| 1 | |
| 2 | RFC: Common Trace Format (CTF) Proposal (pre-v1.7) |
| 3 | |
| 4 | Mathieu Desnoyers, EfficiOS Inc. |
| 5 | |
| 6 | The goal of the present document is to propose a trace format that suits the |
| 7 | needs of the embedded, telecom, high-performance and kernel communities. It is |
| 8 | based on the Common Trace Format Requirements (v1.4) document. It is designed to |
| 9 | allow traces to be natively generated by the Linux kernel, Linux user-space |
| 10 | applications written in C/C++, and hardware components. |
| 11 | |
| 12 | The latest version of this document can be found at: |
| 13 | |
| 14 | git tree: git://git.efficios.com/ctf.git |
| 15 | gitweb: http://git.efficios.com/?p=ctf.git |
| 16 | |
| 17 | A reference implementation of a library to read and write this trace format is |
| 18 | being implemented within the BabelTrace project, a converter between trace |
| 19 | formats. The development tree is available at: |
| 20 | |
| 21 | git tree: git://git.efficios.com/babeltrace.git |
| 22 | gitweb: http://git.efficios.com/?p=babeltrace.git |
| 23 | |
| 24 | |
| 25 | 1. Preliminary definitions |
| 26 | |
| 27 | - Event Trace: An ordered sequence of events. |
| 28 | - Event Stream: An ordered sequence of events, containing a subset of the |
| 29 | trace event types. |
| 30 | - Event Packet: A sequence of physically contiguous events within an event |
| 31 | stream. |
| 32 | - Event: This is the basic entry in a trace. (aka: a trace record). |
| 33 | - An event identifier (ID) relates to the class (a type) of event within |
| 34 | an event stream. |
| 35 | e.g. event: irq_entry. |
| 36 | - An event (or event record) relates to a specific instance of an event |
| 37 | class. |
| 38 | e.g. event: irq_entry, at time X, on CPU Y |
| 39 | - Source Architecture: Architecture writing the trace. |
| 40 | - Reader Architecture: Architecture reading the trace. |
| 41 | |
| 42 | |
| 43 | 2. High-level representation of a trace |
| 44 | |
| 45 | A trace is divided into multiple event streams. Each event stream contains a |
| 46 | subset of the trace event types. |
| 47 | |
| 48 | The final output of the trace, after its generation and optional transport over |
| 49 | the network, is expected to be either on permanent or temporary storage in a |
| 50 | virtual file system. Because each event stream is appended to while a trace is |
| 51 | being recorded, each is associated with a separate file for output. Therefore, |
| 52 | a stored trace can be represented as a directory containing one file per stream. |
| 53 | |
| 54 | A metadata event stream contains information on trace event types. It describes: |
| 55 | |
| 56 | - Trace version. |
| 57 | - Types available. |
| 58 | - Per-stream event header description. |
| 59 | - Per-stream event header selection. |
| 60 | - Per-stream event context fields. |
| 61 | - Per-event |
| 62 | - Event type to stream mapping. |
| 63 | - Event type to name mapping. |
| 64 | - Event type to ID mapping. |
| 65 | - Event fields description. |
| 66 | |
| 67 | |
| 68 | 3. Event stream |
| 69 | |
| 70 | An event stream is divided in contiguous event packets of variable size. These |
| 71 | subdivisions have a variable size. An event packet can contain a certain amount |
| 72 | of padding at the end. The rationale for the event stream design choices is |
| 73 | explained in Appendix B. Stream Header Rationale. |
| 74 | |
| 75 | An event stream is divided in contiguous event packets of variable size. These |
| 76 | subdivisions have a variable size. An event packet can contain a certain amount |
| 77 | of padding at the end. The stream header is repeated at the beginning of each |
| 78 | event packet. |
| 79 | |
| 80 | The event stream header will therefore be referred to as the "event packet |
| 81 | header" throughout the rest of this document. |
| 82 | |
| 83 | |
| 84 | 4. Types |
| 85 | |
| 86 | 4.1 Basic types |
| 87 | |
| 88 | A basic type is a scalar type, as described in this section. |
| 89 | |
| 90 | 4.1.1 Type inheritance |
| 91 | |
| 92 | Type specifications can be inherited to allow deriving types from a |
| 93 | type class. For example, see the uint32_t named type derived from the "integer" |
| 94 | type class below ("Integers" section). Types have a precise binary |
| 95 | representation in the trace. A type class has methods to read and write these |
| 96 | types, but must be derived into a type to be usable in an event field. |
| 97 | |
| 98 | 4.1.2 Alignment |
| 99 | |
| 100 | We define "byte-packed" types as aligned on the byte size, namely 8-bit. |
| 101 | We define "bit-packed" types as following on the next bit, as defined by the |
| 102 | "bitfields" section. |
| 103 | |
