1 Common Trace Format (CTF) Specification (v1.8.2)
3 Mathieu Desnoyers, EfficiOS Inc.
5 The goal of the present document is to specify a trace format that suits the
6 needs of the embedded, telecom, high-performance and kernel communities. It is
7 based on the Common Trace Format Requirements (v1.4) document. It is designed to
8 allow traces to be natively generated by the Linux kernel, Linux user-space
9 applications written in C/C++, and hardware components. One major element of
10 CTF is the Trace Stream Description Language (TSDL) which flexibility
11 enables description of various binary trace stream layouts.
13 The latest version of this document can be found at:
15 git tree: git://git.efficios.com/ctf.git
16 gitweb: http://git.efficios.com/?p=ctf.git
18 A reference implementation of a library to read and write this trace format is
19 being implemented within the BabelTrace project, a converter between trace
20 formats. The development tree is available at:
22 git tree: git://git.efficios.com/babeltrace.git
23 gitweb: http://git.efficios.com/?p=babeltrace.git
25 The CE Workgroup of the Linux Foundation, Ericsson, and EfficiOS have
31 1. Preliminary definitions
32 2. High-level representation of a trace
36 4.1.1 Type inheritance
46 4.2.2 Variants (Discriminated/Tagged Unions)
50 5. Event Packet Header
51 5.1 Event Packet Header Description
52 5.2 Event Packet Context Description
55 6.1.1 Type 1 - Few event IDs
56 6.1.2 Type 2 - Many event IDs
61 7. Trace Stream Description Language (TSDL)
63 7.2 Declaration vs Definition
66 7.3.2 Static and Dynamic Scopes
71 1. Preliminary definitions
73 - Event Trace: An ordered sequence of events.
74 - Event Stream: An ordered sequence of events, containing a subset of the
76 - Event Packet: A sequence of physically contiguous events within an event
78 - Event: This is the basic entry in a trace. (aka: a trace record).
79 - An event identifier (ID) relates to the class (a type) of event within
81 e.g. event: irq_entry.
82 - An event (or event record) relates to a specific instance of an event
84 e.g. event: irq_entry, at time X, on CPU Y
85 - Source Architecture: Architecture writing the trace.
86 - Reader Architecture: Architecture reading the trace.
89 2. High-level representation of a trace
91 A trace is divided into multiple event streams. Each event stream contains a
92 subset of the trace event types.
94 The final output of the trace, after its generation and optional transport over
95 the network, is expected to be either on permanent or temporary storage in a
96 virtual file system. Because each event stream is appended to while a trace is
97 being recorded, each is associated with a distinct set of files for
98 output. Therefore, a stored trace can be represented as a directory
99 containing zero, one or more files per stream.
101 Meta-data description associated with the trace contains information on
102 trace event types expressed in the Trace Stream Description Language
103 (TSDL). This language describes:
107 - Per-trace event header description.
108 - Per-stream event header description.
109 - Per-stream event context description.
111 - Event type to stream mapping.
112 - Event type to name mapping.
113 - Event type to ID mapping.
114 - Event context description.
115 - Event fields description.
120 An event stream can be divided into contiguous event packets of variable
121 size. An event packet can contain a certain amount of padding at the
122 end. The stream header is repeated at the beginning of each event
123 packet. The rationale for the event stream design choices is explained
124 in Appendix B. Stream Header Rationale.
126 The event stream header will therefore be referred to as the "event packet
127 header" throughout the rest of this document.
132 Types are organized as type classes. Each type class belong to either of two
133 kind of types: basic types or compound types.
137 A basic type is a scalar type, as described in this section. It includes
138 integers, GNU/C bitfields, enumerations, and floating point values.
140 4.1.1 Type inheritance
142 Type specifications can be inherited to allow deriving types from a
143 type class. For example, see the uint32_t named type derived from the "integer"
144 type class below ("Integers" section). Types have a precise binary
145 representation in the trace. A type class has methods to read and write these
146 types, but must be derived into a type to be usable in an event field.
