1 ===================== 1 =====================
2 BPF Type Format (BTF) 2 BPF Type Format (BTF)
3 ===================== 3 =====================
4 4
5 1. Introduction 5 1. Introduction
6 =============== !! 6 ***************
7 7
8 BTF (BPF Type Format) is the metadata format w 8 BTF (BPF Type Format) is the metadata format which encodes the debug info
9 related to BPF program/map. The name BTF was u 9 related to BPF program/map. The name BTF was used initially to describe data
10 types. The BTF was later extended to include f 10 types. The BTF was later extended to include function info for defined
11 subroutines, and line info for source/line inf 11 subroutines, and line info for source/line information.
12 12
13 The debug info is used for map pretty print, f 13 The debug info is used for map pretty print, function signature, etc. The
14 function signature enables better bpf program/ 14 function signature enables better bpf program/function kernel symbol. The line
15 info helps generate source annotated translate 15 info helps generate source annotated translated byte code, jited code and
16 verifier log. 16 verifier log.
17 17
18 The BTF specification contains two parts, 18 The BTF specification contains two parts,
19 * BTF kernel API 19 * BTF kernel API
20 * BTF ELF file format 20 * BTF ELF file format
21 21
22 The kernel API is the contract between user sp 22 The kernel API is the contract between user space and kernel. The kernel
23 verifies the BTF info before using it. The ELF 23 verifies the BTF info before using it. The ELF file format is a user space
24 contract between ELF file and libbpf loader. 24 contract between ELF file and libbpf loader.
25 25
26 The type and string sections are part of the B 26 The type and string sections are part of the BTF kernel API, describing the
27 debug info (mostly types related) referenced b 27 debug info (mostly types related) referenced by the bpf program. These two
28 sections are discussed in details in :ref:`BTF 28 sections are discussed in details in :ref:`BTF_Type_String`.
29 29
30 .. _BTF_Type_String: 30 .. _BTF_Type_String:
31 31
32 2. BTF Type and String Encoding 32 2. BTF Type and String Encoding
33 =============================== !! 33 *******************************
34 34
35 The file ``include/uapi/linux/btf.h`` provides 35 The file ``include/uapi/linux/btf.h`` provides high-level definition of how
36 types/strings are encoded. 36 types/strings are encoded.
37 37
38 The beginning of data blob must be:: 38 The beginning of data blob must be::
39 39
40 struct btf_header { 40 struct btf_header {
41 __u16 magic; 41 __u16 magic;
42 __u8 version; 42 __u8 version;
43 __u8 flags; 43 __u8 flags;
44 __u32 hdr_len; 44 __u32 hdr_len;
45 45
46 /* All offsets are in bytes relative t 46 /* All offsets are in bytes relative to the end of this header */
47 __u32 type_off; /* offset of t 47 __u32 type_off; /* offset of type section */
48 __u32 type_len; /* length of t 48 __u32 type_len; /* length of type section */
49 __u32 str_off; /* offset of s 49 __u32 str_off; /* offset of string section */
50 __u32 str_len; /* length of s 50 __u32 str_len; /* length of string section */
51 }; 51 };
52 52
53 The magic is ``0xeB9F``, which has different e 53 The magic is ``0xeB9F``, which has different encoding for big and little
54 endian systems, and can be used to test whethe 54 endian systems, and can be used to test whether BTF is generated for big- or
55 little-endian target. The ``btf_header`` is de 55 little-endian target. The ``btf_header`` is designed to be extensible with
56 ``hdr_len`` equal to ``sizeof(struct btf_heade 56 ``hdr_len`` equal to ``sizeof(struct btf_header)`` when a data blob is
57 generated. 57 generated.
58 58
59 2.1 String Encoding 59 2.1 String Encoding
60 ------------------- !! 60 ===================
61 61
62 The first string in the string section must be 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- 63 string table is a concatenation of other null-terminated strings.
64 64
65 2.2 Type Encoding 65 2.2 Type Encoding
66 ----------------- !! 66 =================
67 67
68 The type id ``0`` is reserved for ``void`` typ 68 The type id ``0`` is reserved for ``void`` type. The type section is parsed
69 sequentially and type id is assigned to each r 69 sequentially and type id is assigned to each recognized type starting from id
70 ``1``. Currently, the following types are supp 70 ``1``. Currently, the following types are supported::
71 71
72 #define BTF_KIND_INT 1 /* 72 #define BTF_KIND_INT 1 /* Integer */
73 #define BTF_KIND_PTR 2 /* 73 #define BTF_KIND_PTR 2 /* Pointer */
74 #define BTF_KIND_ARRAY 3 /* 74 #define BTF_KIND_ARRAY 3 /* Array */
75 #define BTF_KIND_STRUCT 4 /* 75 #define BTF_KIND_STRUCT 4 /* Struct */
76 #define BTF_KIND_UNION 5 /* 76 #define BTF_KIND_UNION 5 /* Union */
77 #define BTF_KIND_ENUM 6 /* !! 77 #define BTF_KIND_ENUM 6 /* Enumeration */
78 #define BTF_KIND_FWD 7 /* 78 #define BTF_KIND_FWD 7 /* Forward */
79 #define BTF_KIND_TYPEDEF 8 /* 79 #define BTF_KIND_TYPEDEF 8 /* Typedef */
80 #define BTF_KIND_VOLATILE 9 /* 80 #define BTF_KIND_VOLATILE 9 /* Volatile */
81 #define BTF_KIND_CONST 10 /* 81 #define BTF_KIND_CONST 10 /* Const */
82 #define BTF_KIND_RESTRICT 11 /* 82 #define BTF_KIND_RESTRICT 11 /* Restrict */
83 #define BTF_KIND_FUNC 12 /* 83 #define BTF_KIND_FUNC 12 /* Function */
84 #define BTF_KIND_FUNC_PROTO 13 /* 84 #define BTF_KIND_FUNC_PROTO 13 /* Function Proto */
85 #define BTF_KIND_VAR 14 /* 85 #define BTF_KIND_VAR 14 /* Variable */
86 #define BTF_KIND_DATASEC 15 /* 86 #define BTF_KIND_DATASEC 15 /* Section */
87 #define BTF_KIND_FLOAT 16 /* <<
88 #define BTF_KIND_DECL_TAG 17 /* <<
89 #define BTF_KIND_TYPE_TAG 18 /* <<
90 #define BTF_KIND_ENUM64 19 /* <<
91 87
92 Note that the type section encodes debug info, 88 Note that the type section encodes debug info, not just pure types.
93 ``BTF_KIND_FUNC`` is not a type, and it repres 89 ``BTF_KIND_FUNC`` is not a type, and it represents a defined subprogram.
94 90
95 Each type contains the following common data:: 91 Each type contains the following common data::
96 92
97 struct btf_type { 93 struct btf_type {
98 __u32 name_off; 94 __u32 name_off;
99 /* "info" bits arrangement 95 /* "info" bits arrangement
100 * bits 0-15: vlen (e.g. # of struct' 96 * bits 0-15: vlen (e.g. # of struct's members)
101 * bits 16-23: unused 97 * bits 16-23: unused
102 * bits 24-28: kind (e.g. int, ptr, ar !! 98 * bits 24-27: kind (e.g. int, ptr, array...etc)
103 * bits 29-30: unused !! 99 * bits 28-30: unused
104 * bit 31: kind_flag, currently us 100 * bit 31: kind_flag, currently used by
105 * struct, union, fwd, enu !! 101 * struct, union and fwd
106 */ 102 */
107 __u32 info; 103 __u32 info;
108 /* "size" is used by INT, ENUM, STRUCT !! 104 /* "size" is used by INT, ENUM, STRUCT and UNION.
109 * "size" tells the size of the type i 105 * "size" tells the size of the type it is describing.
110 * 106 *
111 * "type" is used by PTR, TYPEDEF, VOL 107 * "type" is used by PTR, TYPEDEF, VOLATILE, CONST, RESTRICT,
112 * FUNC, FUNC_PROTO, DECL_TAG and TYPE !! 108 * FUNC and FUNC_PROTO.
113 * "type" is a type_id referring to an 109 * "type" is a type_id referring to another type.
114 */ 110 */
115 union { 111 union {
116 __u32 size; 112 __u32 size;
117 __u32 type; 113 __u32 type;
118 }; 114 };
119 }; 115 };
120 116
121 For certain kinds, the common data are followe 117 For certain kinds, the common data are followed by kind-specific data. The
122 ``name_off`` in ``struct btf_type`` specifies 118 ``name_off`` in ``struct btf_type`` specifies the offset in the string table.