| 104 | All basic types, except bitfields, are either aligned on an architecture-defined |
| 105 | specific alignment or byte-packed, depending on the architecture preference. |
| 106 | Architectures providing fast unaligned write byte-packed basic types to save |
| 107 | space, aligning each type on byte boundaries (8-bit). Architectures with slow |
| 108 | unaligned writes align types on specific alignment values. If no specific |
| 109 | alignment is declared for a type nor its parents, it is assumed to be bit-packed |
| 110 | for bitfields and byte-packed for other types. |
| 111 | |
| 112 | Metadata attribute representation of a specific alignment: |
| 113 | |
| 114 | align = value; /* value in bits */ |
| 115 | |
| 116 | 4.1.3 Byte order |
| 117 | |
| 118 | By default, the native endianness of the source architecture the trace is used. |
| 119 | Byte order can be overridden for a basic type by specifying a "byte_order" |
| 120 | attribute. Typical use-case is to specify the network byte order (big endian: |
| 121 | "be") to save data captured from the network into the trace without conversion. |
| 122 | If not specified, the byte order is native. |
| 123 | |
| 124 | Metadata representation: |
| 125 | |
| 126 | byte_order = native OR network OR be OR le; /* network and be are aliases */ |
| 127 | |
| 128 | 4.1.4 Size |
| 129 | |
| 130 | Type size, in bits, for integers and floats is that returned by "sizeof()" in C |
| 131 | multiplied by CHAR_BIT. |
| 132 | We require the size of "char" and "unsigned char" types (CHAR_BIT) to be fixed |
| 133 | to 8 bits for cross-endianness compatibility. |
| 134 | |
| 135 | Metadata representation: |
| 136 | |
| 137 | size = value; (value is in bits) |
| 138 | |
| 139 | 4.1.5 Integers |
| 140 | |
| 141 | Signed integers are represented in two-complement. Integer alignment, size, |
| 142 | signedness and byte ordering are defined in the metadata. Integers aligned on |
| 143 | byte size (8-bit) and with length multiple of byte size (8-bit) correspond to |
| 144 | the C99 standard integers. In addition, integers with alignment and/or size that |
| 145 | are _not_ a multiple of the byte size are permitted; these correspond to the C99 |
| 146 | standard bitfields, with the added specification that the CTF integer bitfields |
| 147 | have a fixed binary representation. A MIT-licensed reference implementation of |
| 148 | the CTF portable bitfields is available at: |
| 149 | |
| 150 | http://git.efficios.com/?p=babeltrace.git;a=blob;f=include/babeltrace/bitfield.h |
| 151 | |
| 152 | Binary representation of integers: |
| 153 | |
| 154 | - On little and big endian: |
| 155 | - Within a byte, high bits correspond to an integer high bits, and low bits |
| 156 | correspond to low bits. |
| 157 | - On little endian: |
| 158 | - Integer across multiple bytes are placed from the less significant to the |
| 159 | most significant. |
| 160 | - Consecutive integers are placed from lower bits to higher bits (even within |
| 161 | a byte). |
| 162 | - On big endian: |
| 163 | - Integer across multiple bytes are placed from the most significant to the |
| 164 | less significant. |
| 165 | - Consecutive integers are placed from higher bits to lower bits (even within |
| 166 | a byte). |
| 167 | |
| 168 | This binary representation is derived from the bitfield implementation in GCC |
| 169 | for little and big endian. However, contrary to what GCC does, integers can |
| 170 | cross units boundaries (no padding is required). Padding can be explicitely |
| 171 | added (see 4.1.6 GNU/C bitfields) to follow the GCC layout if needed. |
| 172 | |
| 173 | Metadata representation: |
| 174 | |
| 175 | integer { |
| 176 | signed = true OR false; /* default false */ |
| 177 | byte_order = native OR network OR be OR le; /* default native */ |
| 178 | size = value; /* value in bits, no default */ |
| 179 | align = value; /* value in bits */ |
| 180 | } |
| 181 | |
| 182 | Example of type inheritance (creation of a uint32_t named type): |
| 183 | |
| 184 | typedef integer { |
| 185 | size = 32; |
| 186 | signed = false; |
| 187 | align = 32; |
| 188 | } uint32_t; |
| 189 | |
| 190 | Definition of a named 5-bit signed bitfield: |
| 191 | |
| 192 | typedef integer { |
| 193 | size = 5; |
| 194 | signed = true; |
| 195 | align = 1; |
| 196 | } int5_t; |
| 197 | |
| 198 | 4.1.6 GNU/C bitfields |
| 199 | |
| 200 | The GNU/C bitfields follow closely the integer representation, with a |
| 201 | particularity on alignment: if a bitfield cannot fit in the current unit, the |
| 202 | unit is padded and the bitfield starts at the following unit. The unit size is |
| 203 | defined by the size of the type "unit_type". |
| 204 | |
| 205 | Metadata representation: |
| 206 | |
| 207 | unit_type name:size: |