150 We define "byte-packed" types as aligned on the byte size, namely 8-bit.
151 We define "bit-packed" types as following on the next bit, as defined by the
154 Each basic type must specify its alignment, in bits. Examples of
155 possible alignments are: bit-packed (align = 1), byte-packed (align =
156 8), or word-aligned (e.g. align = 32 or align = 64). The choice depends
157 on the architecture preference and compactness vs performance trade-offs
158 of the implementation. Architectures providing fast unaligned write
159 byte-packed basic types to save space, aligning each type on byte
160 boundaries (8-bit). Architectures with slow unaligned writes align types
161 on specific alignment values. If no specific alignment is declared for a
162 type, it is assumed to be bit-packed for integers with size not multiple
163 of 8 bits and for gcc bitfields. All other basic types are byte-packed
164 by default. It is however recommended to always specify the alignment
165 explicitly. Alignment values must be power of two. Compound types are
166 aligned as specified in their individual specification.
168 TSDL meta-data attribute representation of a specific alignment:
170 align = value; /* value in bits */
174 By default, the native endianness of the source architecture is used.
175 Byte order can be overridden for a basic type by specifying a "byte_order"
176 attribute. Typical use-case is to specify the network byte order (big endian:
177 "be") to save data captured from the network into the trace without conversion.
178 If not specified, the byte order is native.
180 TSDL meta-data representation:
182 byte_order = native OR network OR be OR le; /* network and be are aliases */
184 Even though the trace description section is not per se a type, for sake
185 of clarity, it should be noted that native and network byte orders are
186 only allowed within type declaration. The byte_order specified in the
187 trace description section only accepts be OR le values.
191 Type size, in bits, for integers and floats is that returned by "sizeof()" in C
192 multiplied by CHAR_BIT.
193 We require the size of "char" and "unsigned char" types (CHAR_BIT) to be fixed
194 to 8 bits for cross-endianness compatibility.
196 TSDL meta-data representation:
198 size = value; (value is in bits)
202 Signed integers are represented in two-complement. Integer alignment,
203 size, signedness and byte ordering are defined in the TSDL meta-data.
204 Integers aligned on byte size (8-bit) and with length multiple of byte
205 size (8-bit) correspond to the C99 standard integers. In addition,
206 integers with alignment and/or size that are _not_ a multiple of the
207 byte size are permitted; these correspond to the C99 standard bitfields,
208 with the added specification that the CTF integer bitfields have a fixed
209 binary representation. A MIT-licensed reference implementation of the
210 CTF portable bitfields is available at:
212 http://git.efficios.com/?p=babeltrace.git;a=blob;f=include/babeltrace/bitfield.h
214 Binary representation of integers:
216 - On little and big endian:
217 - Within a byte, high bits correspond to an integer high bits, and low bits
218 correspond to low bits.
220 - Integer across multiple bytes are placed from the less significant to the
222 - Consecutive integers are placed from lower bits to higher bits (even within
225 - Integer across multiple bytes are placed from the most significant to the
227 - Consecutive integers are placed from higher bits to lower bits (even within
230 This binary representation is derived from the bitfield implementation in GCC
231 for little and big endian. However, contrary to what GCC does, integers can
232 cross units boundaries (no padding is required). Padding can be explicitly
233 added (see 4.1.6 GNU/C bitfields) to follow the GCC layout if needed.
235 TSDL meta-data representation:
238 signed = true OR false; /* default false */
239 byte_order = native OR network OR be OR le; /* default native */
240 size = value; /* value in bits, no default */
241 align = value; /* value in bits */
242 /* based used for pretty-printing output, default: decimal. */
243 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
244 OR octal OR oct OR o OR 8 OR binary OR b OR 2;
245 /* character encoding, default: none */
246 encoding = none or UTF8 or ASCII;
249 Example of type inheritance (creation of a uint32_t named type):
257 Definition of a named 5-bit signed bitfield:
265 The character encoding field can be used to specify that the integer
266 must be printed as a text character when read. e.g.:
276 4.1.6 GNU/C bitfields
278 The GNU/C bitfields follow closely the integer representation, with a
279 particularity on alignment: if a bitfield cannot fit in the current unit, the
280 unit is padded and the bitfield starts at the following unit. The unit size is
281 defined by the size of the type "unit_type".