123 The following sections detail encoding of each 119 The following sections detail encoding of each kind.
124 120
125 2.2.1 BTF_KIND_INT 121 2.2.1 BTF_KIND_INT
126 ~~~~~~~~~~~~~~~~~~ 122 ~~~~~~~~~~~~~~~~~~
127 123
128 ``struct btf_type`` encoding requirement: 124 ``struct btf_type`` encoding requirement:
129 * ``name_off``: any valid offset 125 * ``name_off``: any valid offset
130 * ``info.kind_flag``: 0 126 * ``info.kind_flag``: 0
131 * ``info.kind``: BTF_KIND_INT 127 * ``info.kind``: BTF_KIND_INT
132 * ``info.vlen``: 0 128 * ``info.vlen``: 0
133 * ``size``: the size of the int type in bytes 129 * ``size``: the size of the int type in bytes.
134 130
135 ``btf_type`` is followed by a ``u32`` with the 131 ``btf_type`` is followed by a ``u32`` with the following bits arrangement::
136 132
137 #define BTF_INT_ENCODING(VAL) (((VAL) & 0x 133 #define BTF_INT_ENCODING(VAL) (((VAL) & 0x0f000000) >> 24)
138 #define BTF_INT_OFFSET(VAL) (((VAL) & 0x 134 #define BTF_INT_OFFSET(VAL) (((VAL) & 0x00ff0000) >> 16)
139 #define BTF_INT_BITS(VAL) ((VAL) & 0x 135 #define BTF_INT_BITS(VAL) ((VAL) & 0x000000ff)
140 136
141 The ``BTF_INT_ENCODING`` has the following att 137 The ``BTF_INT_ENCODING`` has the following attributes::
142 138
143 #define BTF_INT_SIGNED (1 << 0) 139 #define BTF_INT_SIGNED (1 << 0)
144 #define BTF_INT_CHAR (1 << 1) 140 #define BTF_INT_CHAR (1 << 1)
145 #define BTF_INT_BOOL (1 << 2) 141 #define BTF_INT_BOOL (1 << 2)
146 142
147 The ``BTF_INT_ENCODING()`` provides extra info 143 The ``BTF_INT_ENCODING()`` provides extra information: signedness, char, or
148 bool, for the int type. The char and bool enco 144 bool, for the int type. The char and bool encoding are mostly useful for
149 pretty print. At most one encoding can be spec 145 pretty print. At most one encoding can be specified for the int type.
150 146
151 The ``BTF_INT_BITS()`` specifies the number of 147 The ``BTF_INT_BITS()`` specifies the number of actual bits held by this int
152 type. For example, a 4-bit bitfield encodes `` 148 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 149 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_B 150 for the type. The maximum value of ``BTF_INT_BITS()`` is 128.
155 151
156 The ``BTF_INT_OFFSET()`` specifies the startin 152 The ``BTF_INT_OFFSET()`` specifies the starting bit offset to calculate values
157 for this int. For example, a bitfield struct m 153 for this int. For example, a bitfield struct member has:
158 154
159 * btf member bit offset 100 from the start of 155 * btf member bit offset 100 from the start of the structure,
160 * btf member pointing to an int type, 156 * btf member pointing to an int type,
161 * the int type has ``BTF_INT_OFFSET() = 2`` a 157 * the int type has ``BTF_INT_OFFSET() = 2`` and ``BTF_INT_BITS() = 4``
162 158
163 Then in the struct memory layout, this member 159 Then in the struct memory layout, this member will occupy ``4`` bits starting
164 from bits ``100 + 2 = 102``. 160 from bits ``100 + 2 = 102``.
165 161
166 Alternatively, the bitfield struct member can 162 Alternatively, the bitfield struct member can be the following to access the
167 same bits as the above: 163 same bits as the above:
168 164
169 * btf member bit offset 102, 165 * btf member bit offset 102,
170 * btf member pointing to an int type, 166 * btf member pointing to an int type,
171 * the int type has ``BTF_INT_OFFSET() = 0`` a 167 * the int type has ``BTF_INT_OFFSET() = 0`` and ``BTF_INT_BITS() = 4``
172 168
173 The original intention of ``BTF_INT_OFFSET()`` 169 The original intention of ``BTF_INT_OFFSET()`` is to provide flexibility of
174 bitfield encoding. Currently, both llvm and pa 170 bitfield encoding. Currently, both llvm and pahole generate
175 ``BTF_INT_OFFSET() = 0`` for all int types. 171 ``BTF_INT_OFFSET() = 0`` for all int types.
176 172
177 2.2.2 BTF_KIND_PTR 173 2.2.2 BTF_KIND_PTR
178 ~~~~~~~~~~~~~~~~~~ 174 ~~~~~~~~~~~~~~~~~~
179 175
180 ``struct btf_type`` encoding requirement: 176 ``struct btf_type`` encoding requirement:
181 * ``name_off``: 0 177 * ``name_off``: 0
182 * ``info.kind_flag``: 0 178 * ``info.kind_flag``: 0
183 * ``info.kind``: BTF_KIND_PTR 179 * ``info.kind``: BTF_KIND_PTR
184 * ``info.vlen``: 0 180 * ``info.vlen``: 0
185 * ``type``: the pointee type of the pointer 181 * ``type``: the pointee type of the pointer
186 182
187 No additional type data follow ``btf_type``. 183 No additional type data follow ``btf_type``.
188 184
189 2.2.3 BTF_KIND_ARRAY 185 2.2.3 BTF_KIND_ARRAY
190 ~~~~~~~~~~~~~~~~~~~~ 186 ~~~~~~~~~~~~~~~~~~~~
191 187
192 ``struct btf_type`` encoding requirement: 188 ``struct btf_type`` encoding requirement:
193 * ``name_off``: 0 189 * ``name_off``: 0
194 * ``info.kind_flag``: 0 190 * ``info.kind_flag``: 0
195 * ``info.kind``: BTF_KIND_ARRAY 191 * ``info.kind``: BTF_KIND_ARRAY
196 * ``info.vlen``: 0 192 * ``info.vlen``: 0
197 * ``size/type``: 0, not used 193 * ``size/type``: 0, not used
198 194
199 ``btf_type`` is followed by one ``struct btf_a 195 ``btf_type`` is followed by one ``struct btf_array``::
200 196
201 struct btf_array { 197 struct btf_array {
202 __u32 type; 198 __u32 type;
203 __u32 index_type; 199 __u32 index_type;
204 __u32 nelems; 200 __u32 nelems;
205 }; 201 };
206 202
207 The ``struct btf_array`` encoding: 203 The ``struct btf_array`` encoding:
208 * ``type``: the element type 204 * ``type``: the element type
209 * ``index_type``: the index type 205 * ``index_type``: the index type
210 * ``nelems``: the number of elements for thi 206 * ``nelems``: the number of elements for this array (``0`` is also allowed).
211 207
212 The ``index_type`` can be any regular int type 208 The ``index_type`` can be any regular int type (``u8``, ``u16``, ``u32``,
213 ``u64``, ``unsigned __int128``). The original 209 ``u64``, ``unsigned __int128``). The original design of including
214 ``index_type`` follows DWARF, which has an ``i 210 ``index_type`` follows DWARF, which has an ``index_type`` for its array type.
215 Currently in BTF, beyond type verification, th 211 Currently in BTF, beyond type verification, the ``index_type`` is not used.
216 212
217 The ``struct btf_array`` allows chaining throu 213 The ``struct btf_array`` allows chaining through element type to represent
218 multidimensional arrays. For example, for ``in 214 multidimensional arrays. For example, for ``int a[5][6]``, the following type
219 information illustrates the chaining: 215 information illustrates the chaining:
220 216
221 * [1]: int 217 * [1]: int
222 * [2]: array, ``btf_array.type = [1]``, ``bt 218 * [2]: array, ``btf_array.type = [1]``, ``btf_array.nelems = 6``
223 * [3]: array, ``btf_array.type = [2]``, ``bt 219 * [3]: array, ``btf_array.type = [2]``, ``btf_array.nelems = 5``
224 220
225 Currently, both pahole and llvm collapse multi 221 Currently, both pahole and llvm collapse multidimensional array into
226 one-dimensional array, e.g., for ``a[5][6]``, 222 one-dimensional array, e.g., for ``a[5][6]``, the ``btf_array.nelems`` is
227 equal to ``30``. This is because the original 223 equal to ``30``. This is because the original use case is map pretty print
228 where the whole array is dumped out so one-dim 224 where the whole array is dumped out so one-dimensional array is enough. As
229 more BTF usage is explored, pahole and llvm ca 225 more BTF usage is explored, pahole and llvm can be changed to generate proper
230 chained representation for multidimensional ar 226 chained representation for multidimensional arrays.