| 208 | |
| 209 | As an example, the following structure declared in C compiled by GCC: |
| 210 | |
| 211 | struct example { |
| 212 | short a:12; |
| 213 | short b:5; |
| 214 | }; |
| 215 | |
| 216 | The example structure is aligned on the largest element (short). The second |
| 217 | bitfield would be aligned on the next unit boundary, because it would not fit in |
| 218 | the current unit. |
| 219 | |
| 220 | 4.1.7 Floating point |
| 221 | |
| 222 | The floating point values byte ordering is defined in the metadata. |
| 223 | |
| 224 | Floating point values follow the IEEE 754-2008 standard interchange formats. |
| 225 | Description of the floating point values include the exponent and mantissa size |
| 226 | in bits. Some requirements are imposed on the floating point values: |
| 227 | |
| 228 | - FLT_RADIX must be 2. |
| 229 | - mant_dig is the number of digits represented in the mantissa. It is specified |
| 230 | by the ISO C99 standard, section 5.2.4, as FLT_MANT_DIG, DBL_MANT_DIG and |
| 231 | LDBL_MANT_DIG as defined by <float.h>. |
| 232 | - exp_dig is the number of digits represented in the exponent. Given that |
| 233 | mant_dig is one bit more than its actual size in bits (leading 1 is not |
| 234 | needed) and also given that the sign bit always takes one bit, exp_dig can be |
| 235 | specified as: |
| 236 | |
| 237 | - sizeof(float) * CHAR_BIT - FLT_MANT_DIG |
| 238 | - sizeof(double) * CHAR_BIT - DBL_MANT_DIG |
| 239 | - sizeof(long double) * CHAR_BIT - LDBL_MANT_DIG |
| 240 | |
| 241 | Metadata representation: |
| 242 | |
| 243 | floating_point { |
| 244 | exp_dig = value; |
| 245 | mant_dig = value; |
| 246 | byte_order = native OR network OR be OR le; |
| 247 | } |
| 248 | |
| 249 | Example of type inheritance: |
| 250 | |
| 251 | typedef floating_point { |
| 252 | exp_dig = 8; /* sizeof(float) * CHAR_BIT - FLT_MANT_DIG */ |
| 253 | mant_dig = 24; /* FLT_MANT_DIG */ |
| 254 | byte_order = native; |
| 255 | } float; |
| 256 | |
| 257 | TODO: define NaN, +inf, -inf behavior. |
| 258 | |
| 259 | 4.1.8 Enumerations |
| 260 | |
| 261 | Enumerations are a mapping between an integer type and a table of strings. The |
| 262 | numerical representation of the enumeration follows the integer type specified |
| 263 | by the metadata. The enumeration mapping table is detailed in the enumeration |
| 264 | description within the metadata. The mapping table maps inclusive value ranges |
| 265 | (or single values) to strings. Instead of being limited to simple |
| 266 | "value -> string" mappings, these enumerations map |
| 267 | "[ start_value ... end_value ] -> string", which map inclusive ranges of |
| 268 | values to strings. An enumeration from the C language can be represented in |
| 269 | this format by having the same start_value and end_value for each element, which |
| 270 | is in fact a range of size 1. This single-value range is supported without |
| 271 | repeating the start and end values with the value = string declaration. |
| 272 | |
| 273 | If a numeric value is encountered between < >, it represents the integer type |
| 274 | size used to hold the enumeration, in bits. |
| 275 | |
| 276 | enum <integer_type OR size> name { |
| 277 | string = start_value1 ... end_value1, |
| 278 | "other string" = start_value2 ... end_value2, |
| 279 | yet_another_string, /* will be assigned to end_value2 + 1 */ |
| 280 | "some other string" = value, |
| 281 | ... |
| 282 | }; |
| 283 | |
| 284 | If the values are omitted, the enumeration starts at 0 and increment of 1 for |
| 285 | each entry: |
| 286 | |
| 287 | enum <32> name { |
| 288 | ZERO, |
| 289 | ONE, |
| 290 | TWO, |
| 291 | TEN = 10, |
| 292 | ELEVEN, |
| 293 | }; |
| 294 | |
| 295 | Overlapping ranges within a single enumeration are implementation defined. |
| 296 | |
| 297 | A nameless enumeration can be declared as a field type or as part of a typedef: |
| 298 | |
| 299 | enum <integer_type> { |
| 300 | ... |
| 301 | } |
| 302 | |
| 303 | 4.2 Compound types |
| 304 | |
| 305 | 4.2.1 Structures |
| 306 | |
| 307 | Structures are aligned on the largest alignment required by basic types |
| 308 | contained within the structure. (This follows the ISO/C standard for structures) |
| 309 | |
| 310 | Metadata representation of a named structure: |
| 311 | |
| 312 | struct name { |
| 313 | field_type field_name; |
| 314 | field_type field_name; |
| 315 | ... |
| 316 | }; |
| 317 | |
| 318 | Example: |
| 319 | |
| 320 | struct example { |
| 321 | integer { /* Nameless type */ |
| 322 | size = 16; |
| 323 | signed = true; |
| 324 | align = 16; |
| 325 | } first_field_name; |
| 326 | uint64_t second_field_name; /* Named type declared in the metadata */ |
| 327 | }; |
| 328 | |
