2 RFC: Common Trace Format (CTF) Proposal (pre-v1.7)
4 Mathieu Desnoyers, EfficiOS Inc.
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. One major element of
11 CTF is the Trace Stream Description Language (TSDL) which flexibility
12 enables description of various binary trace stream layouts.
14 The latest version of this document can be found at:
16 git tree: git://git.efficios.com/ctf.git
17 gitweb: http://git.efficios.com/?p=ctf.git
19 A reference implementation of a library to read and write this trace format is
20 being implemented within the BabelTrace project, a converter between trace
21 formats. The development tree is available at:
23 git tree: git://git.efficios.com/babeltrace.git
24 gitweb: http://git.efficios.com/?p=babeltrace.git
27 1. Preliminary definitions
29 - Event Trace: An ordered sequence of events.
30 - Event Stream: An ordered sequence of events, containing a subset of the
32 - Event Packet: A sequence of physically contiguous events within an event
34 - Event: This is the basic entry in a trace. (aka: a trace record).
35 - An event identifier (ID) relates to the class (a type) of event within
37 e.g. event: irq_entry.
38 - An event (or event record) relates to a specific instance of an event
40 e.g. event: irq_entry, at time X, on CPU Y
41 - Source Architecture: Architecture writing the trace.
42 - Reader Architecture: Architecture reading the trace.
45 2. High-level representation of a trace
47 A trace is divided into multiple event streams. Each event stream contains a
48 subset of the trace event types.
50 The final output of the trace, after its generation and optional transport over
51 the network, is expected to be either on permanent or temporary storage in a
52 virtual file system. Because each event stream is appended to while a trace is
53 being recorded, each is associated with a separate file for output. Therefore,
54 a stored trace can be represented as a directory containing one file per stream.
56 Meta-data description associated with the trace contains information on
57 trace event types expressed in the Trace Stream Description Language
58 (TSDL). This language describes:
62 - Per-trace event header description.
63 - Per-stream event header description.
64 - Per-stream event context description.
66 - Event type to stream mapping.
67 - Event type to name mapping.
68 - Event type to ID mapping.
69 - Event context description.
70 - Event fields description.
75 An event stream can be divided into contiguous event packets of variable
76 size. These subdivisions have a variable size. An event packet can
77 contain a certain amount of padding at the end. The stream header is
78 repeated at the beginning of each event packet. The rationale for the
79 event stream design choices is explained in Appendix B. Stream Header
82 The event stream header will therefore be referred to as the "event packet
83 header" throughout the rest of this document.
88 Types are organized as type classes. Each type class belong to either of two
89 kind of types: basic types or compound types.
93 A basic type is a scalar type, as described in this section. It includes
94 integers, GNU/C bitfields, enumerations, and floating point values.
96 4.1.1 Type inheritance
98 Type specifications can be inherited to allow deriving types from a
99 type class. For example, see the uint32_t named type derived from the "integer"
100 type class below ("Integers" section). Types have a precise binary
101 representation in the trace. A type class has methods to read and write these
102 types, but must be derived into a type to be usable in an event field.
106 We define "byte-packed" types as aligned on the byte size, namely 8-bit.
107 We define "bit-packed" types as following on the next bit, as defined by the
110 Each basic type must specify its alignment, in bits. Examples of
111 possible alignments are: bit-packed (align = 1), byte-packed (align =
112 8), or word-aligned (e.g. align = 32 or align = 64). The choice depends
113 on the architecture preference and compactness vs performance trade-offs
114 of the implementation. Architectures providing fast unaligned write
115 byte-packed basic types to save space, aligning each type on byte
116 boundaries (8-bit). Architectures with slow unaligned writes align types
117 on specific alignment values. If no specific alignment is declared for a
118 type, it is assumed to be bit-packed for integers with size not multiple
119 of 8 bits and for gcc bitfields. All other basic types are byte-packed
120 by default. It is however recommended to always specify the alignment
121 explicitly. Alignment values must be power of two. Compound types are
122 aligned as specified in their individual specification.
124 TSDL meta-data attribute representation of a specific alignment:
126 align = value; /* value in bits */
130 By default, the native endianness of the source architecture the trace is used.
131 Byte order can be overridden for a basic type by specifying a "byte_order"
132 attribute. Typical use-case is to specify the network byte order (big endian:
133 "be") to save data captured from the network into the trace without conversion.
134 If not specified, the byte order is native.
136 TSDL meta-data representation:
138 byte_order = native OR network OR be OR le; /* network and be are aliases */
142 Type size, in bits, for integers and floats is that returned by "sizeof()" in C
143 multiplied by CHAR_BIT.
144 We require the size of "char" and "unsigned char" types (CHAR_BIT) to be fixed
145 to 8 bits for cross-endianness compatibility.