283 TSDL meta-data representation:
287 As an example, the following structure declared in C compiled by GCC:
294 The example structure is aligned on the largest element (short). The second
295 bitfield would be aligned on the next unit boundary, because it would not fit in
300 The floating point values byte ordering is defined in the TSDL meta-data.
302 Floating point values follow the IEEE 754-2008 standard interchange formats.
303 Description of the floating point values include the exponent and mantissa size
304 in bits. Some requirements are imposed on the floating point values:
306 - FLT_RADIX must be 2.
307 - mant_dig is the number of digits represented in the mantissa. It is specified
308 by the ISO C99 standard, section 5.2.4, as FLT_MANT_DIG, DBL_MANT_DIG and
309 LDBL_MANT_DIG as defined by <float.h>.
310 - exp_dig is the number of digits represented in the exponent. Given that
311 mant_dig is one bit more than its actual size in bits (leading 1 is not
312 needed) and also given that the sign bit always takes one bit, exp_dig can be
315 - sizeof(float) * CHAR_BIT - FLT_MANT_DIG
316 - sizeof(double) * CHAR_BIT - DBL_MANT_DIG
317 - sizeof(long double) * CHAR_BIT - LDBL_MANT_DIG
319 TSDL meta-data representation:
324 byte_order = native OR network OR be OR le;
328 Example of type inheritance:
330 typealias floating_point {
331 exp_dig = 8; /* sizeof(float) * CHAR_BIT - FLT_MANT_DIG */
332 mant_dig = 24; /* FLT_MANT_DIG */
337 TODO: define NaN, +inf, -inf behavior.
339 Bit-packed, byte-packed or larger alignments can be used for floating
340 point values, similarly to integers.
344 Enumerations are a mapping between an integer type and a table of strings. The
345 numerical representation of the enumeration follows the integer type specified
346 by the meta-data. The enumeration mapping table is detailed in the enumeration
347 description within the meta-data. The mapping table maps inclusive value
348 ranges (or single values) to strings. Instead of being limited to simple
349 "value -> string" mappings, these enumerations map
350 "[ start_value ... end_value ] -> string", which map inclusive ranges of
351 values to strings. An enumeration from the C language can be represented in
352 this format by having the same start_value and end_value for each element, which
353 is in fact a range of size 1. This single-value range is supported without
354 repeating the start and end values with the value = string declaration.
356 enum name : integer_type {
357 somestring = start_value1 ... end_value1,
358 "other string" = start_value2 ... end_value2,
359 yet_another_string, /* will be assigned to end_value2 + 1 */
360 "some other string" = value,
364 If the values are omitted, the enumeration starts at 0 and increment of 1 for
365 each entry. An entry with omitted value that follows a range entry takes
366 as value the end_value of the previous range + 1:
368 enum name : unsigned int {
376 Overlapping ranges within a single enumeration are implementation defined.
378 A nameless enumeration can be declared as a field type or as part of a typedef:
380 enum : integer_type {
384 Enumerations omitting the container type ": integer_type" use the "int"
385 type (for compatibility with C99). The "int" type must be previously
388 typealias integer { size = 32; align = 32; signed = true } := int;
397 Compound are aggregation of type declarations. Compound types include
398 structures, variant, arrays, sequences, and strings.
402 Structures are aligned on the largest alignment required by basic types
403 contained within the structure. (This follows the ISO/C standard for structures)
405 TSDL meta-data representation of a named structure:
408 field_type field_name;
409 field_type field_name;
416 integer { /* Nameless type */
421 uint64_t second_field_name; /* Named type declared in the meta-data */
424 The fields are placed in a sequence next to each other. They each
425 possess a field name, which is a unique identifier within the structure.