| 329 | The fields are placed in a sequence next to each other. They each possess a |
| 330 | field name, which is a unique identifier within the structure. |
| 331 | |
| 332 | A nameless structure can be declared as a field type or as part of a typedef: |
| 333 | |
| 334 | struct { |
| 335 | ... |
| 336 | } |
| 337 | |
| 338 | 4.2.2 Variants (Discriminated Unions) |
| 339 | |
| 340 | A CTF variant is a selection between different types. A CTF variant must always |
| 341 | be defined within the scope of a structure or within fields contained within a |
| 342 | structure (defined recursively). A "tag" enumeration field must appear in either |
| 343 | the same lexical scope or an uppermost scope, prior to the variant field (in |
| 344 | field declaration order). The type selection is indicated by the mapping from |
| 345 | the enumeration value to the string used as variant type selector. The field to |
| 346 | use as tag is specified by the "tag_field", specified between "< >" after the |
| 347 | "variant" keyword for unnamed variants, and after "variant name" for named |
| 348 | variants. |
| 349 | |
| 350 | The alignment of the variant is the alignment of the type as selected by the tag |
| 351 | value for the specific instance of the variant. The alignment of the type |
| 352 | containing the variant is independent of the variant alignment. The size of the |
| 353 | variant is the size as selected by the tag value for the specific instance of |
| 354 | the variant. |
| 355 | |
| 356 | A named variant declaration followed by its definition within a structure |
| 357 | declaration: |
| 358 | |
| 359 | variant name { |
| 360 | field_type sel1; |
| 361 | field_type sel2; |
| 362 | field_type sel3; |
| 363 | ... |
| 364 | }; |
| 365 | |
| 366 | struct { |
| 367 | enum <integer_type or size> { sel1, sel2, sel3, ... } tag_field; |
| 368 | ... |
| 369 | variant name <tag_field> v; |
| 370 | } |
| 371 | |
| 372 | An unnamed variant definition within a structure is expressed by the following |
| 373 | metadata: |
| 374 | |
| 375 | struct { |
| 376 | enum <integer_type or size> { sel1, sel2, sel3, ... } tag_field; |
| 377 | ... |
| 378 | variant <tag_field> { |
| 379 | field_type sel1; |
| 380 | field_type sel2; |
| 381 | field_type sel3; |
| 382 | ... |
| 383 | } v; |
| 384 | } |
| 385 | |
| 386 | Example of a named variant within a sequence that refers to a single tag field: |
| 387 | |
| 388 | variant example { |
| 389 | uint32_t a; |
| 390 | uint64_t b; |
| 391 | short c; |
| 392 | }; |
| 393 | |
| 394 | struct { |
| 395 | enum <uint2_t> { a, b, c } choice; |
| 396 | variant example <choice> v[unsigned int]; |
| 397 | } |
| 398 | |
| 399 | Example of an unnamed variant: |
| 400 | |
| 401 | struct { |
| 402 | enum <uint2_t> { a, b, c, d } choice; |
| 403 | /* Unrelated fields can be added between the variant and its tag */ |
| 404 | int32_t somevalue; |
| 405 | variant <choice> { |
| 406 | uint32_t a; |
| 407 | uint64_t b; |
| 408 | short c; |
| 409 | struct { |
| 410 | unsigned int field1; |
| 411 | uint64_t field2; |
| 412 | } d; |
| 413 | } s; |
| 414 | } |
| 415 | |
| 416 | Example of an unnamed variant within an array: |
| 417 | |
| 418 | struct { |
| 419 | enum <uint2_t> { a, b, c } choice; |
| 420 | variant <choice> { |
| 421 | uint32_t a; |
| 422 | uint64_t b; |
| 423 | short c; |
| 424 | } v[10]; |
| 425 | } |
| 426 | |
| 427 | Example of a variant type definition within a structure, where the defined type |
| 428 | is then declared within an array of structures. This variant refers to a tag |
| 429 | located in an upper lexical scope. This example clearly shows that a variant |
| 430 | type definition referring to the tag "x" uses the closest preceding field from |
| 431 | the lexical scope of the type definition. |
| 432 | |
| 433 | struct { |
| 434 | enum <uint2_t> { a, b, c, d } x; |
| 435 | |
| 436 | typedef variant <x> { /* |
| 437 | * "x" refers to the preceding "x" enumeration in the |
| 438 | * lexical scope of the type definition. |
| 439 | */ |
| 440 | uint32_t a; |
| 441 | uint64_t b; |
| 442 | short c; |
| 443 | } example_variant; |
| 444 | |
| 445 | struct { |
| 446 | enum <int> { x, y, z } x; /* This enumeration is not used by "v". */ |
| 447 | example_variant v; /* |
| 448 | * "v" uses the "enum <uint2_t> { a, b, c, d }" |
| 449 | * tag. |
| 450 | */ |
| 451 | } a[10]; |
| 452 | } |
| 453 | |
| 454 | 4.2.3 Arrays |
| 455 | |
| 456 | Arrays are fixed-length. Their length is declared in the type declaration within |
| 457 | the metadata. They contain an array of "inner type" elements, which can refer to |
| 458 | any type not containing the type of the array being declared (no circular |
| 459 | dependency). The length is the number of elements in an array. |
