restrict/typedef
522 and finally the base type. The base type is one of
523 int, ptr, array, struct, union, enum, func_proto and float types.
524
525 2.2.19 BTF_KIND_ENUM64
526 ~~~~~~~~~~~~~~~~~~~~~~
527
528 ``struct btf_type`` encoding requirement:
529 * ``name_off``: 0 or offset to a valid C identifier
530 * ``info.kind_flag``: 0 for unsigned, 1 for signed
531 * ``info.kind``: BTF_KIND_ENUM64
532 * ``info.vlen``: number of enum values
533 * ``size``: 1/2/4/8
534
535 ``btf_type`` is followed by ``info.vlen`` number of ``struct btf_enum64``.::
536
537 struct btf_enum64 {
538 __u32 name_off;
539 __u32 val_lo32;
540 __u32 val_hi32;
541 };
542
543 The ``btf_enum64`` encoding:
544 * ``name_off``: offset to a valid C identifier
545 * ``val_lo32``: lower 32-bit value for a 64-bit value
546 * ``val_hi32``: high 32-bit value for a 64-bit value
547
548 If the original enum value is signed and the size is less than 8,
549 that value will be sign extended into 8 bytes.
550
551 2.3 Constant Values
552 -------------------
553
554 .. _BTF_Function_Linkage_Constants:
555
556 2.3.1 Function Linkage Constant Values
557 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
558 .. table:: Function Linkage Values and Meanings
559
560 =================== ===== ===========
561 kind value description
562 =================== ===== ===========
563 ``BTF_FUNC_STATIC`` 0x0 definition of subprogram not visible outside containing compilation unit
564 ``BTF_FUNC_GLOBAL`` 0x1 definition of subprogram visible outside containing compilation unit
565 ``BTF_FUNC_EXTERN`` 0x2 declaration of a subprogram whose definition is outside the containing compilation unit
566 =================== ===== ===========
567
568
569 .. _BTF_Var_Linkage_Constants:
570
571 2.3.2 Variable Linkage Constant Values
572 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
573 .. table:: Variable Linkage Values and Meanings
574
575 ============================ ===== ===========
576 kind value description
577 ============================ ===== ===========
578 ``BTF_VAR_STATIC`` 0x0 definition of global variable not visible outside containing compilation unit
579 ``BTF_VAR_GLOBAL_ALLOCATED`` 0x1 definition of global variable visible outside containing compilation unit
580 ``BTF_VAR_GLOBAL_EXTERN`` 0x2 declaration of global variable whose definition is outside the containing compilation unit
581 ============================ ===== ===========
582
583 3. BTF Kernel API
584 =================
585
586 The following bpf syscall command involves BTF:
587 * BPF_BTF_LOAD: load a blob of BTF data into kernel
588 * BPF_MAP_CREATE: map creation with btf key and value type info.
589 * BPF_PROG_LOAD: prog load with btf function and line info.
590 * BPF_BTF_GET_FD_BY_ID: get a btf fd
591 * BPF_OBJ_GET_INFO_BY_FD: btf, func_info, line_info
592 and other btf related info are returned.
593
594 The workflow typically looks like:
595 ::
596
597 Application:
598 BPF_BTF_LOAD
599 |
600 v
601 BPF_MAP_CREATE and BPF_PROG_LOAD
602 |
603 V
604 ......
605
606 Introspection tool:
607 ......
608 BPF_{PROG,MAP}_GET_NEXT_ID (get prog/map id's)
609 |
610 V
611 BPF_{PROG,MAP}_GET_FD_BY_ID (get a prog/map fd)
612 |
613 V
614 BPF_OBJ_GET_INFO_BY_FD (get bpf_prog_info/bpf_map_info with btf_id)
615 | |
616 V |
617 BPF_BTF_GET_FD_BY_ID (get btf_fd) |
618 | |
619 V |
620 BPF_OBJ_GET_INFO_BY_FD (get btf) |
621 | |
622 V V
623 pretty print types, dump func signatures and line info, etc.
624
625
626 3.1 BPF_BTF_LOAD
627 ----------------
628
629 Load a blob of BTF data into kernel. A blob of data, described in
630 :ref:`BTF_Type_String`, can be directly loaded into the kernel. A ``btf_fd``
631 is returned to a userspace.