426 The identifier is not allowed to use any reserved keyword
427 (see Section C.1.2). Replacing reserved keywords with
428 underscore-prefixed field names is recommended. Fields starting with an
429 underscore should have their leading underscore removed by the CTF trace
432 A nameless structure can be declared as a field type or as part of a typedef:
438 Alignment for a structure compound type can be forced to a minimum value
439 by adding an "align" specifier after the declaration of a structure
440 body. This attribute is read as: align(value). The value is specified in
441 bits. The structure will be aligned on the maximum value between this
442 attribute and the alignment required by the basic types contained within
449 4.2.2 Variants (Discriminated/Tagged Unions)
451 A CTF variant is a selection between different types. A CTF variant must
452 always be defined within the scope of a structure or within fields
453 contained within a structure (defined recursively). A "tag" enumeration
454 field must appear in either the same static scope, prior to the variant
455 field (in field declaration order), in an upper static scope , or in an
456 upper dynamic scope (see Section 7.3.2). The type selection is indicated
457 by the mapping from the enumeration value to the string used as variant
458 type selector. The field to use as tag is specified by the "tag_field",
459 specified between "< >" after the "variant" keyword for unnamed
460 variants, and after "variant name" for named variants.
462 The alignment of the variant is the alignment of the type as selected by
463 the tag value for the specific instance of the variant. The size of the
464 variant is the size as selected by the tag value for the specific
465 instance of the variant.
467 The alignment of the type containing the variant is independent of the
468 variant alignment. For instance, if a structure contains two fields, a
469 32-bit integer, aligned on 32 bits, and a variant, which contains two
470 choices: either a 32-bit field, aligned on 32 bits, or a 64-bit field,
471 aligned on 64 bits, the alignment of the outmost structure will be
472 32-bit (the alignment of its largest field, disregarding the alignment
473 of the variant). The alignment of the variant will depend on the
474 selector: if the variant's 32-bit field is selected, its alignment will
475 be 32-bit, or 64-bit otherwise. It is important to note that variants
476 are specifically tailored for compactness in a stream. Therefore, the
477 relative offsets of compound type fields can vary depending on
478 the offset at which the compound type starts if it contains a variant
479 that itself contains a type with alignment larger than the largest field
480 contained within the compound type. This is caused by the fact that the
481 compound type may contain the enumeration that select the variant's
482 choice, and therefore the alignment to be applied to the compound type
483 cannot be determined before encountering the enumeration.
485 Each variant type selector possess a field name, which is a unique
486 identifier within the variant. The identifier is not allowed to use any
487 reserved keyword (see Section C.1.2). Replacing reserved keywords with
488 underscore-prefixed field names is recommended. Fields starting with an
489 underscore should have their leading underscore removed by the CTF trace
493 A named variant declaration followed by its definition within a structure
504 enum : integer_type { sel1, sel2, sel3, ... } tag_field;
506 variant name <tag_field> v;
509 An unnamed variant definition within a structure is expressed by the following
513 enum : integer_type { sel1, sel2, sel3, ... } tag_field;
515 variant <tag_field> {
523 Example of a named variant within a sequence that refers to a single tag field:
532 enum : uint2_t { a, b, c } choice;
534 variant example <choice> v[seqlen];
537 Example of an unnamed variant:
540 enum : uint2_t { a, b, c, d } choice;
541 /* Unrelated fields can be added between the variant and its tag */
554 Example of an unnamed variant within an array:
557 enum : uint2_t { a, b, c } choice;
565 Example of a variant type definition within a structure, where the defined type
566 is then declared within an array of structures. This variant refers to a tag
567 located in an upper static scope. This example clearly shows that a variant
568 type definition referring to the tag "x" uses the closest preceding field from
569 the static scope of the type definition.
572 enum : uint2_t { a, b, c, d } x;
574 typedef variant <x> { /*
575 * "x" refers to the preceding "x" enumeration in the
576 * static scope of the type definition.
584 enum : int { x, y, z } x; /* This enumeration is not used by "v". */
585 example_variant v; /*
586 * "v" uses the "enum : uint2_t { a, b, c, d }"
594 Arrays are fixed-length. Their length is declared in the type
595 declaration within the meta-data. They contain an array of "inner type"
596 elements, which can refer to any type not containing the type of the
597 array being declared (no circular dependency). The length is the number
598 of elements in an array.