147 TSDL meta-data representation:
149 size = value; (value is in bits)
153 Signed integers are represented in two-complement. Integer alignment,
154 size, signedness and byte ordering are defined in the TSDL meta-data.
155 Integers aligned on byte size (8-bit) and with length multiple of byte
156 size (8-bit) correspond to the C99 standard integers. In addition,
157 integers with alignment and/or size that are _not_ a multiple of the
158 byte size are permitted; these correspond to the C99 standard bitfields,
159 with the added specification that the CTF integer bitfields have a fixed
160 binary representation. A MIT-licensed reference implementation of the
161 CTF portable bitfields is available at:
163 http://git.efficios.com/?p=babeltrace.git;a=blob;f=include/babeltrace/bitfield.h
165 Binary representation of integers:
167 - On little and big endian:
168 - Within a byte, high bits correspond to an integer high bits, and low bits
169 correspond to low bits.
171 - Integer across multiple bytes are placed from the less significant to the
173 - Consecutive integers are placed from lower bits to higher bits (even within
176 - Integer across multiple bytes are placed from the most significant to the
178 - Consecutive integers are placed from higher bits to lower bits (even within
181 This binary representation is derived from the bitfield implementation in GCC
182 for little and big endian. However, contrary to what GCC does, integers can
183 cross units boundaries (no padding is required). Padding can be explicitly
184 added (see 4.1.6 GNU/C bitfields) to follow the GCC layout if needed.
186 TSDL meta-data representation:
189 signed = true OR false; /* default false */
190 byte_order = native OR network OR be OR le; /* default native */
191 size = value; /* value in bits, no default */
192 align = value; /* value in bits */
193 /* based used for pretty-printing output, default: decimal. */
194 base = decimal OR dec OR OR d OR i OR u OR 10 OR hexadecimal OR hex OR x OR X OR p OR 16
195 OR octal OR oct OR o OR 8 OR binary OR b OR 2;
198 Example of type inheritance (creation of a uint32_t named type):
206 Definition of a named 5-bit signed bitfield:
214 4.1.6 GNU/C bitfields
216 The GNU/C bitfields follow closely the integer representation, with a
217 particularity on alignment: if a bitfield cannot fit in the current unit, the
218 unit is padded and the bitfield starts at the following unit. The unit size is
219 defined by the size of the type "unit_type".
221 TSDL meta-data representation:
225 As an example, the following structure declared in C compiled by GCC:
232 The example structure is aligned on the largest element (short). The second
233 bitfield would be aligned on the next unit boundary, because it would not fit in
238 The floating point values byte ordering is defined in the TSDL meta-data.
240 Floating point values follow the IEEE 754-2008 standard interchange formats.
241 Description of the floating point values include the exponent and mantissa size
242 in bits. Some requirements are imposed on the floating point values:
244 - FLT_RADIX must be 2.
245 - mant_dig is the number of digits represented in the mantissa. It is specified
246 by the ISO C99 standard, section 5.2.4, as FLT_MANT_DIG, DBL_MANT_DIG and
247 LDBL_MANT_DIG as defined by <float.h>.
248 - exp_dig is the number of digits represented in the exponent. Given that
249 mant_dig is one bit more than its actual size in bits (leading 1 is not
250 needed) and also given that the sign bit always takes one bit, exp_dig can be
253 - sizeof(float) * CHAR_BIT - FLT_MANT_DIG
254 - sizeof(double) * CHAR_BIT - DBL_MANT_DIG
255 - sizeof(long double) * CHAR_BIT - LDBL_MANT_DIG
257 TSDL meta-data representation:
262 byte_order = native OR network OR be OR le;
266 Example of type inheritance:
268 typealias floating_point {
269 exp_dig = 8; /* sizeof(float) * CHAR_BIT - FLT_MANT_DIG */
270 mant_dig = 24; /* FLT_MANT_DIG */
275 TODO: define NaN, +inf, -inf behavior.
277 Bit-packed, byte-packed or larger alignments can be used for floating
278 point values, similarly to integers.
282 Enumerations are a mapping between an integer type and a table of strings. The
283 numerical representation of the enumeration follows the integer type specified
284 by the meta-data. The enumeration mapping table is detailed in the enumeration
285 description within the meta-data. The mapping table maps inclusive value
286 ranges (or single values) to strings. Instead of being limited to simple
287 "value -> string" mappings, these enumerations map
288 "[ start_value ... end_value ] -> string", which map inclusive ranges of
289 values to strings. An enumeration from the C language can be represented in
290 this format by having the same start_value and end_value for each element, which
291 is in fact a range of size 1. This single-value range is supported without
292 repeating the start and end values with the value = string declaration.