| 460 | |
| 461 | Metadata representation of a named array: |
| 462 | |
| 463 | typedef elem_type name[length]; |
| 464 | |
| 465 | A nameless array can be declared as a field type within a structure, e.g.: |
| 466 | |
| 467 | uint8_t field_name[10]; |
| 468 | |
| 469 | |
| 470 | 4.2.4 Sequences |
| 471 | |
| 472 | Sequences are dynamically-sized arrays. They start with an integer that specify |
| 473 | the length of the sequence, followed by an array of "inner type" elements. |
| 474 | The length is the number of elements in the sequence. |
| 475 | |
| 476 | Metadata representation for a named sequence: |
| 477 | |
| 478 | typedef elem_type name[length_type]; |
| 479 | |
| 480 | A nameless sequence can be declared as a field type, e.g.: |
| 481 | |
| 482 | long field_name[int]; |
| 483 | |
| 484 | The length type follows the integer types specifications, and the sequence |
| 485 | elements follow the "array" specifications. |
| 486 | |
| 487 | 4.2.5 Strings |
| 488 | |
| 489 | Strings are an array of bytes of variable size and are terminated by a '\0' |
| 490 | "NULL" character. Their encoding is described in the metadata. In absence of |
| 491 | encoding attribute information, the default encoding is UTF-8. |
| 492 | |
| 493 | Metadata representation of a named string type: |
| 494 | |
| 495 | typedef string { |
| 496 | encoding = UTF8 OR ASCII; |
| 497 | } name; |
| 498 | |
| 499 | A nameless string type can be declared as a field type: |
| 500 | |
| 501 | string field_name; /* Use default UTF8 encoding */ |
| 502 | |
| 503 | 5. Event Packet Header |
| 504 | |
| 505 | The event packet header consists of two part: one is mandatory and have a fixed |
| 506 | layout. The second part, the "event packet context", has its layout described in |
| 507 | the metadata. |
| 508 | |
| 509 | - Aligned on page size. Fixed size. Fields either aligned or packed (depending |
| 510 | on the architecture preference). |
| 511 | No padding at the end of the event packet header. Native architecture byte |
| 512 | ordering. |
| 513 | |
| 514 | Fixed layout (event packet header): |
| 515 | |
| 516 | - Magic number (CTF magic numbers: 0xC1FC1FC1 and its reverse endianness |
| 517 | representation: 0xC11FFCC1) It needs to have a non-symmetric bytewise |
| 518 | representation. Used to distinguish between big and little endian traces (this |
| 519 | information is determined by knowing the endianness of the architecture |
| 520 | reading the trace and comparing the magic number against its value and the |
| 521 | reverse, 0xC11FFCC1). This magic number specifies that we use the CTF metadata |
| 522 | description language described in this document. Different magic numbers |
| 523 | should be used for other metadata description languages. |
| 524 | - Trace UUID, used to ensure the event packet match the metadata used. |
| 525 | (note: we cannot use a metadata checksum because metadata can be appended to |
| 526 | while tracing is active) |
| 527 | - Stream ID, used as reference to stream description in metadata. |
| 528 | |
| 529 | Metadata-defined layout (event packet context): |
| 530 | |
| 531 | - Event packet content size (in bytes). |
| 532 | - Event packet size (in bytes, includes padding). |
| 533 | - Event packet content checksum (optional). Checksum excludes the event packet |
| 534 | header. |
| 535 | - Per-stream event packet sequence count (to deal with UDP packet loss). The |
| 536 | number of significant sequence counter bits should also be present, so |
| 537 | wrap-arounds are deal with correctly. |
| 538 | - Timestamp at the beginning and timestamp at the end of the event packet. |
| 539 | Both timestamps are written in the packet header, but sampled respectively |
| 540 | while (or before) writing the first event and while (or after) writing the |
| 541 | last event in the packet. The inclusive range between these timestamps should |
| 542 | include all event timestamps assigned to events contained within the packet. |
| 543 | - Events discarded count |
| 544 | - Snapshot of a per-stream free-running counter, counting the number of |
| 545 | events discarded that were supposed to be written in the stream prior to |
| 546 | the first event in the event packet. |
| 547 | * Note: producer-consumer buffer full condition should fill the current |
| 548 | event packet with padding so we know exactly where events have been |
| 549 | discarded. |
| 550 | - Lossless compression scheme used for the event packet content. Applied |
| 551 | directly to raw data. New types of compression can be added in following |
| 552 | versions of the format. |
| 553 | 0: no compression scheme |
| 554 | 1: bzip2 |
| 555 | 2: gzip |
| 556 | 3: xz |
| 557 | - Cypher used for the event packet content. Applied after compression. |