632
633 3.2 BPF_MAP_CREATE
634 ------------------
635
636 A map can be created with ``btf_fd`` and specified key/value type id.::
637
638 __u32 btf_fd; /* fd pointing to a BTF type data */
639 __u32 btf_key_type_id; /* BTF type_id of the key */
640 __u32 btf_value_type_id; /* BTF type_id of the value */
641
642 In libbpf, the map can be defined with extra annotation like below:
643 ::
644
645 struct {
646 __uint(type, BPF_MAP_TYPE_ARRAY);
647 __type(key, int);
648 __type(value, struct ipv_counts);
649 __uint(max_entries, 4);
650 } btf_map SEC(".maps");
651
652 During ELF parsing, libbpf is able to extract key/value type_id's and assign
653 them to BPF_MAP_CREATE attributes automatically.
654
655 .. _BPF_Prog_Load:
656
657 3.3 BPF_PROG_LOAD
658 -----------------
659
660 During prog_load, func_info and line_info can be passed to kernel with proper
661 values for the following attributes:
662 ::
663
664 __u32 insn_cnt;
665 __aligned_u64 insns;
666 ......
667 __u32 prog_btf_fd; /* fd pointing to BTF type data */
668 __u32 func_info_rec_size; /* userspace bpf_func_info size */
669 __aligned_u64 func_info; /* func info */
670 __u32 func_info_cnt; /* number of bpf_func_info records */
671 __u32 line_info_rec_size; /* userspace bpf_line_info size */
672 __aligned_u64 line_info; /* line info */
673 __u32 line_info_cnt; /* number of bpf_line_info records */
674
675 The func_info and line_info are an array of below, respectively.::
676
677 struct bpf_func_info {
678 __u32 insn_off; /* [0, insn_cnt - 1] */
679 __u32 type_id; /* pointing to a BTF_KIND_FUNC type */
680 };
681 struct bpf_line_info {
682 __u32 insn_off; /* [0, insn_cnt - 1] */
683 __u32 file_name_off; /* offset to string table for the filename */
684 __u32 line_off; /* offset to string table for the source line */
685 __u32 line_col; /* line number and column number */
686 };
687
688 func_info_rec_size is the size of each func_info record, and
689 line_info_rec_size is the size of each line_info record. Passing the record
690 size to kernel make it possible to extend the record itself in the future.
691
692 Below are requirements for func_info:
693 * func_info[0].insn_off must be 0.
694 * the func_info insn_off is in strictly increasing order and matches
695 bpf func boundaries.
696
697 Below are requirements for line_info:
698 * the first insn in each func must have a line_info record pointing to it.
699 * the line_info insn_off is in strictly increasing order.
700
701 For line_info, the line number and column number are defined as below:
702 ::
703
704 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
705 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
706
707 3.4 BPF_{PROG,MAP}_GET_NEXT_ID
708 ------------------------------
709
710 In kernel, every loaded program, map or btf has a unique id. The id won't
711 change during the lifetime of a program, map, or btf.
712
713 The bpf syscall command BPF_{PROG,MAP}_GET_NEXT_ID returns all id's, one for
714 each command, to user space, for bpf program or maps, respectively, so an
715 inspection tool can inspect all programs and maps.
716
717 3.5 BPF_{PROG,MAP}_GET_FD_BY_ID
718 -------------------------------
719
720 An introspection tool cannot use id to get details about program or maps.
721 A file descriptor needs to be obtained first for reference-counting purpose.
722
723 3.6 BPF_OBJ_GET_INFO_BY_FD
724 --------------------------
725
726 Once a program/map fd is acquired, an introspection tool can get the detailed
727 information from kernel about this fd, some of which are BTF-related. For
728 example, ``bpf_map_info`` returns ``btf_id`` and key/value type ids.
729 ``bpf_prog_info`` returns ``btf_id``, func_info, and line info for translated
730 bpf byte codes, and jited_line_info.
731
732 3.7 BPF_BTF_GET_FD_BY_ID
733 ------------------------
734
735 With ``btf_id`` obtained in ``bpf_map_info`` and ``bpf_prog_info``, bpf
736 syscall command BPF_BTF_GET_FD_BY_ID can retrieve a btf fd. Then, with
737 command BPF_OBJ_GET_INFO_BY_FD, the btf blob, originally loaded into the
738 kernel with BPF_BTF_LOAD, can be retrieved.
739
740 With the btf blob, ``bpf_map_info``, and ``bpf_prog_info``, an introspection
741 tool has full btf knowledge and is able to pretty print map key/values, dump
742 func signatures and line info, along with byte/jit codes.
743
744 4. ELF File Format Interface
745 ============================
746
747 4.1 .BTF section
748 ----------------
749
750 The .BTF section contains type and string data. The format of this section is
751 same as the one describe in :ref:`BTF_Type_String`.