600 TSDL meta-data representation of a named array:
602 typedef elem_type name[length];
604 A nameless array can be declared as a field type within a structure, e.g.:
606 uint8_t field_name[10];
608 Arrays are always aligned on their element alignment requirement.
612 Sequences are dynamically-sized arrays. They refer to a "length"
613 unsigned integer field, which must appear in either the same static scope,
614 prior to the sequence field (in field declaration order), in an upper
615 static scope, or in an upper dynamic scope (see Section 7.3.2). This
616 length field represents the number of elements in the sequence. The
617 sequence per se is an array of "inner type" elements.
619 TSDL meta-data representation for a sequence type definition:
622 unsigned int length_field;
623 typedef elem_type typename[length_field];
624 typename seq_field_name;
627 A sequence can also be declared as a field type, e.g.:
630 unsigned int length_field;
631 long seq_field_name[length_field];
634 Multiple sequences can refer to the same length field, and these length
635 fields can be in a different upper dynamic scope:
637 e.g., assuming the stream.event.header defines:
642 event.header := struct {
651 long seq_a[stream.event.header.seq_len];
652 char seq_b[stream.event.header.seq_len];
656 The sequence elements follow the "array" specifications.
660 Strings are an array of bytes of variable size and are terminated by a '\0'
661 "NULL" character. Their encoding is described in the TSDL meta-data. In
662 absence of encoding attribute information, the default encoding is
665 TSDL meta-data representation of a named string type:
668 encoding = UTF8 OR ASCII;
671 A nameless string type can be declared as a field type:
673 string field_name; /* Use default UTF8 encoding */
675 Strings are always aligned on byte size.
677 5. Event Packet Header
679 The event packet header consists of two parts: the "event packet header"
680 is the same for all streams of a trace. The second part, the "event
681 packet context", is described on a per-stream basis. Both are described
682 in the TSDL meta-data. The packets are aligned on architecture-page-sized
685 Event packet header (all fields are optional, specified by TSDL meta-data):
687 - Magic number (CTF magic number: 0xC1FC1FC1) specifies that this is a
688 CTF packet. This magic number is optional, but when present, it should
689 come at the very beginning of the packet.
690 - Trace UUID, used to ensure the event packet match the meta-data used.
691 (note: we cannot use a meta-data checksum in every cases instead of a
692 UUID because meta-data can be appended to while tracing is active)
693 This field is optional.
694 - Stream ID, used as reference to stream description in meta-data.
695 This field is optional if there is only one stream description in the
696 meta-data, but becomes required if there are more than one stream in
697 the TSDL meta-data description.
699 Event packet context (all fields are optional, specified by TSDL meta-data):
701 - Event packet content size (in bits).
702 - Event packet size (in bits, includes padding).
703 - Event packet content checksum. Checksum excludes the event packet
705 - Per-stream event packet sequence count (to deal with UDP packet loss). The
706 number of significant sequence counter bits should also be present, so
707 wrap-arounds are dealt with correctly.
708 - Time-stamp at the beginning and time-stamp at the end of the event packet.
709 Both timestamps are written in the packet header, but sampled respectively
710 while (or before) writing the first event and while (or after) writing the
711 last event in the packet. The inclusive range between these timestamps should
712 include all event timestamps assigned to events contained within the packet.
713 - Events discarded count
714 - Snapshot of a per-stream free-running counter, counting the number of
715 events discarded that were supposed to be written in the stream after
716 the last event in the event packet.
717 * Note: producer-consumer buffer full condition can fill the current
718 event packet with padding so we know exactly where events have been
719 discarded. However, if the buffer full condition chooses not
720 to fill the current event packet with padding, all we know
721 about the timestamp range in which the events have been
722 discarded is that it is somewhere between the beginning and
723 the end of the packet.
724 - Lossless compression scheme used for the event packet content. Applied
725 directly to raw data. New types of compression can be added in following
726 versions of the format.
727 0: no compression scheme
731 - Cypher used for the event packet content. Applied after compression.
734 - Checksum scheme used for the event packet content. Applied after encryption.
740 5.1 Event Packet Header Description
742 The event packet header layout is indicated by the trace packet.header
743 field. Here is a recommended structure type for the packet header with
744 the fields typically expected (although these fields are each optional):
746 struct event_packet_header {
754 packet.header := struct event_packet_header;
757 If the magic number is not present, tools such as "file" will have no
758 mean to discover the file type.