231 227
232 2.2.4 BTF_KIND_STRUCT 228 2.2.4 BTF_KIND_STRUCT
233 ~~~~~~~~~~~~~~~~~~~~~ 229 ~~~~~~~~~~~~~~~~~~~~~
234 2.2.5 BTF_KIND_UNION 230 2.2.5 BTF_KIND_UNION
235 ~~~~~~~~~~~~~~~~~~~~ 231 ~~~~~~~~~~~~~~~~~~~~
236 232
237 ``struct btf_type`` encoding requirement: 233 ``struct btf_type`` encoding requirement:
238 * ``name_off``: 0 or offset to a valid C ide 234 * ``name_off``: 0 or offset to a valid C identifier
239 * ``info.kind_flag``: 0 or 1 235 * ``info.kind_flag``: 0 or 1
240 * ``info.kind``: BTF_KIND_STRUCT or BTF_KIND 236 * ``info.kind``: BTF_KIND_STRUCT or BTF_KIND_UNION
241 * ``info.vlen``: the number of struct/union 237 * ``info.vlen``: the number of struct/union members
242 * ``info.size``: the size of the struct/unio 238 * ``info.size``: the size of the struct/union in bytes
243 239
244 ``btf_type`` is followed by ``info.vlen`` numb 240 ``btf_type`` is followed by ``info.vlen`` number of ``struct btf_member``.::
245 241
246 struct btf_member { 242 struct btf_member {
247 __u32 name_off; 243 __u32 name_off;
248 __u32 type; 244 __u32 type;
249 __u32 offset; 245 __u32 offset;
250 }; 246 };
251 247
252 ``struct btf_member`` encoding: 248 ``struct btf_member`` encoding:
253 * ``name_off``: offset to a valid C identifi 249 * ``name_off``: offset to a valid C identifier
254 * ``type``: the member type 250 * ``type``: the member type
255 * ``offset``: <see below> 251 * ``offset``: <see below>
256 252
257 If the type info ``kind_flag`` is not set, the 253 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 254 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 typ 255 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 256 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. 257 ``BTF_INT_BITS()`` encodes the bitfield size.
262 258
263 If the ``kind_flag`` is set, the ``btf_member. 259 If the ``kind_flag`` is set, the ``btf_member.offset`` contains both member
264 bitfield size and bit offset. The bitfield siz 260 bitfield size and bit offset. The bitfield size and bit offset are calculated
265 as below.:: 261 as below.::
266 262
267 #define BTF_MEMBER_BITFIELD_SIZE(val) ((va 263 #define BTF_MEMBER_BITFIELD_SIZE(val) ((val) >> 24)
268 #define BTF_MEMBER_BIT_OFFSET(val) ((va 264 #define BTF_MEMBER_BIT_OFFSET(val) ((val) & 0xffffff)
269 265
270 In this case, if the base type is an int type, 266 In this case, if the base type is an int type, it must be a regular int type:
271 267
272 * ``BTF_INT_OFFSET()`` must be 0. 268 * ``BTF_INT_OFFSET()`` must be 0.
273 * ``BTF_INT_BITS()`` must be equal to ``{1,2 269 * ``BTF_INT_BITS()`` must be equal to ``{1,2,4,8,16} * 8``.
274 270
275 Commit 9d5f9f701b18 introduced ``kind_flag`` a !! 271 The following kernel patch introduced ``kind_flag`` and explained why both
276 exist. !! 272 modes exist:
>> 273
>> 274 https://github.com/torvalds/linux/commit/9d5f9f701b1891466fb3dbb1806ad97716f95cc3#diff-fa650a64fdd3968396883d2fe8215ff3
277 275
278 2.2.6 BTF_KIND_ENUM 276 2.2.6 BTF_KIND_ENUM
279 ~~~~~~~~~~~~~~~~~~~ 277 ~~~~~~~~~~~~~~~~~~~
280 278
281 ``struct btf_type`` encoding requirement: 279 ``struct btf_type`` encoding requirement:
282 * ``name_off``: 0 or offset to a valid C ide 280 * ``name_off``: 0 or offset to a valid C identifier
283 * ``info.kind_flag``: 0 for unsigned, 1 for !! 281 * ``info.kind_flag``: 0
284 * ``info.kind``: BTF_KIND_ENUM 282 * ``info.kind``: BTF_KIND_ENUM
285 * ``info.vlen``: number of enum values 283 * ``info.vlen``: number of enum values
286 * ``size``: 1/2/4/8 !! 284 * ``size``: 4
287 285
288 ``btf_type`` is followed by ``info.vlen`` numb 286 ``btf_type`` is followed by ``info.vlen`` number of ``struct btf_enum``.::
289 287
290 struct btf_enum { 288 struct btf_enum {
291 __u32 name_off; 289 __u32 name_off;
292 __s32 val; 290 __s32 val;
293 }; 291 };
294 292
295 The ``btf_enum`` encoding: 293 The ``btf_enum`` encoding:
296 * ``name_off``: offset to a valid C identifi 294 * ``name_off``: offset to a valid C identifier
297 * ``val``: any value 295 * ``val``: any value
298 296
299 If the original enum value is signed and the s <<
300 that value will be sign extended into 4 bytes. <<
301 the value will be truncated into 4 bytes. <<
302 <<
303 2.2.7 BTF_KIND_FWD 297 2.2.7 BTF_KIND_FWD
304 ~~~~~~~~~~~~~~~~~~ 298 ~~~~~~~~~~~~~~~~~~
305 299
306 ``struct btf_type`` encoding requirement: 300 ``struct btf_type`` encoding requirement:
307 * ``name_off``: offset to a valid C identifi 301 * ``name_off``: offset to a valid C identifier
308 * ``info.kind_flag``: 0 for struct, 1 for un 302 * ``info.kind_flag``: 0 for struct, 1 for union
309 * ``info.kind``: BTF_KIND_FWD 303 * ``info.kind``: BTF_KIND_FWD
310 * ``info.vlen``: 0 304 * ``info.vlen``: 0
311 * ``type``: 0 305 * ``type``: 0
312 306
313 No additional type data follow ``btf_type``. 307 No additional type data follow ``btf_type``.
314 308
315 2.2.8 BTF_KIND_TYPEDEF 309 2.2.8 BTF_KIND_TYPEDEF
316 ~~~~~~~~~~~~~~~~~~~~~~ 310 ~~~~~~~~~~~~~~~~~~~~~~
317 311
318 ``struct btf_type`` encoding requirement: 312 ``struct btf_type`` encoding requirement:
319 * ``name_off``: offset to a valid C identifi 313 * ``name_off``: offset to a valid C identifier
320 * ``info.kind_flag``: 0 314 * ``info.kind_flag``: 0
321 * ``info.kind``: BTF_KIND_TYPEDEF 315 * ``info.kind``: BTF_KIND_TYPEDEF
322 * ``info.vlen``: 0 316 * ``info.vlen``: 0
323 * ``type``: the type which can be referred b 317 * ``type``: the type which can be referred by name at ``name_off``
324 318
325 No additional type data follow ``btf_type``. 319 No additional type data follow ``btf_type``.
326 320
327 2.2.9 BTF_KIND_VOLATILE 321 2.2.9 BTF_KIND_VOLATILE
328 ~~~~~~~~~~~~~~~~~~~~~~~ 322 ~~~~~~~~~~~~~~~~~~~~~~~
329 323
330 ``struct btf_type`` encoding requirement: 324 ``struct btf_type`` encoding requirement:
331 * ``name_off``: 0 325 * ``name_off``: 0
332 * ``info.kind_flag``: 0 326 * ``info.kind_flag``: 0
333 * ``info.kind``: BTF_KIND_VOLATILE 327 * ``info.kind``: BTF_KIND_VOLATILE
334 * ``info.vlen``: 0 328 * ``info.vlen``: 0
335 * ``type``: the type with ``volatile`` quali 329 * ``type``: the type with ``volatile`` qualifier
336 330
337 No additional type data follow ``btf_type``. 331 No additional type data follow ``btf_type``.
338 332
339 2.2.10 BTF_KIND_CONST 333 2.2.10 BTF_KIND_CONST
340 ~~~~~~~~~~~~~~~~~~~~~ 334 ~~~~~~~~~~~~~~~~~~~~~
341 335
342 ``struct btf_type`` encoding requirement: 336 ``struct btf_type`` encoding requirement:
343 * ``name_off``: 0 337 * ``name_off``: 0
344 * ``info.kind_flag``: 0 338 * ``info.kind_flag``: 0
345 * ``info.kind``: BTF_KIND_CONST 339 * ``info.kind``: BTF_KIND_CONST
346 * ``info.vlen``: 0 340 * ``info.vlen``: 0
347 * ``type``: the type with ``const`` qualifie 341 * ``type``: the type with ``const`` qualifier