752
753 .. _BTF_Ext_Section:
754
755 4.2 .BTF.ext section
756 --------------------
757
758 The .BTF.ext section encodes func_info, line_info and CO-RE relocations
759 which needs loader manipulation before loading into the kernel.
760
761 The specification for .BTF.ext section is defined at ``tools/lib/bpf/btf.h``
762 and ``tools/lib/bpf/btf.c``.
763
764 The current header of .BTF.ext section::
765
766 struct btf_ext_header {
767 __u16 magic;
768 __u8 version;
769 __u8 flags;
770 __u32 hdr_len;
771
772 /* All offsets are in bytes relative to the end of this header */
773 __u32 func_info_off;
774 __u32 func_info_len;
775 __u32 line_info_off;
776 __u32 line_info_len;
777
778 /* optional part of .BTF.ext header */
779 __u32 core_relo_off;
780 __u32 core_relo_len;
781 };
782
783 It is very similar to .BTF section. Instead of type/string section, it
784 contains func_info, line_info and core_relo sub-sections.
785 See :ref:`BPF_Prog_Load` for details about func_info and line_info
786 record format.
787
788 The func_info is organized as below.::
789
790 func_info_rec_size /* __u32 value */
791 btf_ext_info_sec for section #1 /* func_info for section #1 */
792 btf_ext_info_sec for section #2 /* func_info for section #2 */
793 ...
794
795 ``func_info_rec_size`` specifies the size of ``bpf_func_info`` structure when
796 .BTF.ext is generated. ``btf_ext_info_sec``, defined below, is a collection of
797 func_info for each specific ELF section.::
798
799 struct btf_ext_info_sec {
800 __u32 sec_name_off; /* offset to section name */
801 __u32 num_info;
802 /* Followed by num_info * record_size number of bytes */
803 __u8 data[0];
804 };
805
806 Here, num_info must be greater than 0.
807
808 The line_info is organized as below.::
809
810 line_info_rec_size /* __u32 value */
811 btf_ext_info_sec for section #1 /* line_info for section #1 */
812 btf_ext_info_sec for section #2 /* line_info for section #2 */
813 ...
814
815 ``line_info_rec_size`` specifies the size of ``bpf_line_info`` structure when
816 .BTF.ext is generated.
817
818 The interpretation of ``bpf_func_info->insn_off`` and
819 ``bpf_line_info->insn_off`` is different between kernel API and ELF API. For
820 kernel API, the ``insn_off`` is the instruction offset in the unit of ``struct
821 bpf_insn``. For ELF API, the ``insn_off`` is the byte offset from the
822 beginning of section (``btf_ext_info_sec->sec_name_off``).
823
824 The core_relo is organized as below.::
825
826 core_relo_rec_size /* __u32 value */
827 btf_ext_info_sec for section #1 /* core_relo for section #1 */
828 btf_ext_info_sec for section #2 /* core_relo for section #2 */
829
830 ``core_relo_rec_size`` specifies the size of ``bpf_core_relo``
831 structure when .BTF.ext is generated. All ``bpf_core_relo`` structures
832 within a single ``btf_ext_info_sec`` describe relocations applied to
833 section named by ``btf_ext_info_sec->sec_name_off``.
834
835 See :ref:`Documentation/bpf/llvm_reloc.rst <btf-co-re-relocations>`
836 for more information on CO-RE relocations.
837
838 4.2 .BTF_ids section
839 --------------------
840
841 The .BTF_ids section encodes BTF ID values that are used within the kernel.
842
843 This section is created during the kernel compilation with the help of
844 macros defined in ``include/linux/btf_ids.h`` header file. Kernel code can
845 use them to create lists and sets (sorted lists) of BTF ID values.
846
847 The ``BTF_ID_LIST`` and ``BTF_ID`` macros define unsorted list of BTF ID values,
848 with following syntax::
849
850 BTF_ID_LIST(list)
851 BTF_ID(type1, name1)
852 BTF_ID(type2, name2)
853
854 resulting in following layout in .BTF_ids section::
855
856 __BTF_ID__type1__name1__1:
857 .zero 4
858 __BTF_ID__type2__name2__2:
859 .zero 4
860
861 The ``u32 list[];`` variable is defined to access the list.
862
863 The ``BTF_ID_UNUSED`` macro defines 4 zero bytes. It's used when we
864 want to define unused entry in BTF_ID_LIST, like::
865
866 BTF_ID_LIST(bpf_skb_output_btf_ids)
867 BTF_ID(struct, sk_buff)
868 BTF_ID_UNUSED
869 BTF_ID(struct, task_struct)
870
871 The ``BTF_SET_START/END`` macros pair defines sorted list of BTF ID values
872 and their count, with following syntax::
873
874 BTF_SET_START(set)
875 BTF_ID(type1, name1)
876 BTF_ID(type2, name2)
877 BTF_SET_END(set)
878
879 resulting in following layout in .BTF_ids section::
880
881 __BTF_ID__set__set:
882 .zero 4
883 __BTF_ID__type1__name1__3:
884 .zero 4
885 __BTF_ID__type2__name2__4:
886 .zero 4
887
888 The ``struct btf_id_set set;`` variable is defined to access the list.
889
890 The ``typeX`` name can be one of following::
891
892 struct, union, typedef, func
893
894 and is used as a filter when resolving the BTF ID value.
895
896 All the BTF ID lists and sets are compiled in the .BTF_ids section and
897 resolved during the linking phase of kernel build by ``resolve_btfids`` tool.
898
899 5. Using BTF
900 ============
901
902 5.1 bpftool map pretty print
903 ----------------------------
904
905 With BTF, the map key/value can be printed based on fields rather than simply
906 raw bytes. This is especially valuable for large structure or if your data
907 structure has bitfields. For example, for the following map,::
908
909 enum A { A1, A2, A3, A4, A5 };
910 typedef enum A ___A;
911 struct tmp_t {
912 char a1:4;
913 int a2:4;
914 int :4;
915 __u32 a3:4;
916 int b;
917 ___A b1:4;
918 enum A b2:4;
919 };
920 struct {
921 __uint(type, BPF_MAP_TYPE_ARRAY);
922 __type(key, int);
923 __type(value, struct tmp_t);
924 __uint(max_entries, 1);
925 } tmpmap SEC(".maps");
926
927 bpftool is able to pretty print like below:
928 ::
929
930 [{
931 "key": 0,
932 "value": {
933 "a1": 0x2,
934 "a2": 0x4,
935 "a3": 0x6,
936 "b": 7,
937 "b1": 0x8,
938 "b2": 0xa
939 }
940 }
941 ]
942
943 5.2 bpftool prog dump
944 ---------------------
945
946 The following is an example showing how func_info and line_info can help prog
947 dump with better kernel symbol names, function prototypes and line
948 information.::
949
950 $ bpftool prog dump jited pinned /sys/fs/bpf/test_btf_haskv
951 [...]
952 int test_long_fname_2(struct dummy_tracepoint_args * arg):
953 bpf_prog_44a040bf25481309_test_long_fname_2:
954 ; static int test_long_fname_2(struct dummy_tracepoint_args *arg)
955 0: push %rbp
956 1: mov %rsp,%rbp
957 4: sub $0x30,%rsp
958 b: sub $0x28,%rbp
959 f: mov %rbx,0x0(%rbp)
960 13: mov %r13,0x8(%rbp)
961 17: mov %r14,0x10(%rbp)
962 1b: mov %r15,0x18(%rbp)
963 1f: xor %eax,%eax
964 21: mov %rax,0x20(%rbp)
965 25: xor %esi,%esi
966 ; int key = 0;
967 27: mov %esi,-0x4(%rbp)
968 ; if (!arg->sock)
969 2a: mov 0x8(%rdi),%rdi
970 ; if (!arg->sock)
971 2e: cmp $0x0,%rdi
972 32: je 0x0000000000000070
973 34: mov %rbp,%rsi
974 ; counts = bpf_map_lookup_elem(&btf_map, &key);
975 [...]
976
977 5.3 Verifier Log
978 ----------------
979
980 The following is an example of how line_info can help debugging verification
981 failure.::
982
983 /* The code at tools/testing/selftests/bpf/test_xdp_noinline.c
984 * is modified as below.