760 If the uuid is not present, no validation that the meta-data actually
761 corresponds to the stream is performed.
763 If the stream_id packet header field is missing, the trace can only
764 contain a single stream. Its "id" field can be left out, and its events
765 don't need to declare a "stream_id" field.
768 5.2 Event Packet Context Description
770 Event packet context example. These are declared within the stream declaration
771 in the meta-data. All these fields are optional. If the packet size field is
772 missing, the whole stream only contains a single packet. If the content
773 size field is missing, the packet is filled (no padding). The content
774 and packet sizes include all headers.
776 An example event packet context type:
778 struct event_packet_context {
779 uint64_t timestamp_begin;
780 uint64_t timestamp_end;
782 uint32_t stream_packet_count;
783 uint32_t events_discarded;
785 uint64_t/uint32_t/uint16_t content_size;
786 uint64_t/uint32_t/uint16_t packet_size;
787 uint8_t compression_scheme;
788 uint8_t encryption_scheme;
789 uint8_t checksum_scheme;
795 The overall structure of an event is:
797 1 - Stream Packet Context (as specified by the stream meta-data)
798 2 - Event Header (as specified by the stream meta-data)
799 3 - Stream Event Context (as specified by the stream meta-data)
800 4 - Event Context (as specified by the event meta-data)
801 5 - Event Payload (as specified by the event meta-data)
803 This structure defines an implicit dynamic scoping, where variants
804 located in inner structures (those with a higher number in the listing
805 above) can refer to the fields of outer structures (with lower number in
806 the listing above). See Section 7.3 TSDL Scopes for more detail.
810 Event headers can be described within the meta-data. We hereby propose, as an
811 example, two types of events headers. Type 1 accommodates streams with less than
812 31 event IDs. Type 2 accommodates streams with 31 or more event IDs.
814 One major factor can vary between streams: the number of event IDs assigned to
815 a stream. Luckily, this information tends to stay relatively constant (modulo
816 event registration while trace is being recorded), so we can specify different
817 representations for streams containing few event IDs and streams containing
818 many event IDs, so we end up representing the event ID and time-stamp as
819 densely as possible in each case.
821 The header is extended in the rare occasions where the information cannot be
822 represented in the ranges available in the standard event header. They are also
823 used in the rare occasions where the data required for a field could not be
824 collected: the flag corresponding to the missing field within the missing_fields
825 array is then set to 1.
827 Types uintX_t represent an X-bit unsigned integer, as declared with
830 typealias integer { size = X; align = X; signed = false } := uintX_t;
834 typealias integer { size = X; align = 1; signed = false } := uintX_t;
836 6.1.1 Type 1 - Few event IDs
838 - Aligned on 32-bit (or 8-bit if byte-packed, depending on the architecture
840 - Native architecture byte ordering.
841 - For "compact" selection
842 - Fixed size: 32 bits.
843 - For "extended" selection
844 - Size depends on the architecture and variant alignment.
846 struct event_header_1 {
849 * id 31 is reserved to indicate an extended header.
851 enum : uint5_t { compact = 0 ... 30, extended = 31 } id;
857 uint32_t id; /* 32-bit event IDs */
858 uint64_t timestamp; /* 64-bit timestamps */
861 } align(32); /* or align(8) */
864 6.1.2 Type 2 - Many event IDs
866 - Aligned on 16-bit (or 8-bit if byte-packed, depending on the architecture
868 - Native architecture byte ordering.
869 - For "compact" selection
870 - Size depends on the architecture and variant alignment.
871 - For "extended" selection
872 - Size depends on the architecture and variant alignment.
874 struct event_header_2 {
876 * id: range: 0 - 65534.
877 * id 65535 is reserved to indicate an extended header.
879 enum : uint16_t { compact = 0 ... 65534, extended = 65535 } id;
885 uint32_t id; /* 32-bit event IDs */
886 uint64_t timestamp; /* 64-bit timestamps */
889 } align(16); /* or align(8) */
894 The event context contains information relative to the current event.
895 The choice and meaning of this information is specified by the TSDL
896 stream and event meta-data descriptions. The stream context is applied
897 to all events within the stream. The stream context structure follows
898 the event header. The event context is applied to specific events. Its
899 structure follows the stream context structure.