348 342
349 No additional type data follow ``btf_type``. 343 No additional type data follow ``btf_type``.
350 344
351 2.2.11 BTF_KIND_RESTRICT 345 2.2.11 BTF_KIND_RESTRICT
352 ~~~~~~~~~~~~~~~~~~~~~~~~ 346 ~~~~~~~~~~~~~~~~~~~~~~~~
353 347
354 ``struct btf_type`` encoding requirement: 348 ``struct btf_type`` encoding requirement:
355 * ``name_off``: 0 349 * ``name_off``: 0
356 * ``info.kind_flag``: 0 350 * ``info.kind_flag``: 0
357 * ``info.kind``: BTF_KIND_RESTRICT 351 * ``info.kind``: BTF_KIND_RESTRICT
358 * ``info.vlen``: 0 352 * ``info.vlen``: 0
359 * ``type``: the type with ``restrict`` quali 353 * ``type``: the type with ``restrict`` qualifier
360 354
361 No additional type data follow ``btf_type``. 355 No additional type data follow ``btf_type``.
362 356
363 2.2.12 BTF_KIND_FUNC 357 2.2.12 BTF_KIND_FUNC
364 ~~~~~~~~~~~~~~~~~~~~ 358 ~~~~~~~~~~~~~~~~~~~~
365 359
366 ``struct btf_type`` encoding requirement: 360 ``struct btf_type`` encoding requirement:
367 * ``name_off``: offset to a valid C identifi 361 * ``name_off``: offset to a valid C identifier
368 * ``info.kind_flag``: 0 362 * ``info.kind_flag``: 0
369 * ``info.kind``: BTF_KIND_FUNC 363 * ``info.kind``: BTF_KIND_FUNC
370 * ``info.vlen``: linkage information (BTF_FU !! 364 * ``info.vlen``: 0
371 or BTF_FUNC_EXTERN - see :r <<
372 * ``type``: a BTF_KIND_FUNC_PROTO type 365 * ``type``: a BTF_KIND_FUNC_PROTO type
373 366
374 No additional type data follow ``btf_type``. 367 No additional type data follow ``btf_type``.
375 368
376 A BTF_KIND_FUNC defines not a type, but a subp 369 A BTF_KIND_FUNC defines not a type, but a subprogram (function) whose
377 signature is defined by ``type``. The subprogr 370 signature is defined by ``type``. The subprogram is thus an instance of that
378 type. The BTF_KIND_FUNC may in turn be referen 371 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 argumen 372 :ref:`BTF_Ext_Section` (ELF) or in the arguments to :ref:`BPF_Prog_Load`
380 (ABI). 373 (ABI).
381 374
382 Currently, only linkage values of BTF_FUNC_STA <<
383 supported in the kernel. <<
384 <<
385 2.2.13 BTF_KIND_FUNC_PROTO 375 2.2.13 BTF_KIND_FUNC_PROTO
386 ~~~~~~~~~~~~~~~~~~~~~~~~~~ 376 ~~~~~~~~~~~~~~~~~~~~~~~~~~
387 377
388 ``struct btf_type`` encoding requirement: 378 ``struct btf_type`` encoding requirement:
389 * ``name_off``: 0 379 * ``name_off``: 0
390 * ``info.kind_flag``: 0 380 * ``info.kind_flag``: 0
391 * ``info.kind``: BTF_KIND_FUNC_PROTO 381 * ``info.kind``: BTF_KIND_FUNC_PROTO
392 * ``info.vlen``: # of parameters 382 * ``info.vlen``: # of parameters
393 * ``type``: the return type 383 * ``type``: the return type
394 384
395 ``btf_type`` is followed by ``info.vlen`` numb 385 ``btf_type`` is followed by ``info.vlen`` number of ``struct btf_param``.::
396 386
397 struct btf_param { 387 struct btf_param {
398 __u32 name_off; 388 __u32 name_off;
399 __u32 type; 389 __u32 type;
400 }; 390 };
401 391
402 If a BTF_KIND_FUNC_PROTO type is referred by a 392 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 393 ``btf_param.name_off`` must point to a valid C identifier except for the
404 possible last argument representing the variab 394 possible last argument representing the variable argument. The btf_param.type
405 refers to parameter type. 395 refers to parameter type.
406 396
407 If the function has variable arguments, the la 397 If the function has variable arguments, the last parameter is encoded with
408 ``name_off = 0`` and ``type = 0``. 398 ``name_off = 0`` and ``type = 0``.
409 399
410 2.2.14 BTF_KIND_VAR 400 2.2.14 BTF_KIND_VAR
411 ~~~~~~~~~~~~~~~~~~~ 401 ~~~~~~~~~~~~~~~~~~~
412 402
413 ``struct btf_type`` encoding requirement: 403 ``struct btf_type`` encoding requirement:
414 * ``name_off``: offset to a valid C identifi 404 * ``name_off``: offset to a valid C identifier
415 * ``info.kind_flag``: 0 405 * ``info.kind_flag``: 0
416 * ``info.kind``: BTF_KIND_VAR 406 * ``info.kind``: BTF_KIND_VAR
417 * ``info.vlen``: 0 407 * ``info.vlen``: 0
418 * ``type``: the type of the variable 408 * ``type``: the type of the variable
419 409
420 ``btf_type`` is followed by a single ``struct 410 ``btf_type`` is followed by a single ``struct btf_variable`` with the
421 following data:: 411 following data::
422 412
423 struct btf_var { 413 struct btf_var {
424 __u32 linkage; 414 __u32 linkage;
425 }; 415 };
426 416
427 ``btf_var.linkage`` may take the values: BTF_V !! 417 ``struct btf_var`` encoding:
428 see :ref:`BTF_Var_Linkage_Constants`. !! 418 * ``linkage``: currently only static variable 0, or globally allocated
>> 419 variable in ELF sections 1
429 420
430 Not all type of global variables are supported 421 Not all type of global variables are supported by LLVM at this point.
431 The following is currently available: 422 The following is currently available:
432 423
433 * static variables with or without section a 424 * static variables with or without section attributes
434 * global variables with section attributes 425 * global variables with section attributes
435 426
436 The latter is for future extraction of map key 427 The latter is for future extraction of map key/value type id's from a
437 map definition. 428 map definition.