294 enum name : integer_type {
295 somestring = start_value1 ... end_value1,
296 "other string" = start_value2 ... end_value2,
297 yet_another_string, /* will be assigned to end_value2 + 1 */
298 "some other string" = value,
302 If the values are omitted, the enumeration starts at 0 and increment of 1 for
305 enum name : unsigned int {
313 Overlapping ranges within a single enumeration are implementation defined.
315 A nameless enumeration can be declared as a field type or as part of a typedef:
317 enum : integer_type {
321 Enumerations omitting the container type ": integer_type" use the "int"
322 type (for compatibility with C99). The "int" type must be previously
325 typealias integer { size = 32; align = 32; signed = true } := int;
334 Compound are aggregation of type declarations. Compound types include
335 structures, variant, arrays, sequences, and strings.
339 Structures are aligned on the largest alignment required by basic types
340 contained within the structure. (This follows the ISO/C standard for structures)
342 TSDL meta-data representation of a named structure:
345 field_type field_name;
346 field_type field_name;
353 integer { /* Nameless type */
358 uint64_t second_field_name; /* Named type declared in the meta-data */
361 The fields are placed in a sequence next to each other. They each possess a
362 field name, which is a unique identifier within the structure.
364 A nameless structure can be declared as a field type or as part of a typedef:
370 Alignment for a structure compound type can be forced to a minimum value
371 by adding an "align" specifier after the declaration of a structure
372 body. This attribute is read as: align(value). The value is specified in
373 bits. The structure will be aligned on the maximum value between this
374 attribute and the alignment required by the basic types contained within
381 4.2.2 Variants (Discriminated/Tagged Unions)
383 A CTF variant is a selection between different types. A CTF variant must
384 always be defined within the scope of a structure or within fields
385 contained within a structure (defined recursively). A "tag" enumeration
386 field must appear in either the same lexical scope, prior to the variant
387 field (in field declaration order), in an upper lexical scope (see
388 Section 7.3.1), or in an upper dynamic scope (see Section 7.3.2). The
389 type selection is indicated by the mapping from the enumeration value to
390 the string used as variant type selector. The field to use as tag is
391 specified by the "tag_field", specified between "< >" after the
392 "variant" keyword for unnamed variants, and after "variant name" for
395 The alignment of the variant is the alignment of the type as selected by the tag
396 value for the specific instance of the variant. The alignment of the type
397 containing the variant is independent of the variant alignment. The size of the
398 variant is the size as selected by the tag value for the specific instance of
401 A named variant declaration followed by its definition within a structure
412 enum : integer_type { sel1, sel2, sel3, ... } tag_field;
414 variant name <tag_field> v;
417 An unnamed variant definition within a structure is expressed by the following
421 enum : integer_type { sel1, sel2, sel3, ... } tag_field;
423 variant <tag_field> {
431 Example of a named variant within a sequence that refers to a single tag field:
440 enum : uint2_t { a, b, c } choice;
442 variant example <choice> v[seqlen];
445 Example of an unnamed variant:
448 enum : uint2_t { a, b, c, d } choice;
449 /* Unrelated fields can be added between the variant and its tag */
462 Example of an unnamed variant within an array:
465 enum : uint2_t { a, b, c } choice;
473 Example of a variant type definition within a structure, where the defined type
474 is then declared within an array of structures. This variant refers to a tag
475 located in an upper lexical scope. This example clearly shows that a variant
476 type definition referring to the tag "x" uses the closest preceding field from
477 the lexical scope of the type definition.
480 enum : uint2_t { a, b, c, d } x;
482 typedef variant <x> { /*
483 * "x" refers to the preceding "x" enumeration in the
484 * lexical scope of the type definition.
492 enum : int { x, y, z } x; /* This enumeration is not used by "v". */
493 example_variant v; /*
494 * "v" uses the "enum : uint2_t { a, b, c, d }"
502 Arrays are fixed-length. Their length is declared in the type
503 declaration within the meta-data. They contain an array of "inner type"
504 elements, which can refer to any type not containing the type of the
505 array being declared (no circular dependency). The length is the number
506 of elements in an array.
508 TSDL meta-data representation of a named array:
510 typedef elem_type name[length];
512 A nameless array can be declared as a field type within a structure, e.g.:
514 uint8_t field_name[10];
516 Arrays are always aligned on their element alignment requirement.
520 Sequences are dynamically-sized arrays. They refer to a a "length"
521 unsigned integer field, which must appear in either the same lexical scope,
522 prior to the sequence field (in field declaration order), in an upper
523 lexical scope (see Section 7.3.1), or in an upper dynamic scope (see
524 Section 7.3.2). This length field represents the number of elements in
525 the sequence. The sequence per se is an array of "inner type" elements.