901 An example of stream-level event context is to save the event payload size with
902 each event, or to save the current PID with each event. These are declared
903 within the stream declaration within the meta-data:
907 event.context := struct {
909 uint16_t payload_size;
913 An example of event-specific event context is to declare a bitmap of missing
914 fields, only appended after the stream event context if the extended event
915 header is selected. NR_FIELDS is the number of fields within the event (a
923 uint1_t missing_fields[NR_FIELDS]; /* missing event fields bitmap */
932 An event payload contains fields specific to a given event type. The fields
933 belonging to an event type are described in the event-specific meta-data
934 within a structure type.
938 No padding at the end of the event payload. This differs from the ISO/C standard
939 for structures, but follows the CTF standard for structures. In a trace, even
940 though it makes sense to align the beginning of a structure, it really makes no
941 sense to add padding at the end of the structure, because structures are usually
942 not followed by a structure of the same type.
944 This trick can be done by adding a zero-length "end" field at the end of the C
945 structures, and by using the offset of this field rather than using sizeof()
946 when calculating the size of a structure (see Appendix "A. Helper macros").
950 The event payload is aligned on the largest alignment required by types
951 contained within the payload. (This follows the ISO/C standard for structures)
954 7. Trace Stream Description Language (TSDL)
956 The Trace Stream Description Language (TSDL) allows expression of the
957 binary trace streams layout in a C99-like Domain Specific Language
963 The trace stream layout description is located in the trace meta-data.
964 The meta-data is itself located in a stream identified by its name:
967 The meta-data description can be expressed in two different formats:
968 text-only and packet-based. The text-only description facilitates
969 generation of meta-data and provides a convenient way to enter the
970 meta-data information by hand. The packet-based meta-data provides the
971 CTF stream packet facilities (checksumming, compression, encryption,
972 network-readiness) for meta-data stream generated and transported by a
975 The text-only meta-data file is a plain-text TSDL description. This file
976 must begin with the following characters to identify the file as a CTF
977 TSDL text-based metadata file (without the double-quotes) :
981 It must be followed by a space, and the version of the specification
982 followed by the CTF trace, e.g.:
986 These characters allow automated discovery of file type and CTF
987 specification version. They are interpreted as a the beginning of a
988 comment by the TSDL metadata parser. The comment can be continued to
989 contain extra commented characters before it is closed.
991 The packet-based meta-data is made of "meta-data packets", which each
992 start with a meta-data packet header. The packet-based meta-data
993 description is detected by reading the magic number "0x75D11D57" at the
994 beginning of the file. This magic number is also used to detect the
995 endianness of the architecture by trying to read the CTF magic number
996 and its counterpart in reversed endianness. The events within the
997 meta-data stream have no event header nor event context. Each event only
998 contains a special "sequence" payload, which is a sequence of bits which
999 length is implicitly calculated by using the
1000 "trace.packet.header.content_size" field, minus the packet header size.
1001 The formatting of this sequence of bits is a plain-text representation
1002 of the TSDL description. Each meta-data packet start with a special
1003 packet header, specific to the meta-data stream, which contains,
1006 struct metadata_packet_header {
1007 uint32_t magic; /* 0x75D11D57 */
1008 uint8_t uuid[16]; /* Unique Universal Identifier */
1009 uint32_t checksum; /* 0 if unused */
1010 uint32_t content_size; /* in bits */
1011 uint32_t packet_size; /* in bits */
1012 uint8_t compression_scheme; /* 0 if unused */
1013 uint8_t encryption_scheme; /* 0 if unused */
1014 uint8_t checksum_scheme; /* 0 if unused */
1015 uint8_t major; /* CTF spec version major number */
1016 uint8_t minor; /* CTF spec version minor number */
1019 The packet-based meta-data can be converted to a text-only meta-data by
1020 concatenating all the strings it contains.