438 429
439 2.2.15 BTF_KIND_DATASEC 430 2.2.15 BTF_KIND_DATASEC
440 ~~~~~~~~~~~~~~~~~~~~~~~ 431 ~~~~~~~~~~~~~~~~~~~~~~~
441 432
442 ``struct btf_type`` encoding requirement: 433 ``struct btf_type`` encoding requirement:
443 * ``name_off``: offset to a valid name assoc 434 * ``name_off``: offset to a valid name associated with a variable or
444 one of .data/.bss/.rodata 435 one of .data/.bss/.rodata
445 * ``info.kind_flag``: 0 436 * ``info.kind_flag``: 0
446 * ``info.kind``: BTF_KIND_DATASEC 437 * ``info.kind``: BTF_KIND_DATASEC
447 * ``info.vlen``: # of variables 438 * ``info.vlen``: # of variables
448 * ``size``: total section size in bytes (0 a 439 * ``size``: total section size in bytes (0 at compilation time, patched
449 to actual size by BPF loaders su 440 to actual size by BPF loaders such as libbpf)
450 441
451 ``btf_type`` is followed by ``info.vlen`` numb 442 ``btf_type`` is followed by ``info.vlen`` number of ``struct btf_var_secinfo``.::
452 443
453 struct btf_var_secinfo { 444 struct btf_var_secinfo {
454 __u32 type; 445 __u32 type;
455 __u32 offset; 446 __u32 offset;
456 __u32 size; 447 __u32 size;
457 }; 448 };
458 449
459 ``struct btf_var_secinfo`` encoding: 450 ``struct btf_var_secinfo`` encoding:
460 * ``type``: the type of the BTF_KIND_VAR var 451 * ``type``: the type of the BTF_KIND_VAR variable
461 * ``offset``: the in-section offset of the v 452 * ``offset``: the in-section offset of the variable
462 * ``size``: the size of the variable in byte 453 * ``size``: the size of the variable in bytes
463 454
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 byt <<
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``, <<
485 <<
486 ``btf_type`` is followed by ``struct btf_decl_ <<
487 <<
488 struct btf_decl_tag { <<
489 __u32 component_idx; <<
490 }; <<
491 <<
492 The ``name_off`` encodes btf_decl_tag attribut <<
493 The ``type`` should be ``struct``, ``union``, <<
494 For ``var`` or ``typedef`` type, ``btf_decl_ta <<
495 For the other three types, if the btf_decl_tag <<
496 applied to the ``struct``, ``union`` or ``func <<
497 ``btf_decl_tag.component_idx`` must be ``-1``. <<
498 the attribute is applied to a ``struct``/``uni <<
499 a ``func`` argument, and ``btf_decl_tag.compon <<
500 valid index (starting from 0) pointing to a me <<
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`` at <<
511 <<
512 Currently, ``BTF_KIND_TYPE_TAG`` is only emitt <<
513 It has the following btf type chain: <<
514 :: <<
515 <<
516 ptr -> [type_tag]* <<
517 -> [const | volatile | restrict | typede <<
518 -> base_type <<
519 <<
520 Basically, a pointer type points to zero or mo <<
521 type_tag, then zero or more const/volatile/res <<
522 and finally the base type. The base type is on <<
523 int, ptr, array, struct, union, enum, func_pro <<
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 ide <<
530 * ``info.kind_flag``: 0 for unsigned, 1 for <<
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`` numb <<
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 identifi <<
545 * ``val_lo32``: lower 32-bit value for a 64- <<
546 * ``val_hi32``: high 32-bit value for a 64-b <<
547 <<
548 If the original enum value is signed and the s <<
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 Meaning <<
559 <<
560 =================== ===== =========== <<
561 kind value description <<
562 =================== ===== =========== <<
563 ``BTF_FUNC_STATIC`` 0x0 definition of su <<
564 ``BTF_FUNC_GLOBAL`` 0x1 definition of su <<
565 ``BTF_FUNC_EXTERN`` 0x2 declaration of a <<
566 =================== ===== =========== <<
567 <<
568 <<
569 .. _BTF_Var_Linkage_Constants: <<
570 <<
571 2.3.2 Variable Linkage Constant Values <<
572 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ <<
573 .. table:: Variable Linkage Values and Meaning <<
574 <<
575 ============================ ===== ======= <<
576 kind value descrip <<
577 ============================ ===== ======= <<
578 ``BTF_VAR_STATIC`` 0x0 definit <<
579 ``BTF_VAR_GLOBAL_ALLOCATED`` 0x1 definit <<
580 ``BTF_VAR_GLOBAL_EXTERN`` 0x2 declara <<
581 ============================ ===== ======= <<
582 <<
583 3. BTF Kernel API 455 3. BTF Kernel API
584 ================= !! 456 *****************
585 457
586 The following bpf syscall command involves BTF 458 The following bpf syscall command involves BTF:
587 * BPF_BTF_LOAD: load a blob of BTF data int 459 * BPF_BTF_LOAD: load a blob of BTF data into kernel
588 * BPF_MAP_CREATE: map creation with btf key 460 * BPF_MAP_CREATE: map creation with btf key and value type info.
589 * BPF_PROG_LOAD: prog load with btf functio 461 * BPF_PROG_LOAD: prog load with btf function and line info.
590 * BPF_BTF_GET_FD_BY_ID: get a btf fd 462 * BPF_BTF_GET_FD_BY_ID: get a btf fd
591 * BPF_OBJ_GET_INFO_BY_FD: btf, func_info, l 463 * BPF_OBJ_GET_INFO_BY_FD: btf, func_info, line_info
592 and other btf related info are returned. 464 and other btf related info are returned.
593 465
594 The workflow typically looks like: 466 The workflow typically looks like:
595 :: 467 ::
596 468
597 Application: 469 Application:
598 BPF_BTF_LOAD 470 BPF_BTF_LOAD
599 | 471 |
600 v 472 v
601 BPF_MAP_CREATE and BPF_PROG_LOAD 473 BPF_MAP_CREATE and BPF_PROG_LOAD
602 | 474 |
603 V 475 V
604 ...... 476 ......
605 477
606 Introspection tool: 478 Introspection tool:
607 ...... 479 ......
608 BPF_{PROG,MAP}_GET_NEXT_ID (get prog/map 480 BPF_{PROG,MAP}_GET_NEXT_ID (get prog/map id's)
609 | 481 |
610 V 482 V
611 BPF_{PROG,MAP}_GET_FD_BY_ID (get a prog/ 483 BPF_{PROG,MAP}_GET_FD_BY_ID (get a prog/map fd)
612 | 484 |
613 V 485 V
614 BPF_OBJ_GET_INFO_BY_FD (get bpf_prog_inf 486 BPF_OBJ_GET_INFO_BY_FD (get bpf_prog_info/bpf_map_info with btf_id)
615 | 487 | |
616 V 488 V |
617 BPF_BTF_GET_FD_BY_ID (get btf_fd) 489 BPF_BTF_GET_FD_BY_ID (get btf_fd) |
618 | 490 | |
619 V 491 V |
620 BPF_OBJ_GET_INFO_BY_FD (get btf) 492 BPF_OBJ_GET_INFO_BY_FD (get btf) |
621 | 493 | |
622 V 494 V V
623 pretty print types, dump func signatures 495 pretty print types, dump func signatures and line info, etc.
624 496
625 497
626 3.1 BPF_BTF_LOAD 498 3.1 BPF_BTF_LOAD
627 ---------------- !! 499 ================
628 500
629 Load a blob of BTF data into kernel. A blob of 501 Load a blob of BTF data into kernel. A blob of data, described in
630 :ref:`BTF_Type_String`, can be directly loaded 502 :ref:`BTF_Type_String`, can be directly loaded into the kernel. A ``btf_fd``
631 is returned to a userspace. 503 is returned to a userspace.
632 504
633 3.2 BPF_MAP_CREATE 505 3.2 BPF_MAP_CREATE
634 ------------------ !! 506 ==================
635 507
636 A map can be created with ``btf_fd`` and speci 508 A map can be created with ``btf_fd`` and specified key/value type id.::
637 509
638 __u32 btf_fd; /* fd pointing to 510 __u32 btf_fd; /* fd pointing to a BTF type data */
639 __u32 btf_key_type_id; /* BTF typ 511 __u32 btf_key_type_id; /* BTF type_id of the key */
640 __u32 btf_value_type_id; /* BTF typ 512 __u32 btf_value_type_id; /* BTF type_id of the value */
641 513
642 In libbpf, the map can be defined with extra a 514 In libbpf, the map can be defined with extra annotation like below:
643 :: 515 ::
644 516
645 struct { !! 517 struct bpf_map_def SEC("maps") btf_map = {
646 __uint(type, BPF_MAP_TYPE_ARRAY); !! 518 .type = BPF_MAP_TYPE_ARRAY,
647 __type(key, int); !! 519 .key_size = sizeof(int),
648 __type(value, struct ipv_counts); !! 520 .value_size = sizeof(struct ipv_counts),
649 __uint(max_entries, 4); !! 521 .max_entries = 4,
650 } btf_map SEC(".maps"); !! 522 };
651 !! 523 BPF_ANNOTATE_KV_PAIR(btf_map, int, struct ipv_counts);
652 During ELF parsing, libbpf is able to extract !! 524
653 them to BPF_MAP_CREATE attributes automaticall !! 525 Here, the parameters for macro BPF_ANNOTATE_KV_PAIR are map name, key and
>> 526 value types for the map. During ELF parsing, libbpf is able to extract
>> 527 key/value type_id's and assign them to BPF_MAP_CREATE attributes
>> 528 automatically.
654 529
655 .. _BPF_Prog_Load: 530 .. _BPF_Prog_Load:
656 531
657 3.3 BPF_PROG_LOAD 532 3.3 BPF_PROG_LOAD
658 ----------------- !! 533 =================
659 534
660 During prog_load, func_info and line_info can 535 During prog_load, func_info and line_info can be passed to kernel with proper
661 values for the following attributes: 536 values for the following attributes:
662 :: 537 ::
663 538
664 __u32 insn_cnt; 539 __u32 insn_cnt;
665 __aligned_u64 insns; 540 __aligned_u64 insns;
666 ...... 541 ......