527 TSDL meta-data representation for a sequence type definition:
530 unsigned int length_field;
531 typedef elem_type typename[length_field];
532 typename seq_field_name;
535 A sequence can also be declared as a field type, e.g.:
538 unsigned int length_field;
539 long seq_field_name[length_field];
542 Multiple sequences can refer to the same length field, and these length
543 fields can be in a different upper dynamic scope:
545 e.g., assuming the stream.event.header defines:
550 event.header := struct {
559 long seq_a[stream.event.header.seq_len];
560 char seq_b[stream.event.header.seq_len];
564 The sequence elements follow the "array" specifications.
568 Strings are an array of bytes of variable size and are terminated by a '\0'
569 "NULL" character. Their encoding is described in the TSDL meta-data. In
570 absence of encoding attribute information, the default encoding is
573 TSDL meta-data representation of a named string type:
576 encoding = UTF8 OR ASCII;
579 A nameless string type can be declared as a field type:
581 string field_name; /* Use default UTF8 encoding */
583 Strings are always aligned on byte size.
585 5. Event Packet Header
587 The event packet header consists of two parts: the "event packet header"
588 is the same for all streams of a trace. The second part, the "event
589 packet context", is described on a per-stream basis. Both are described
590 in the TSDL meta-data. The packets are aligned on architecture-page-sized
593 Event packet header (all fields are optional, specified by TSDL meta-data):
595 - Magic number (CTF magic number: 0xC1FC1FC1) specifies that this is a
596 CTF packet. This magic number is optional, but when present, it should
597 come at the very beginning of the packet.
598 - Trace UUID, used to ensure the event packet match the meta-data used.
599 (note: we cannot use a meta-data checksum in every cases instead of a
600 UUID because meta-data can be appended to while tracing is active)
601 This field is optional.
602 - Stream ID, used as reference to stream description in meta-data.
603 This field is optional if there is only one stream description in the
604 meta-data, but becomes required if there are more than one stream in
605 the TSDL meta-data description.
607 Event packet context (all fields are optional, specified by TSDL meta-data):
609 - Event packet content size (in bytes).
610 - Event packet size (in bytes, includes padding).
611 - Event packet content checksum (optional). Checksum excludes the event packet
613 - Per-stream event packet sequence count (to deal with UDP packet loss). The
614 number of significant sequence counter bits should also be present, so
615 wrap-arounds are dealt with correctly.
616 - Time-stamp at the beginning and time-stamp at the end of the event packet.
617 Both timestamps are written in the packet header, but sampled respectively
618 while (or before) writing the first event and while (or after) writing the
619 last event in the packet. The inclusive range between these timestamps should
620 include all event timestamps assigned to events contained within the packet.
621 - Events discarded count
622 - Snapshot of a per-stream free-running counter, counting the number of
623 events discarded that were supposed to be written in the stream prior to
624 the first event in the event packet.
625 * Note: producer-consumer buffer full condition should fill the current
626 event packet with padding so we know exactly where events have been
628 - Lossless compression scheme used for the event packet content. Applied
629 directly to raw data. New types of compression can be added in following
630 versions of the format.
631 0: no compression scheme
635 - Cypher used for the event packet content. Applied after compression.
638 - Checksum scheme used for the event packet content. Applied after encryption.
644 5.1 Event Packet Header Description
646 The event packet header layout is indicated by the trace packet.header
647 field. Here is a recommended structure type for the packet header with
648 the fields typically expected (although these fields are each optional):
650 struct event_packet_header {
652 uint8_t trace_uuid[16];
658 packet.header := struct event_packet_header;
661 If the magic number is not present, tools such as "file" will have no
662 mean to discover the file type.
664 If the trace_uuid is not present, no validation that the meta-data
665 actually corresponds to the stream is performed.
667 If the stream_id packet header field is missing, the trace can only
668 contain a single stream. Its "id" field can be left out, and its events
669 don't need to declare a "stream_id" field.
672 5.2 Event Packet Context Description
674 Event packet context example. These are declared within the stream declaration
675 in the meta-data. All these fields are optional. If the packet size field is
676 missing, the whole stream only contains a single packet. If the content
677 size field is missing, the packet is filled (no padding). The content
678 and packet sizes include all headers.
680 An example event packet context type:
682 struct event_packet_context {
683 uint64_t timestamp_begin;
684 uint64_t timestamp_end;
686 uint32_t stream_packet_count;
687 uint32_t events_discarded;
689 uint32_t/uint16_t content_size;
690 uint32_t/uint16_t packet_size;
691 uint8_t stream_packet_count_bits; /* Significant counter bits */
692 uint8_t compression_scheme;
693 uint8_t encryption_scheme;
694 uint8_t checksum_scheme;
700 The overall structure of an event is:
702 1 - Stream Packet Context (as specified by the stream meta-data)
703 2 - Event Header (as specified by the stream meta-data)
704 3 - Stream Event Context (as specified by the stream meta-data)
705 4 - Event Context (as specified by the event meta-data)
706 5 - Event Payload (as specified by the event meta-data)
708 This structure defines an implicit dynamic scoping, where variants
709 located in inner structures (those with a higher number in the listing
710 above) can refer to the fields of outer structures (with lower number in
711 the listing above). See Section 7.3 TSDL Scopes for more detail.