| 558 | 0: no encryption |
| 559 | 1: AES |
| 560 | - Checksum scheme used for the event packet content. Applied after encryption. |
| 561 | 0: no checksum |
| 562 | 1: md5 |
| 563 | 2: sha1 |
| 564 | 3: crc32 |
| 565 | |
| 566 | 5.1 Event Packet Header Fixed Layout Description |
| 567 | |
| 568 | struct event_packet_header { |
| 569 | uint32_t magic; |
| 570 | uint8_t trace_uuid[16]; |
| 571 | uint32_t stream_id; |
| 572 | }; |
| 573 | |
| 574 | 5.2 Event Packet Context Description |
| 575 | |
| 576 | Event packet context example. These are declared within the stream declaration |
| 577 | in the metadata. All these fields are optional except for "content_size" and |
| 578 | "packet_size", which must be present in the context. |
| 579 | |
| 580 | An example event packet context type: |
| 581 | |
| 582 | struct event_packet_context { |
| 583 | uint64_t timestamp_begin; |
| 584 | uint64_t timestamp_end; |
| 585 | uint32_t checksum; |
| 586 | uint32_t stream_packet_count; |
| 587 | uint32_t events_discarded; |
| 588 | uint32_t cpu_id; |
| 589 | uint32_t/uint16_t content_size; |
| 590 | uint32_t/uint16_t packet_size; |
| 591 | uint8_t stream_packet_count_bits; /* Significant counter bits */ |
| 592 | uint8_t compression_scheme; |
| 593 | uint8_t encryption_scheme; |
| 594 | uint8_t checksum; |
| 595 | }; |
| 596 | |
| 597 | |
| 598 | 6. Event Structure |
| 599 | |
| 600 | The overall structure of an event is: |
| 601 | |
| 602 | 1 - Stream Packet Context (as specified by the stream metadata) |
| 603 | 2 - Event Header (as specifed by the stream metadata) |
| 604 | 3 - Stream Event Context (as specified by the stream metadata) |
| 605 | 4 - Event Context (as specified by the event metadata) |
| 606 | 5 - Event Payload (as specified by the event metadata) |
| 607 | |
| 608 | 6.1 Lexical Scope |
| 609 | |
| 610 | The lexical scope of each structure (stream packet context, header, stream event |
| 611 | context, event context and payload) is extended in the following way: lower |
| 612 | levels (e.g. 3) can refer to fields defined in prior levels (e.g. 2 and 1). The |
| 613 | field in the closest level has priority in case of field name conflict. |
| 614 | |
| 615 | This allows, for instance, the event context to define a variant refering to the |
| 616 | "id" field of the event header as selector. |
| 617 | |
| 618 | 6.2 Event Header |
| 619 | |
| 620 | Event headers can be described within the metadata. We hereby propose, as an |
| 621 | example, two types of events headers. Type 1 accommodates streams with less than |
| 622 | 31 event IDs. Type 2 accommodates streams with 31 or more event IDs. |
| 623 | |
| 624 | One major factor can vary between streams: the number of event IDs assigned to |
| 625 | a stream. Luckily, this information tends to stay relatively constant (modulo |
| 626 | event registration while trace is being recorded), so we can specify different |
| 627 | representations for streams containing few event IDs and streams containing |
| 628 | many event IDs, so we end up representing the event ID and timestamp as densely |
| 629 | as possible in each case. |
| 630 | |
| 631 | The header is extended in the rare occasions where the information cannot be |
| 632 | represented in the ranges available in the standard event header. They are also |
| 633 | used in the rare occasions where the data required for a field could not be |
| 634 | collected: the flag corresponding to the missing field within the missing_fields |
| 635 | array is then set to 1. |
| 636 | |
| 637 | Types uintX_t represent an X-bit unsigned integer. |
| 638 | |
| 639 | |
| 640 | 6.2.1 Type 1 - Few event IDs |
| 641 | |
| 642 | - Aligned on 32-bit (or 8-bit if byte-packed, depending on the architecture |
| 643 | preference). |
| 644 | - Native architecture byte ordering. |
| 645 | - For "compact" selection |
| 646 | - Fixed size: 32 bits. |
| 647 | - For "extended" selection |
| 648 | - Size depends on the architecture and variant alignment. |
| 649 | |
| 650 | struct event_header_1 { |
| 651 | /* |
| 652 | * id: range: 0 - 30. |
| 653 | * id 31 is reserved to indicate an extended header. |
| 654 | */ |
| 655 | enum <uint5_t> { compact = 0 ... 30, extended = 31 } id; |
| 656 | variant <id> { |
| 657 | struct { |
| 658 | uint27_t timestamp; |
| 659 | } compact; |
| 660 | struct { |
| 661 | uint32_t id; /* 32-bit event IDs */ |
| 662 | uint64_t timestamp; /* 64-bit timestamps */ |
| 663 | } extended; |
| 664 | } v; |
| 665 | }; |
| 666 | |
| 667 | |
| 668 | 6.2.2 Type 2 - Many event IDs |
| 669 | |
| 670 | - Aligned on 16-bit (or 8-bit if byte-packed, depending on the architecture |
| 671 | preference). |
| 672 | - Native architecture byte ordering. |