667 __u32 prog_btf_fd; /* fd poin 542 __u32 prog_btf_fd; /* fd pointing to BTF type data */
668 __u32 func_info_rec_size; /* 543 __u32 func_info_rec_size; /* userspace bpf_func_info size */
669 __aligned_u64 func_info; /* func in 544 __aligned_u64 func_info; /* func info */
670 __u32 func_info_cnt; /* number 545 __u32 func_info_cnt; /* number of bpf_func_info records */
671 __u32 line_info_rec_size; /* 546 __u32 line_info_rec_size; /* userspace bpf_line_info size */
672 __aligned_u64 line_info; /* line in 547 __aligned_u64 line_info; /* line info */
673 __u32 line_info_cnt; /* number 548 __u32 line_info_cnt; /* number of bpf_line_info records */
674 549
675 The func_info and line_info are an array of be 550 The func_info and line_info are an array of below, respectively.::
676 551
677 struct bpf_func_info { 552 struct bpf_func_info {
678 __u32 insn_off; /* [0, insn_cnt - 1] 553 __u32 insn_off; /* [0, insn_cnt - 1] */
679 __u32 type_id; /* pointing to a BTF 554 __u32 type_id; /* pointing to a BTF_KIND_FUNC type */
680 }; 555 };
681 struct bpf_line_info { 556 struct bpf_line_info {
682 __u32 insn_off; /* [0, insn_cnt - 1] 557 __u32 insn_off; /* [0, insn_cnt - 1] */
683 __u32 file_name_off; /* offset to st 558 __u32 file_name_off; /* offset to string table for the filename */
684 __u32 line_off; /* offset to string 559 __u32 line_off; /* offset to string table for the source line */
685 __u32 line_col; /* line number and c 560 __u32 line_col; /* line number and column number */
686 }; 561 };
687 562
688 func_info_rec_size is the size of each func_in 563 func_info_rec_size is the size of each func_info record, and
689 line_info_rec_size is the size of each line_in 564 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 565 size to kernel make it possible to extend the record itself in the future.
691 566
692 Below are requirements for func_info: 567 Below are requirements for func_info:
693 * func_info[0].insn_off must be 0. 568 * func_info[0].insn_off must be 0.
694 * the func_info insn_off is in strictly incr 569 * the func_info insn_off is in strictly increasing order and matches
695 bpf func boundaries. 570 bpf func boundaries.
696 571
697 Below are requirements for line_info: 572 Below are requirements for line_info:
698 * the first insn in each func must have a li 573 * the first insn in each func must have a line_info record pointing to it.
699 * the line_info insn_off is in strictly incr 574 * the line_info insn_off is in strictly increasing order.
700 575
701 For line_info, the line number and column numb 576 For line_info, the line number and column number are defined as below:
702 :: 577 ::
703 578
704 #define BPF_LINE_INFO_LINE_NUM(line_col) 579 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
705 #define BPF_LINE_INFO_LINE_COL(line_col) 580 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
706 581
707 3.4 BPF_{PROG,MAP}_GET_NEXT_ID 582 3.4 BPF_{PROG,MAP}_GET_NEXT_ID
708 ------------------------------ !! 583 ==============================
709 584
710 In kernel, every loaded program, map or btf ha 585 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, 586 change during the lifetime of a program, map, or btf.
712 587
713 The bpf syscall command BPF_{PROG,MAP}_GET_NEX 588 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 o 589 each command, to user space, for bpf program or maps, respectively, so an
715 inspection tool can inspect all programs and m 590 inspection tool can inspect all programs and maps.
716 591
717 3.5 BPF_{PROG,MAP}_GET_FD_BY_ID 592 3.5 BPF_{PROG,MAP}_GET_FD_BY_ID
718 ------------------------------- !! 593 ===============================
719 594
720 An introspection tool cannot use id to get det 595 An introspection tool cannot use id to get details about program or maps.
721 A file descriptor needs to be obtained first f 596 A file descriptor needs to be obtained first for reference-counting purpose.
722 597
723 3.6 BPF_OBJ_GET_INFO_BY_FD 598 3.6 BPF_OBJ_GET_INFO_BY_FD
724 -------------------------- !! 599 ==========================
725 600
726 Once a program/map fd is acquired, an introspe 601 Once a program/map fd is acquired, an introspection tool can get the detailed
727 information from kernel about this fd, some of 602 information from kernel about this fd, some of which are BTF-related. For
728 example, ``bpf_map_info`` returns ``btf_id`` a 603 example, ``bpf_map_info`` returns ``btf_id`` and key/value type ids.
729 ``bpf_prog_info`` returns ``btf_id``, func_inf 604 ``bpf_prog_info`` returns ``btf_id``, func_info, and line info for translated
730 bpf byte codes, and jited_line_info. 605 bpf byte codes, and jited_line_info.
731 606
732 3.7 BPF_BTF_GET_FD_BY_ID 607 3.7 BPF_BTF_GET_FD_BY_ID
733 ------------------------ !! 608 ========================
734 609
735 With ``btf_id`` obtained in ``bpf_map_info`` a 610 With ``btf_id`` obtained in ``bpf_map_info`` and ``bpf_prog_info``, bpf
736 syscall command BPF_BTF_GET_FD_BY_ID can retri 611 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, 612 command BPF_OBJ_GET_INFO_BY_FD, the btf blob, originally loaded into the
738 kernel with BPF_BTF_LOAD, can be retrieved. 613 kernel with BPF_BTF_LOAD, can be retrieved.
739 614
740 With the btf blob, ``bpf_map_info``, and ``bpf 615 With the btf blob, ``bpf_map_info``, and ``bpf_prog_info``, an introspection
741 tool has full btf knowledge and is able to pre 616 tool has full btf knowledge and is able to pretty print map key/values, dump
742 func signatures and line info, along with byte 617 func signatures and line info, along with byte/jit codes.
743 618
744 4. ELF File Format Interface 619 4. ELF File Format Interface
745 ============================ !! 620 ****************************
746 621
747 4.1 .BTF section 622 4.1 .BTF section
748 ---------------- !! 623 ================
749 624
750 The .BTF section contains type and string data 625 The .BTF section contains type and string data. The format of this section is
751 same as the one describe in :ref:`BTF_Type_Str 626 same as the one describe in :ref:`BTF_Type_String`.
752 627
753 .. _BTF_Ext_Section: 628 .. _BTF_Ext_Section:
754 629
755 4.2 .BTF.ext section 630 4.2 .BTF.ext section
756 -------------------- !! 631 ====================
757 632
758 The .BTF.ext section encodes func_info, line_i !! 633 The .BTF.ext section encodes func_info and line_info which needs loader
759 which needs loader manipulation before loading !! 634 manipulation before loading into the kernel.
760 635
761 The specification for .BTF.ext section is defi 636 The specification for .BTF.ext section is defined at ``tools/lib/bpf/btf.h``
762 and ``tools/lib/bpf/btf.c``. 637 and ``tools/lib/bpf/btf.c``.
763 638
764 The current header of .BTF.ext section:: 639 The current header of .BTF.ext section::
765 640
766 struct btf_ext_header { 641 struct btf_ext_header {
767 __u16 magic; 642 __u16 magic;
768 __u8 version; 643 __u8 version;
769 __u8 flags; 644 __u8 flags;
770 __u32 hdr_len; 645 __u32 hdr_len;
771 646
772 /* All offsets are in bytes relative t 647 /* All offsets are in bytes relative to the end of this header */
773 __u32 func_info_off; 648 __u32 func_info_off;
774 __u32 func_info_len; 649 __u32 func_info_len;
775 __u32 line_info_off; 650 __u32 line_info_off;
776 __u32 line_info_len; 651 __u32 line_info_len;
777 <<
778 /* optional part of .BTF.ext header */ <<
779 __u32 core_relo_off; <<
780 __u32 core_relo_len; <<
781 }; 652 };
782 653
783 It is very similar to .BTF section. Instead of 654 It is very similar to .BTF section. Instead of type/string section, it
784 contains func_info, line_info and core_relo su !! 655 contains func_info and line_info section. See :ref:`BPF_Prog_Load` for details
785 See :ref:`BPF_Prog_Load` for details about fun !! 656 about func_info and line_info record format.
786 record format. <<
787 657
788 The func_info is organized as below.:: 658 The func_info is organized as below.::
789 659
790 func_info_rec_size /* __u32 !! 660 func_info_rec_size
791 btf_ext_info_sec for section #1 /* func_i 661 btf_ext_info_sec for section #1 /* func_info for section #1 */
792 btf_ext_info_sec for section #2 /* func_i 662 btf_ext_info_sec for section #2 /* func_info for section #2 */
793 ... 663 ...