715 Event headers can be described within the meta-data. We hereby propose, as an
716 example, two types of events headers. Type 1 accommodates streams with less than
717 31 event IDs. Type 2 accommodates streams with 31 or more event IDs.
719 One major factor can vary between streams: the number of event IDs assigned to
720 a stream. Luckily, this information tends to stay relatively constant (modulo
721 event registration while trace is being recorded), so we can specify different
722 representations for streams containing few event IDs and streams containing
723 many event IDs, so we end up representing the event ID and time-stamp as
724 densely as possible in each case.
726 The header is extended in the rare occasions where the information cannot be
727 represented in the ranges available in the standard event header. They are also
728 used in the rare occasions where the data required for a field could not be
729 collected: the flag corresponding to the missing field within the missing_fields
730 array is then set to 1.
732 Types uintX_t represent an X-bit unsigned integer, as declared with
735 typealias integer { size = X; align = X; signed = false } := uintX_t;
739 typealias integer { size = X; align = 1; signed = false } := uintX_t;
741 6.1.1 Type 1 - Few event IDs
743 - Aligned on 32-bit (or 8-bit if byte-packed, depending on the architecture
745 - Native architecture byte ordering.
746 - For "compact" selection
747 - Fixed size: 32 bits.
748 - For "extended" selection
749 - Size depends on the architecture and variant alignment.
751 struct event_header_1 {
754 * id 31 is reserved to indicate an extended header.
756 enum : uint5_t { compact = 0 ... 30, extended = 31 } id;
762 uint32_t id; /* 32-bit event IDs */
763 uint64_t timestamp; /* 64-bit timestamps */
769 6.1.2 Type 2 - Many event IDs
771 - Aligned on 16-bit (or 8-bit if byte-packed, depending on the architecture
773 - Native architecture byte ordering.
774 - For "compact" selection
775 - Size depends on the architecture and variant alignment.
776 - For "extended" selection
777 - Size depends on the architecture and variant alignment.
779 struct event_header_2 {
781 * id: range: 0 - 65534.
782 * id 65535 is reserved to indicate an extended header.
784 enum : uint16_t { compact = 0 ... 65534, extended = 65535 } id;
790 uint32_t id; /* 32-bit event IDs */
791 uint64_t timestamp; /* 64-bit timestamps */
799 The event context contains information relative to the current event.
800 The choice and meaning of this information is specified by the TSDL
801 stream and event meta-data descriptions. The stream context is applied
802 to all events within the stream. The stream context structure follows
803 the event header. The event context is applied to specific events. Its
804 structure follows the stream context structure.
806 An example of stream-level event context is to save the event payload size with
807 each event, or to save the current PID with each event. These are declared
808 within the stream declaration within the meta-data:
812 event.context := struct {
814 uint16_t payload_size;
818 An example of event-specific event context is to declare a bitmap of missing
819 fields, only appended after the stream event context if the extended event
820 header is selected. NR_FIELDS is the number of fields within the event (a
828 uint1_t missing_fields[NR_FIELDS]; /* missing event fields bitmap */
837 An event payload contains fields specific to a given event type. The fields
838 belonging to an event type are described in the event-specific meta-data
839 within a structure type.
843 No padding at the end of the event payload. This differs from the ISO/C standard
844 for structures, but follows the CTF standard for structures. In a trace, even
845 though it makes sense to align the beginning of a structure, it really makes no
846 sense to add padding at the end of the structure, because structures are usually
847 not followed by a structure of the same type.
849 This trick can be done by adding a zero-length "end" field at the end of the C
850 structures, and by using the offset of this field rather than using sizeof()
851 when calculating the size of a structure (see Appendix "A. Helper macros").
855 The event payload is aligned on the largest alignment required by types
856 contained within the payload. (This follows the ISO/C standard for structures)
859 7. Trace Stream Description Language (TSDL)