| 673 | - For "compact" selection |
| 674 | - Size depends on the architecture and variant alignment. |
| 675 | - For "extended" selection |
| 676 | - Size depends on the architecture and variant alignment. |
| 677 | |
| 678 | struct event_header_2 { |
| 679 | /* |
| 680 | * id: range: 0 - 65534. |
| 681 | * id 65535 is reserved to indicate an extended header. |
| 682 | */ |
| 683 | enum <uint16_t> { compact = 0 ... 65534, extended = 65535 } id; |
| 684 | variant <id> { |
| 685 | struct { |
| 686 | uint32_t timestamp; |
| 687 | } compact; |
| 688 | struct { |
| 689 | uint32_t id; /* 32-bit event IDs */ |
| 690 | uint64_t timestamp; /* 64-bit timestamps */ |
| 691 | } extended; |
| 692 | } v; |
| 693 | }; |
| 694 | |
| 695 | |
| 696 | 6.2 Event Context |
| 697 | |
| 698 | The event context contains information relative to the current event. The choice |
| 699 | and meaning of this information is specified by the metadata "stream" and |
| 700 | "event" information. The "stream" context is applied to all events within the |
| 701 | stream. The "stream" context structure follows the event header. The "event" |
| 702 | context is applied to specific events. Its structure follows the "stream" |
| 703 | context stucture. |
| 704 | |
| 705 | An example of stream-level event context is to save the event payload size with |
| 706 | each event, or to save the current PID with each event. These are declared |
| 707 | within the stream declaration within the metadata: |
| 708 | |
| 709 | stream { |
| 710 | ... |
| 711 | event { |
| 712 | ... |
| 713 | context = struct { |
| 714 | uint pid; |
| 715 | uint16_t payload_size; |
| 716 | }; |
| 717 | } |
| 718 | }; |
| 719 | |
| 720 | An example of event-specific event context is to declare a bitmap of missing |
| 721 | fields, only appended after the stream event context if the extended event |
| 722 | header is selected. NR_FIELDS is the number of fields within the event (a |
| 723 | numeric value). |
| 724 | |
| 725 | event { |
| 726 | context = struct { |
| 727 | variant <id> { |
| 728 | struct { } compact; |
| 729 | struct { |
| 730 | uint1_t missing_fields[NR_FIELDS]; /* missing event fields bitmap */ |
| 731 | } extended; |
| 732 | } v; |
| 733 | }; |
| 734 | ... |
| 735 | } |
| 736 | |
| 737 | 6.3 Event Payload |
| 738 | |
| 739 | An event payload contains fields specific to a given event type. The fields |
| 740 | belonging to an event type are described in the event-specific metadata |
| 741 | within a structure type. |
| 742 | |
| 743 | 6.3.1 Padding |
| 744 | |
| 745 | No padding at the end of the event payload. This differs from the ISO/C standard |
| 746 | for structures, but follows the CTF standard for structures. In a trace, even |
| 747 | though it makes sense to align the beginning of a structure, it really makes no |
| 748 | sense to add padding at the end of the structure, because structures are usually |
| 749 | not followed by a structure of the same type. |
| 750 | |
| 751 | This trick can be done by adding a zero-length "end" field at the end of the C |
| 752 | structures, and by using the offset of this field rather than using sizeof() |
| 753 | when calculating the size of a structure (see Appendix "A. Helper macros"). |
| 754 | |
| 755 | 6.3.2 Alignment |
| 756 | |
| 757 | The event payload is aligned on the largest alignment required by types |
| 758 | contained within the payload. (This follows the ISO/C standard for structures) |
| 759 | |
| 760 | |
| 761 | 7. Metadata |
| 762 | |
| 763 | The meta-data is located in a stream named "metadata". It is made of "event |
| 764 | packets", which each start with an event packet header. The event type within |
| 765 | the metadata stream have no event header nor event context. Each event only |
| 766 | contains a null-terminated "string" payload, which is a metadata description |
| 767 | entry. The events are packed one next to another. Each event packet start with |
| 768 | an event packet header, which contains, amongst other fields, the magic number |
| 769 | and trace UUID. |
| 770 | |
| 771 | The metadata can be parsed by reading through the metadata strings, skipping |
| 772 | newlines and null-characters. Type names are made of a single identifier, and |
| 773 | can be surrounded by prefix/postfix. Text contained within "/*" and "*/", as |
| 774 | well as within "//" and end of line, are treated as comments. |
| 775 | |
| 776 | The grammar representing the CTF metadata is presented in |
| 777 | Appendix C. CTF Metadata Grammar. |
| 778 | |
| 779 | trace { |
| 780 | major = value; /* Trace format version */ |
| 781 | minor = value; |
| 782 | uuid = value; /* Trace UUID */ |
| 783 | word_size = value; |
| 784 | }; |
| 785 | |
| 786 | stream { |