794 664
795 ``func_info_rec_size`` specifies the size of ` 665 ``func_info_rec_size`` specifies the size of ``bpf_func_info`` structure when
796 .BTF.ext is generated. ``btf_ext_info_sec``, d 666 .BTF.ext is generated. ``btf_ext_info_sec``, defined below, is a collection of
797 func_info for each specific ELF section.:: 667 func_info for each specific ELF section.::
798 668
799 struct btf_ext_info_sec { 669 struct btf_ext_info_sec {
800 __u32 sec_name_off; /* offset to sec 670 __u32 sec_name_off; /* offset to section name */
801 __u32 num_info; 671 __u32 num_info;
802 /* Followed by num_info * record_size 672 /* Followed by num_info * record_size number of bytes */
803 __u8 data[0]; 673 __u8 data[0];
804 }; 674 };
805 675
806 Here, num_info must be greater than 0. 676 Here, num_info must be greater than 0.
807 677
808 The line_info is organized as below.:: 678 The line_info is organized as below.::
809 679
810 line_info_rec_size /* __u32 !! 680 line_info_rec_size
811 btf_ext_info_sec for section #1 /* line_i 681 btf_ext_info_sec for section #1 /* line_info for section #1 */
812 btf_ext_info_sec for section #2 /* line_i 682 btf_ext_info_sec for section #2 /* line_info for section #2 */
813 ... 683 ...
814 684
815 ``line_info_rec_size`` specifies the size of ` 685 ``line_info_rec_size`` specifies the size of ``bpf_line_info`` structure when
816 .BTF.ext is generated. 686 .BTF.ext is generated.
817 687
818 The interpretation of ``bpf_func_info->insn_of 688 The interpretation of ``bpf_func_info->insn_off`` and
819 ``bpf_line_info->insn_off`` is different betwe 689 ``bpf_line_info->insn_off`` is different between kernel API and ELF API. For
820 kernel API, the ``insn_off`` is the instructio 690 kernel API, the ``insn_off`` is the instruction offset in the unit of ``struct
821 bpf_insn``. For ELF API, the ``insn_off`` is t 691 bpf_insn``. For ELF API, the ``insn_off`` is the byte offset from the
822 beginning of section (``btf_ext_info_sec->sec_ 692 beginning of section (``btf_ext_info_sec->sec_name_off``).
823 693
824 The core_relo is organized as below.:: <<
825 <<
826 core_relo_rec_size /* __u32 <<
827 btf_ext_info_sec for section #1 /* core_r <<
828 btf_ext_info_sec for section #2 /* core_r <<
829 <<
830 ``core_relo_rec_size`` specifies the size of ` <<
831 structure when .BTF.ext is generated. All ``bp <<
832 within a single ``btf_ext_info_sec`` describe <<
833 section named by ``btf_ext_info_sec->sec_name_ <<
834 <<
835 See :ref:`Documentation/bpf/llvm_reloc.rst <bt <<
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 tha <<
842 <<
843 This section is created during the kernel comp <<
844 macros defined in ``include/linux/btf_ids.h`` <<
845 use them to create lists and sets (sorted list <<
846 <<
847 The ``BTF_ID_LIST`` and ``BTF_ID`` macros defi <<
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 sect <<
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 acc <<
862 <<
863 The ``BTF_ID_UNUSED`` macro defines 4 zero byt <<
864 want to define unused entry in BTF_ID_LIST, li <<
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 <<
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 sect <<
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 def <<
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 <<
895 <<
896 All the BTF ID lists and sets are compiled in <<
897 resolved during the linking phase of kernel bu <<
898 <<
899 5. Using BTF 694 5. Using BTF
900 ============ !! 695 ************
901 696
902 5.1 bpftool map pretty print 697 5.1 bpftool map pretty print
903 ---------------------------- !! 698 ============================
904 699
905 With BTF, the map key/value can be printed bas 700 With BTF, the map key/value can be printed based on fields rather than simply
906 raw bytes. This is especially valuable for lar 701 raw bytes. This is especially valuable for large structure or if your data
907 structure has bitfields. For example, for the 702 structure has bitfields. For example, for the following map,::
908 703
909 enum A { A1, A2, A3, A4, A5 }; 704 enum A { A1, A2, A3, A4, A5 };
910 typedef enum A ___A; 705 typedef enum A ___A;
911 struct tmp_t { 706 struct tmp_t {
912 char a1:4; 707 char a1:4;
913 int a2:4; 708 int a2:4;
914 int :4; 709 int :4;
915 __u32 a3:4; 710 __u32 a3:4;
916 int b; 711 int b;
917 ___A b1:4; 712 ___A b1:4;
918 enum A b2:4; 713 enum A b2:4;
919 }; 714 };
920 struct { !! 715 struct bpf_map_def SEC("maps") tmpmap = {
921 __uint(type, BPF_MAP_TYPE_ARRAY); !! 716 .type = BPF_MAP_TYPE_ARRAY,
922 __type(key, int); !! 717 .key_size = sizeof(__u32),
923 __type(value, struct tmp_t); !! 718 .value_size = sizeof(struct tmp_t),
924 __uint(max_entries, 1); !! 719 .max_entries = 1,
925 } tmpmap SEC(".maps"); !! 720 };
>> 721 BPF_ANNOTATE_KV_PAIR(tmpmap, int, struct tmp_t);
926 722
927 bpftool is able to pretty print like below: 723 bpftool is able to pretty print like below:
928 :: 724 ::
929 725
930 [{ 726 [{
931 "key": 0, 727 "key": 0,
932 "value": { 728 "value": {
933 "a1": 0x2, 729 "a1": 0x2,
934 "a2": 0x4, 730 "a2": 0x4,
935 "a3": 0x6, 731 "a3": 0x6,
936 "b": 7, 732 "b": 7,
937 "b1": 0x8, 733 "b1": 0x8,
938 "b2": 0xa 734 "b2": 0xa
939 } 735 }
940 } 736 }
941 ] 737 ]
942 738
943 5.2 bpftool prog dump 739 5.2 bpftool prog dump
944 --------------------- !! 740 =====================
945 741
946 The following is an example showing how func_i 742 The following is an example showing how func_info and line_info can help prog
947 dump with better kernel symbol names, function 743 dump with better kernel symbol names, function prototypes and line
948 information.:: 744 information.::
949 745
950 $ bpftool prog dump jited pinned /sys/fs/b 746 $ bpftool prog dump jited pinned /sys/fs/bpf/test_btf_haskv
951 [...] 747 [...]
952 int test_long_fname_2(struct dummy_tracepo 748 int test_long_fname_2(struct dummy_tracepoint_args * arg):
953 bpf_prog_44a040bf25481309_test_long_fname_ 749 bpf_prog_44a040bf25481309_test_long_fname_2:
954 ; static int test_long_fname_2(struct dumm 750 ; static int test_long_fname_2(struct dummy_tracepoint_args *arg)
955 0: push %rbp 751 0: push %rbp
956 1: mov %rsp,%rbp 752 1: mov %rsp,%rbp
957 4: sub $0x30,%rsp 753 4: sub $0x30,%rsp
958 b: sub $0x28,%rbp 754 b: sub $0x28,%rbp
959 f: mov %rbx,0x0(%rbp) 755 f: mov %rbx,0x0(%rbp)
960 13: mov %r13,0x8(%rbp) 756 13: mov %r13,0x8(%rbp)
961 17: mov %r14,0x10(%rbp) 757 17: mov %r14,0x10(%rbp)
962 1b: mov %r15,0x18(%rbp) 758 1b: mov %r15,0x18(%rbp)
963 1f: xor %eax,%eax 759 1f: xor %eax,%eax
964 21: mov %rax,0x20(%rbp) 760 21: mov %rax,0x20(%rbp)
965 25: xor %esi,%esi 761 25: xor %esi,%esi
966 ; int key = 0; 762 ; int key = 0;
967 27: mov %esi,-0x4(%rbp) 763 27: mov %esi,-0x4(%rbp)
968 ; if (!arg->sock) 764 ; if (!arg->sock)
969 2a: mov 0x8(%rdi),%rdi 765 2a: mov 0x8(%rdi),%rdi
970 ; if (!arg->sock) 766 ; if (!arg->sock)
971 2e: cmp $0x0,%rdi 767 2e: cmp $0x0,%rdi
972 32: je 0x0000000000000070 768 32: je 0x0000000000000070
973 34: mov %rbp,%rsi 769 34: mov %rbp,%rsi
974 ; counts = bpf_map_lookup_elem(&btf_map, & 770 ; counts = bpf_map_lookup_elem(&btf_map, &key);
975 [...] 771 [...]
976 772
977 5.3 Verifier Log 773 5.3 Verifier Log
978 ---------------- !! 774 ================
979 775
980 The following is an example of how line_info c 776 The following is an example of how line_info can help debugging verification
981 failure.:: 777 failure.::
982 778
983 /* The code at tools/testing/selftests/ 779 /* The code at tools/testing/selftests/bpf/test_xdp_noinline.c
984 * is modified as below. 780 * is modified as below.