861 The Trace Stream Description Language (TSDL) allows expression of the
862 binary trace streams layout in a C99-like Domain Specific Language
868 The trace stream layout description is located in the trace meta-data.
869 The meta-data is itself located in a stream identified by its name:
872 The meta-data description can be expressed in two different formats:
873 text-only and packet-based. The text-only description facilitates
874 generation of meta-data and provides a convenient way to enter the
875 meta-data information by hand. The packet-based meta-data provides the
876 CTF stream packet facilities (checksumming, compression, encryption,
877 network-readiness) for meta-data stream generated and transported by a
880 The text-only meta-data file is a plain text TSDL description.
882 The packet-based meta-data is made of "meta-data packets", which each
883 start with a meta-data packet header. The packet-based meta-data
884 description is detected by reading the magic number "0x75D11D57" at the
885 beginning of the file. This magic number is also used to detect the
886 endianness of the architecture by trying to read the CTF magic number
887 and its counterpart in reversed endianness. The events within the
888 meta-data stream have no event header nor event context. Each event only
889 contains a "string" payload. Each meta-data packet start with a special
890 packet header, specific to the meta-data stream, which contains,
893 struct metadata_packet_header {
894 uint32_t magic; /* 0x75D11D57 */
895 uint8_t trace_uuid[16]; /* Unique Universal Identifier */
896 uint32_t checksum; /* 0 if unused */
897 uint32_t content_size; /* in bits */
898 uint32_t packet_size; /* in bits */
899 uint8_t compression_scheme; /* 0 if unused */
900 uint8_t encryption_scheme; /* 0 if unused */
901 uint8_t checksum_scheme; /* 0 if unused */
904 The packet-based meta-data can be converted to a text-only meta-data by
905 concatenating all the strings in contains.
907 In the textual representation of the meta-data, the text contained
908 within "/*" and "*/", as well as within "//" and end of line, are
909 treated as comments. Boolean values can be represented as true, TRUE,
910 or 1 for true, and false, FALSE, or 0 for false. Within the string-based
911 meta-data description, the trace UUID is represented as a string of
912 hexadecimal digits and dashes "-". In the event packet header, the trace
913 UUID is represented as an array of bytes.
916 7.2 Declaration vs Definition
918 A declaration associates a layout to a type, without specifying where
919 this type is located in the event structure hierarchy (see Section 6).
920 This therefore includes typedef, typealias, as well as all type
921 specifiers. In certain circumstances (typedef, structure field and
922 variant field), a declaration is followed by a declarator, which specify
923 the newly defined type name (for typedef), or the field name (for
924 declarations located within structure and variants). Array and sequence,
925 declared with square brackets ("[" "]"), are part of the declarator,
926 similarly to C99. The enumeration base type is specified by
927 ": enum_base", which is part of the type specifier. The variant tag
928 name, specified between "<" ">", is also part of the type specifier.
930 A definition associates a type to a location in the event structure
931 hierarchy (see Section 6). This association is denoted by ":=", as shown
937 TSDL uses two different types of scoping: a lexical scope is used for
938 declarations and type definitions, and a dynamic scope is used for
939 variants references to tag fields and for sequence references to length
944 Each of "trace", "stream", "event", "struct" and "variant" have their own
945 nestable declaration scope, within which types can be declared using "typedef"
946 and "typealias". A root declaration scope also contains all declarations
947 located outside of any of the aforementioned declarations. An inner
948 declaration scope can refer to type declared within its container
949 lexical scope prior to the inner declaration scope. Redefinition of a
950 typedef or typealias is not valid, although hiding an upper scope
951 typedef or typealias is allowed within a sub-scope.
955 A dynamic scope consists in the lexical scope augmented with the
956 implicit event structure definition hierarchy presented at Section 6.
957 The dynamic scope is used for variant tag and sequence length
958 definitions. It is used at definition time to look up the location of
959 the tag field associated with a variant, and to lookup up the location
960 of the length field associated with a sequence.
962 Therefore, variants (or sequences) in lower levels in the dynamic scope
963 (e.g. event context) can refer to a tag (or length) field located in
964 upper levels (e.g. in the event header) by specifying, in this case, the
965 associated tag with <header.field_name>. This allows, for instance, the
966 event context to define a variant referring to the "id" field of the
967 event header as selector.
969 The target dynamic scope must be specified explicitly when referring to
970 a field outside of the local static scope. The dynamic scope prefixes
973 - Trace Packet Header: <trace.packet.header. >,
974 - Stream Packet Context: <stream.packet.context. >,
975 - Event Header: <stream.event.header. >,
976 - Stream Event Context: <stream.event.context. >,
977 - Event Context: <event.context. >,
978 - Event Payload: <event.fields. >.
980 Multiple declarations of the same field name within a single scope is
981 not valid. It is however valid to re-use the same field name in
982 different scopes. There is no possible conflict, because the dynamic
983 scope must be specified when a variant refers to a tag field located in
984 a different dynamic scope.
986 The information available in the dynamic scopes can be thought of as the
987 current tracing context. At trace production, information about the
988 current context is saved into the specified scope field levels. At trace
989 consumption, for each event, the current trace context is therefore
990 readable by accessing the upper dynamic scopes.
995 The grammar representing the TSDL meta-data is presented in Appendix C.
996 TSDL Grammar. This section presents a rather lighter reading that
997 consists in examples of TSDL meta-data, with template values.