| 787 | id = stream_id; |
| 788 | event { |
| 789 | header_alignment = value; |
| 790 | /* Type 1 - Few event IDs; Type 2 - Many event IDs. See section 6.2. */ |
| 791 | header = event_header_1 OR event_header_2; |
| 792 | context = struct { |
| 793 | ... |
| 794 | }; |
| 795 | }; |
| 796 | packet { |
| 797 | context = struct { |
| 798 | ... |
| 799 | }; |
| 800 | }; |
| 801 | }; |
| 802 | |
| 803 | event { |
| 804 | name = event_name; |
| 805 | id = value; /* Numeric identifier within the stream */ |
| 806 | stream = stream_id; |
| 807 | context = struct { |
| 808 | ... |
| 809 | }; |
| 810 | fields = struct { |
| 811 | ... |
| 812 | }; |
| 813 | }; |
| 814 | |
| 815 | /* More detail on types in section 4. Types */ |
| 816 | |
| 817 | /* |
| 818 | * Named types: |
| 819 | * |
| 820 | * Type declarations behave similarly to the C standard, with the following |
| 821 | * added feature: new_type can be preceded by a colon to allow creation of a |
| 822 | * type name with prefix/postfix. |
| 823 | */ |
| 824 | |
| 825 | typedef aliased_type_prefix aliased_type new_type aliased_type_postfix; |
| 826 | |
| 827 | /* e.g.: typedef struct example new_type_name[10]; */ |
| 828 | |
| 829 | typedef type_class { |
| 830 | ... |
| 831 | } : new_type_prefix new_type new_type_postfix; |
| 832 | |
| 833 | /* |
| 834 | * e.g.: |
| 835 | * typedef integer { |
| 836 | * size = 32; |
| 837 | * align = 32; |
| 838 | * signed = false; |
| 839 | * } : struct page *; |
| 840 | */ |
| 841 | |
| 842 | struct name { |
| 843 | ... |
| 844 | }; |
| 845 | |
| 846 | variant name { |
| 847 | ... |
| 848 | }; |
| 849 | |
| 850 | enum <integer_type or size> name { |
| 851 | ... |
| 852 | }; |
| 853 | |
| 854 | |
| 855 | /* Unnamed types, contained within compound type fields or typedef. */ |
| 856 | |
| 857 | struct { |
| 858 | ... |
| 859 | } |
| 860 | |
| 861 | variant { |
| 862 | ... |
| 863 | } |
| 864 | |
| 865 | enum <integer_type or size> { |
| 866 | ... |
| 867 | } |
| 868 | |
| 869 | typedef type new_type[length]; |
| 870 | |
| 871 | struct { |
| 872 | type field_name[length]; |
| 873 | } |
| 874 | |
| 875 | typedef type new_type[length_type]; |
| 876 | |
| 877 | struct { |
| 878 | type field_name[length_type]; |
| 879 | } |
| 880 | |
| 881 | integer { |
| 882 | ... |
| 883 | } |
| 884 | |
| 885 | floating_point { |
| 886 | ... |
| 887 | } |
| 888 | |
| 889 | struct { |
| 890 | integer_type field_name:size; /* GNU/C bitfield */ |
| 891 | } |
| 892 | |
| 893 | struct { |
| 894 | string field_name; |
| 895 | } |
| 896 | |
| 897 | |
| 898 | A. Helper macros |
| 899 | |
| 900 | The two following macros keep track of the size of a GNU/C structure without |
| 901 | padding at the end by placing HEADER_END as the last field. A one byte end field |
| 902 | is used for C90 compatibility (C99 flexible arrays could be used here). Note |
| 903 | that this does not affect the effective structure size, which should always be |
| 904 | calculated with the header_sizeof() helper. |
| 905 | |
| 906 | #define HEADER_END char end_field |
| 907 | #define header_sizeof(type) offsetof(typeof(type), end_field) |
| 908 | |
| 909 | |
| 910 | B. Stream Header Rationale |
| 911 | |
| 912 | An event stream is divided in contiguous event packets of variable size. These |
| 913 | subdivisions allow the trace analyzer to perform a fast binary search by time |
| 914 | within the stream (typically requiring to index only the event packet headers) |
| 915 | without reading the whole stream. These subdivisions have a variable size to |
| 916 | eliminate the need to transfer the event packet padding when partially filled |
| 917 | event packets must be sent when streaming a trace for live viewing/analysis. |
| 918 | An event packet can contain a certain amount of padding at the end. Dividing |
| 919 | streams into event packets is also useful for network streaming over UDP and |
| 920 | flight recorder mode tracing (a whole event packet can be swapped out of the |
| 921 | buffer atomically for reading). |
| 922 | |
| 923 | The stream header is repeated at the beginning of each event packet to allow |
| 924 | flexibility in terms of: |
| 925 | |
| 926 | - streaming support, |
| 927 | - allowing arbitrary buffers to be discarded without making the trace |
| 928 | unreadable, |
| 929 | - allow UDP packet loss handling by either dealing with missing event packet |
| 930 | or asking for re-transmission. |
| 931 | - transparently support flight recorder mode, |
| 932 | - transparently support crash dump. |
| 933 | |
| 934 | The event stream header will therefore be referred to as the "event packet |
| 935 | header" throughout the rest of this document. |
| 936 | |
| 937 | C. CTF Metadata Grammar |
| 938 | |
| 939 | TODO |