985 */ 781 */
986 data = (void *)(long)xdp->data; 782 data = (void *)(long)xdp->data;
987 data_end = (void *)(long)xdp->data_end; 783 data_end = (void *)(long)xdp->data_end;
988 /* 784 /*
989 if (data + 4 > data_end) 785 if (data + 4 > data_end)
990 return XDP_DROP; 786 return XDP_DROP;
991 */ 787 */
992 *(u32 *)data = dst->dst; 788 *(u32 *)data = dst->dst;
993 789
994 $ bpftool prog load ./test_xdp_noinline.o 790 $ bpftool prog load ./test_xdp_noinline.o /sys/fs/bpf/test_xdp_noinline type xdp
995 ; data = (void *)(long)xdp->data; 791 ; data = (void *)(long)xdp->data;
996 224: (79) r2 = *(u64 *)(r10 -112) 792 224: (79) r2 = *(u64 *)(r10 -112)
997 225: (61) r2 = *(u32 *)(r2 +0) 793 225: (61) r2 = *(u32 *)(r2 +0)
998 ; *(u32 *)data = dst->dst; 794 ; *(u32 *)data = dst->dst;
999 226: (63) *(u32 *)(r2 +0) = r1 795 226: (63) *(u32 *)(r2 +0) = r1
1000 invalid access to packet, off=0 size= 796 invalid access to packet, off=0 size=4, R2(id=0,off=0,r=0)
1001 R2 offset is outside of the packet 797 R2 offset is outside of the packet
1002 798
1003 6. BTF Generation 799 6. BTF Generation
1004 ================= !! 800 *****************
1005 801
1006 You need latest pahole 802 You need latest pahole
1007 803
1008 https://git.kernel.org/pub/scm/devel/pahole 804 https://git.kernel.org/pub/scm/devel/pahole/pahole.git/
1009 805
1010 or llvm (8.0 or later). The pahole acts as a 806 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 y 807 support .BTF.ext and btf BTF_KIND_FUNC type yet. For example,::
1012 808
1013 -bash-4.4$ cat t.c 809 -bash-4.4$ cat t.c
1014 struct t { 810 struct t {
1015 int a:2; 811 int a:2;
1016 int b:3; 812 int b:3;
1017 int c:2; 813 int c:2;
1018 } g; 814 } g;
1019 -bash-4.4$ gcc -c -O2 -g t.c 815 -bash-4.4$ gcc -c -O2 -g t.c
1020 -bash-4.4$ pahole -JV t.o 816 -bash-4.4$ pahole -JV t.o
1021 File t.o: 817 File t.o:
1022 [1] STRUCT t kind_flag=1 size=4 vlen=3 818 [1] STRUCT t kind_flag=1 size=4 vlen=3
1023 a type_id=2 bitfield_size=2 bit 819 a type_id=2 bitfield_size=2 bits_offset=0
1024 b type_id=2 bitfield_size=3 bit 820 b type_id=2 bitfield_size=3 bits_offset=2
1025 c type_id=2 bitfield_size=2 bit 821 c type_id=2 bitfield_size=2 bits_offset=5
1026 [2] INT int size=4 bit_offset=0 nr_bits 822 [2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED
1027 823
1028 The llvm is able to generate .BTF and .BTF.ex 824 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 825 only. The assembly code (-S) is able to show the BTF encoding in assembly
1030 format.:: 826 format.::
1031 827
1032 -bash-4.4$ cat t2.c 828 -bash-4.4$ cat t2.c
1033 typedef int __int32; 829 typedef int __int32;
1034 struct t2 { 830 struct t2 {
1035 int a2; 831 int a2;
1036 int (*f2)(char q1, __int32 q2, ...); 832 int (*f2)(char q1, __int32 q2, ...);
1037 int (*f3)(); 833 int (*f3)();
1038 } g2; 834 } g2;
1039 int main() { return 0; } 835 int main() { return 0; }
1040 int test() { return 0; } 836 int test() { return 0; }
1041 -bash-4.4$ clang -c -g -O2 --target=bpf t !! 837 -bash-4.4$ clang -c -g -O2 -target bpf t2.c
1042 -bash-4.4$ readelf -S t2.o 838 -bash-4.4$ readelf -S t2.o
1043 ...... 839 ......
1044 [ 8] .BTF PROGBITS 840 [ 8] .BTF PROGBITS 0000000000000000 00000247
1045 000000000000016e 0000000000000000 841 000000000000016e 0000000000000000 0 0 1
1046 [ 9] .BTF.ext PROGBITS 842 [ 9] .BTF.ext PROGBITS 0000000000000000 000003b5
1047 0000000000000060 0000000000000000 843 0000000000000060 0000000000000000 0 0 1
1048 [10] .rel.BTF.ext REL 844 [10] .rel.BTF.ext REL 0000000000000000 000007e0
1049 0000000000000040 0000000000000010 845 0000000000000040 0000000000000010 16 9 8
1050 ...... 846 ......
1051 -bash-4.4$ clang -S -g -O2 --target=bpf t !! 847 -bash-4.4$ clang -S -g -O2 -target bpf t2.c
1052 -bash-4.4$ cat t2.s 848 -bash-4.4$ cat t2.s
1053 ...... 849 ......
1054 .section .BTF,"",@progbits 850 .section .BTF,"",@progbits
1055 .short 60319 # 851 .short 60319 # 0xeb9f
1056 .byte 1 852 .byte 1
1057 .byte 0 853 .byte 0
1058 .long 24 854 .long 24
1059 .long 0 855 .long 0
1060 .long 220 856 .long 220
1061 .long 220 857 .long 220
1062 .long 122 858 .long 122
1063 .long 0 # 859 .long 0 # BTF_KIND_FUNC_PROTO(id = 1)
1064 .long 218103808 # 860 .long 218103808 # 0xd000000
1065 .long 2 861 .long 2
1066 .long 83 # 862 .long 83 # BTF_KIND_INT(id = 2)
1067 .long 16777216 # 863 .long 16777216 # 0x1000000
1068 .long 4 864 .long 4
1069 .long 16777248 # 865 .long 16777248 # 0x1000020
1070 ...... 866 ......
1071 .byte 0 # 867 .byte 0 # string offset=0
1072 .ascii ".text" # 868 .ascii ".text" # string offset=1
1073 .byte 0 869 .byte 0
1074 .ascii "/home/yhs/tmp-pahole/t2. 870 .ascii "/home/yhs/tmp-pahole/t2.c" # string offset=7
1075 .byte 0 871 .byte 0
1076 .ascii "int main() { return 0; } 872 .ascii "int main() { return 0; }" # string offset=33
1077 .byte 0 873 .byte 0
1078 .ascii "int test() { return 0; } 874 .ascii "int test() { return 0; }" # string offset=58
1079 .byte 0 875 .byte 0
1080 .ascii "int" # 876 .ascii "int" # string offset=83
1081 ...... 877 ......
1082 .section .BTF.ext,"",@prog 878 .section .BTF.ext,"",@progbits
1083 .short 60319 # 879 .short 60319 # 0xeb9f
1084 .byte 1 880 .byte 1
1085 .byte 0 881 .byte 0
1086 .long 24 882 .long 24
1087 .long 0 883 .long 0
1088 .long 28 884 .long 28
1089 .long 28 885 .long 28
1090 .long 44 886 .long 44
1091 .long 8 # 887 .long 8 # FuncInfo
1092 .long 1 # 888 .long 1 # FuncInfo section string offset=1
1093 .long 2 889 .long 2
1094 .long .Lfunc_begin0 890 .long .Lfunc_begin0
1095 .long 3 891 .long 3
1096 .long .Lfunc_begin1 892 .long .Lfunc_begin1
1097 .long 5 893 .long 5
1098 .long 16 # 894 .long 16 # LineInfo
1099 .long 1 # 895 .long 1 # LineInfo section string offset=1
1100 .long 2 896 .long 2
1101 .long .Ltmp0 897 .long .Ltmp0
1102 .long 7 898 .long 7
1103 .long 33 899 .long 33
1104 .long 7182 # 900 .long 7182 # Line 7 Col 14
1105 .long .Ltmp3 901 .long .Ltmp3
1106 .long 7 902 .long 7
1107 .long 58 903 .long 58
1108 .long 8206 # 904 .long 8206 # Line 8 Col 14
1109 905
1110 7. Testing 906 7. Testing
1111 ========== !! 907 **********
1112 <<
1113 The kernel BPF selftest `tools/testing/selfte <<
1114 provides an extensive set of BTF-related test <<
1115 908
1116 .. Links !! 909 Kernel bpf selftest `test_btf.c` provides extensive set of BTF-related tests.
1117 .. _tools/testing/selftests/bpf/prog_tests/bt <<
1118 https://git.kernel.org/pub/scm/linux/kerne <<
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.