999 The stream "id" can be left out if there is only one stream in the
1000 trace. The event "id" field can be left out if there is only one event
1004 major = value; /* Trace format version */
1006 uuid = "aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa"; /* Trace UUID */
1007 byte_order = be OR le; /* Endianness (required) */
1008 packet.header := struct {
1010 uint8_t trace_uuid[16];
1017 /* Type 1 - Few event IDs; Type 2 - Many event IDs. See section 6.1. */
1018 event.header := event_header_1 OR event_header_2;
1019 event.context := struct {
1022 packet.context := struct {
1029 id = value; /* Numeric identifier within the stream */
1030 stream_id = stream_id;
1039 /* More detail on types in section 4. Types */
1044 * Type declarations behave similarly to the C standard.
1047 typedef aliased_type_specifiers new_type_declarators;
1049 /* e.g.: typedef struct example new_type_name[10]; */
1054 * The "typealias" declaration can be used to give a name (including
1055 * pointer declarator specifier) to a type. It should also be used to
1056 * map basic C types (float, int, unsigned long, ...) to a CTF type.
1057 * Typealias is a superset of "typedef": it also allows assignment of a
1058 * simple variable identifier to a type.
1061 typealias type_class {
1063 } := type_specifiers type_declarator;
1067 * typealias integer {
1071 * } := struct page *;
1073 * typealias integer {
1088 enum name : integer_type {
1094 * Unnamed types, contained within compound type fields, typedef or typealias.
1109 enum : integer_type {
1113 typedef type new_type[length];
1116 type field_name[length];
1119 typedef type new_type[length_type];
1122 type field_name[length_type];
1134 integer_type field_name:size; /* GNU/C bitfield */
1144 The two following macros keep track of the size of a GNU/C structure without
1145 padding at the end by placing HEADER_END as the last field. A one byte end field
1146 is used for C90 compatibility (C99 flexible arrays could be used here). Note
1147 that this does not affect the effective structure size, which should always be
1148 calculated with the header_sizeof() helper.
1150 #define HEADER_END char end_field
1151 #define header_sizeof(type) offsetof(typeof(type), end_field)
1154 B. Stream Header Rationale
1156 An event stream is divided in contiguous event packets of variable size. These
1157 subdivisions allow the trace analyzer to perform a fast binary search by time
1158 within the stream (typically requiring to index only the event packet headers)
1159 without reading the whole stream. These subdivisions have a variable size to
1160 eliminate the need to transfer the event packet padding when partially filled
1161 event packets must be sent when streaming a trace for live viewing/analysis.
1162 An event packet can contain a certain amount of padding at the end. Dividing
1163 streams into event packets is also useful for network streaming over UDP and
1164 flight recorder mode tracing (a whole event packet can be swapped out of the
1165 buffer atomically for reading).
1167 The stream header is repeated at the beginning of each event packet to allow
1168 flexibility in terms of:
1170 - streaming support,
1171 - allowing arbitrary buffers to be discarded without making the trace
1173 - allow UDP packet loss handling by either dealing with missing event packet
1174 or asking for re-transmission.
1175 - transparently support flight recorder mode,
1176 - transparently support crash dump.
1182 * Common Trace Format (CTF) Trace Stream Description Language (TSDL) Grammar.
1184 * Inspired from the C99 grammar:
1185 * http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1124.pdf (Annex A)
1186 * and c++1x grammar (draft)
1187 * http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3291.pdf (Annex A)
1189 * Specialized for CTF needs by including only constant and declarations from
1190 * C99 (excluding function declarations), and by adding support for variants,
1191 * sequences and CTF-specific specifiers. Enumeration container types
1192 * semantic is inspired from c++1x enum-base.
1197 1.1) Lexical elements
1241 identifier identifier-nondigit
1244 identifier-nondigit:
1246 universal-character-name
1247 any other implementation-defined characters
1251 [a-zA-Z] /* regular expression */
1254 [0-9] /* regular expression */
1256 1.4) Universal character names
1258 universal-character-name:
1260 \U hex-quad hex-quad
1263 hexadecimal-digit hexadecimal-digit hexadecimal-digit hexadecimal-digit
1269 enumeration-constant
1273 decimal-constant integer-suffix-opt
1274 octal-constant integer-suffix-opt
1275 hexadecimal-constant integer-suffix-opt
1279 decimal-constant digit
1283 octal-constant octal-digit
1285 hexadecimal-constant:
1286 hexadecimal-prefix hexadecimal-digit
1287 hexadecimal-constant hexadecimal-digit
1297 unsigned-suffix long-suffix-opt
1298 unsigned-suffix long-long-suffix
1299 long-suffix unsigned-suffix-opt
1300 long-long-suffix unsigned-suffix-opt
1314 enumeration-constant:
1320 L' c-char-sequence '
1324 c-char-sequence c-char
1327 any member of source charset except single-quote ('), backslash
1328 (\), or new-line character.