985 */
986 data = (void *)(long)xdp->data;
987 data_end = (void *)(long)xdp->data_end;
988 /*
989 if (data + 4 > data_end)
990 return XDP_DROP;
991 */
992 *(u32 *)data = dst->dst;
993
994 $ bpftool prog load ./test_xdp_noinline.o /sys/fs/bpf/test_xdp_noinline type xdp
995 ; data = (void *)(long)xdp->data;
996 224: (79) r2 = *(u64 *)(r10 -112)
997 225: (61) r2 = *(u32 *)(r2 +0)
998 ; *(u32 *)data = dst->dst;
999 226: (63) *(u32 *)(r2 +0) = r1
1000 invalid access to packet, off=0 size=4, R2(id=0,off=0,r=0)
1001 R2 offset is outside of the packet
1002
1003 6. BTF Generation
1004 =================
1005
1006 You need latest pahole
1007
1008 https://git.kernel.org/pub/scm/devel/pahole/pahole.git/
1009
1010 or llvm (8.0 or later). The pahole acts as a dwarf2btf converter. It doesn't
1011 support .BTF.ext and btf BTF_KIND_FUNC type yet. For example,::
1012
1013 -bash-4.4$ cat t.c
1014 struct t {
1015 int a:2;
1016 int b:3;
1017 int c:2;
1018 } g;
1019 -bash-4.4$ gcc -c -O2 -g t.c
1020 -bash-4.4$ pahole -JV t.o
1021 File t.o:
1022 [1] STRUCT t kind_flag=1 size=4 vlen=3
1023 a type_id=2 bitfield_size=2 bits_offset=0
1024 b type_id=2 bitfield_size=3 bits_offset=2
1025 c type_id=2 bitfield_size=2 bits_offset=5
1026 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED
1027
1028 The llvm is able to generate .BTF and .BTF.ext directly with -g for bpf target
1029 only. The assembly code (-S) is able to show the BTF encoding in assembly
1030 format.::
1031
1032 -bash-4.4$ cat t2.c
1033 typedef int __int32;
1034 struct t2 {
1035 int a2;
1036 int (*f2)(char q1, __int32 q2, ...);
1037 int (*f3)();
1038 } g2;
1039 int main() { return 0; }
1040 int test() { return 0; }
1041 -bash-4.4$ clang -c -g -O2 --target=bpf t2.c
1042 -bash-4.4$ readelf -S t2.o
1043 ......
1044 [ 8] .BTF PROGBITS 0000000000000000 00000247
1045 000000000000016e 0000000000000000 0 0 1
1046 [ 9] .BTF.ext PROGBITS 0000000000000000 000003b5
1047 0000000000000060 0000000000000000 0 0 1
1048 [10] .rel.BTF.ext REL 0000000000000000 000007e0
1049 0000000000000040 0000000000000010 16 9 8
1050 ......
1051 -bash-4.4$ clang -S -g -O2 --target=bpf t2.c
1052 -bash-4.4$ cat t2.s
1053 ......
1054 .section .BTF,"",@progbits
1055 .short 60319 # 0xeb9f
1056 .byte 1
1057 .byte 0
1058 .long 24
1059 .long 0
1060 .long 220
1061 .long 220
1062 .long 122
1063 .long 0 # BTF_KIND_FUNC_PROTO(id = 1)
1064 .long 218103808 # 0xd000000
1065 .long 2
1066 .long 83 # BTF_KIND_INT(id = 2)
1067 .long 16777216 # 0x1000000
1068 .long 4
1069 .long 16777248 # 0x1000020
1070 ......
1071 .byte 0 # string offset=0
1072 .ascii ".text" # string offset=1
1073 .byte 0
1074 .ascii "/home/yhs/tmp-pahole/t2.c" # string offset=7
1075 .byte 0
1076 .ascii "int main() { return 0; }" # string offset=33
1077 .byte 0
1078 .ascii "int test() { return 0; }" # string offset=58
1079 .byte 0
1080 .ascii "int" # string offset=83
1081 ......
1082 .section .BTF.ext,"",@progbits
1083 .short 60319 # 0xeb9f
1084 .byte 1
1085 .byte 0
1086 .long 24
1087 .long 0
1088 .long 28
1089 .long 28
1090 .long 44
1091 .long 8 # FuncInfo
1092 .long 1 # FuncInfo section string offset=1
1093 .long 2
1094 .long .Lfunc_begin0
1095 .long 3
1096 .long .Lfunc_begin1
1097 .long 5
1098 .long 16 # LineInfo
1099 .long 1 # LineInfo section string offset=1
1100 .long 2
1101 .long .Ltmp0
1102 .long 7
1103 .long 33
1104 .long 7182 # Line 7 Col 14
1105 .long .Ltmp3
1106 .long 7
1107 .long 58
1108 .long 8206 # Line 8 Col 14
1109
1110 7. Testing
1111 ==========
1112
1113 The kernel BPF selftest `tools/testing/selftests/bpf/prog_tests/btf.c`_
1114 provides an extensive set of BTF-related tests.
1115
1116 .. Links
1117 .. _tools/testing/selftests/bpf/prog_tests/btf.c:
1118 https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/tools/testing/selftests/bpf/prog_tests/btf.c
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.