1332 simple-escape-sequence
1333 octal-escape-sequence
1334 hexadecimal-escape-sequence
1335 universal-character-name
1337 simple-escape-sequence: one of
1338 \' \" \? \\ \a \b \f \n \r \t \v
1340 octal-escape-sequence:
1342 \ octal-digit octal-digit
1343 \ octal-digit octal-digit octal-digit
1345 hexadecimal-escape-sequence:
1346 \x hexadecimal-digit
1347 hexadecimal-escape-sequence hexadecimal-digit
1349 1.6) String literals
1352 " s-char-sequence-opt "
1353 L" s-char-sequence-opt "
1357 s-char-sequence s-char
1360 any member of source charset except double-quote ("), backslash
1361 (\), or new-line character.
1367 [ ] ( ) { } . -> * + - < > : ; ... = ,
1370 2) Phrase structure grammar
1376 ( unary-expression )
1380 postfix-expression [ unary-expression ]
1381 postfix-expression . identifier
1382 postfix-expressoin -> identifier
1386 unary-operator postfix-expression
1388 unary-operator: one of
1391 assignment-operator:
1394 type-assignment-operator:
1397 constant-expression-range:
1398 unary-expression ... unary-expression
1403 declaration-specifiers declarator-list-opt ;
1406 declaration-specifiers:
1407 storage-class-specifier declaration-specifiers-opt
1408 type-specifier declaration-specifiers-opt
1409 type-qualifier declaration-specifiers-opt
1413 declarator-list , declarator
1415 abstract-declarator-list:
1417 abstract-declarator-list , abstract-declarator
1419 storage-class-specifier:
1442 align ( unary-expression )
1445 struct identifier-opt { struct-or-variant-declaration-list-opt } align-attribute-opt
1446 struct identifier align-attribute-opt
1448 struct-or-variant-declaration-list:
1449 struct-or-variant-declaration
1450 struct-or-variant-declaration-list struct-or-variant-declaration
1452 struct-or-variant-declaration:
1453 specifier-qualifier-list struct-or-variant-declarator-list ;
1454 declaration-specifiers storage-class-specifier declaration-specifiers declarator-list ;
1455 typealias declaration-specifiers abstract-declarator-list := declaration-specifiers abstract-declarator-list ;
1456 typealias declaration-specifiers abstract-declarator-list := declarator-list ;
1458 specifier-qualifier-list:
1459 type-specifier specifier-qualifier-list-opt
1460 type-qualifier specifier-qualifier-list-opt
1462 struct-or-variant-declarator-list:
1463 struct-or-variant-declarator
1464 struct-or-variant-declarator-list , struct-or-variant-declarator
1466 struct-or-variant-declarator:
1468 declarator-opt : unary-expression
1471 variant identifier-opt variant-tag-opt { struct-or-variant-declaration-list }
1472 variant identifier variant-tag
1478 enum identifier-opt { enumerator-list }
1479 enum identifier-opt { enumerator-list , }
1481 enum identifier-opt : declaration-specifiers { enumerator-list }
1482 enum identifier-opt : declaration-specifiers { enumerator-list , }
1486 enumerator-list , enumerator
1489 enumeration-constant
1490 enumeration-constant = unary-expression
1491 enumeration-constant = constant-expression-range
1497 pointer-opt direct-declarator
1502 direct-declarator [ unary-expression ]
1504 abstract-declarator:
1505 pointer-opt direct-abstract-declarator
1507 direct-abstract-declarator:
1509 ( abstract-declarator )
1510 direct-abstract-declarator [ unary-expression ]
1511 direct-abstract-declarator [ ]
1514 * type-qualifier-list-opt
1515 * type-qualifier-list-opt pointer
1517 type-qualifier-list:
1519 type-qualifier-list type-qualifier
1524 2.3) CTF-specific declarations
1527 event { ctf-assignment-expression-list-opt }
1528 stream { ctf-assignment-expression-list-opt }
1529 trace { ctf-assignment-expression-list-opt }
1530 typealias declaration-specifiers abstract-declarator-list := declaration-specifiers abstract-declarator-list ;
1531 typealias declaration-specifiers abstract-declarator-list := declarator-list ;
1534 floating_point { ctf-assignment-expression-list-opt }
1535 integer { ctf-assignment-expression-list-opt }
1536 string { ctf-assignment-expression-list-opt }
1538 ctf-assignment-expression-list:
1539 ctf-assignment-expression
1540 ctf-assignment-expression-list ; ctf-assignment-expression
1542 ctf-assignment-expression:
1543 unary-expression assignment-operator unary-expression
1544 unary-expression type-assignment-operator type-specifier
1545 declaration-specifiers storage-class-specifier declaration-specifiers declarator-list
1546 typealias declaration-specifiers abstract-declarator-list := declaration-specifiers abstract-declarator-list
1547 typealias declaration-specifiers abstract-declarator-list := declarator-list