1022 In the textual representation of the meta-data, the text contained
1023 within "/*" and "*/", as well as within "//" and end of line, are
1024 treated as comments. Boolean values can be represented as true, TRUE,
1025 or 1 for true, and false, FALSE, or 0 for false. Within the string-based
1026 meta-data description, the trace UUID is represented as a string of
1027 hexadecimal digits and dashes "-". In the event packet header, the trace
1028 UUID is represented as an array of bytes.
1031 7.2 Declaration vs Definition
1033 A declaration associates a layout to a type, without specifying where
1034 this type is located in the event structure hierarchy (see Section 6).
1035 This therefore includes typedef, typealias, as well as all type
1036 specifiers. In certain circumstances (typedef, structure field and
1037 variant field), a declaration is followed by a declarator, which specify
1038 the newly defined type name (for typedef), or the field name (for
1039 declarations located within structure and variants). Array and sequence,
1040 declared with square brackets ("[" "]"), are part of the declarator,
1041 similarly to C99. The enumeration base type is specified by
1042 ": enum_base", which is part of the type specifier. The variant tag
1043 name, specified between "<" ">", is also part of the type specifier.
1045 A definition associates a type to a location in the event structure
1046 hierarchy (see Section 6). This association is denoted by ":=", as shown
1052 TSDL uses three different types of scoping: a lexical scope is used for
1053 declarations and type definitions, and static and dynamic scopes are
1054 used for variants references to tag fields (with relative and absolute
1055 path lookups) and for sequence references to length fields.
1059 Each of "trace", "env", "stream", "event", "struct" and "variant" have
1060 their own nestable declaration scope, within which types can be declared
1061 using "typedef" and "typealias". A root declaration scope also contains
1062 all declarations located outside of any of the aforementioned
1063 declarations. An inner declaration scope can refer to type declared
1064 within its container lexical scope prior to the inner declaration scope.
1065 Redefinition of a typedef or typealias is not valid, although hiding an
1066 upper scope typedef or typealias is allowed within a sub-scope.
1068 7.3.2 Static and Dynamic Scopes
1070 A local static scope consists in the scope generated by the declaration
1071 of fields within a compound type. A static scope is a local static scope
1072 augmented with the nested sub-static-scopes it contains.
1074 A dynamic scope consists in the static scope augmented with the
1075 implicit event structure definition hierarchy presented at Section 6.
1077 Multiple declarations of the same field name within a local static scope
1078 is not valid. It is however valid to re-use the same field name in
1079 different local scopes.
1081 Nested static and dynamic scopes form lookup paths. These are used for
1082 variant tag and sequence length references. They are used at the variant
1083 and sequence definition site to look up the location of the tag field
1084 associated with a variant, and to lookup up the location of the length
1085 field associated with a sequence.
1087 Variants and sequences can refer to a tag field either using a relative
1088 path or an absolute path. The relative path is relative to the scope in
1089 which the variant or sequence performing the lookup is located.
1090 Relative paths are only allowed to lookup within the same static scope,
1091 which includes its nested static scopes. Lookups targeting parent static
1092 scopes need to be performed with an absolute path.
1094 Absolute path lookups use the full path including the dynamic scope
1095 followed by a "." and then the static scope. Therefore, variants (or
1096 sequences) in lower levels in the dynamic scope (e.g. event context) can
1097 refer to a tag (or length) field located in upper levels (e.g. in the
1098 event header) by specifying, in this case, the associated tag with
1099 <stream.event.header.field_name>. This allows, for instance, the event
1100 context to define a variant referring to the "id" field of the event
1103 The dynamic scope prefixes are thus:
1105 - Trace Environment: <env. >,
1106 - Trace Packet Header: <trace.packet.header. >,
1107 - Stream Packet Context: <stream.packet.context. >,
1108 - Event Header: <stream.event.header. >,
1109 - Stream Event Context: <stream.event.context. >,
1110 - Event Context: <event.context. >,
1111 - Event Payload: <event.fields. >.
1114 The target dynamic scope must be specified explicitly when referring to
1115 a field outside of the static scope (absolute scope reference). No
1116 conflict can occur between relative and dynamic paths, because the
1117 keywords "trace", "stream", and "event" are reserved, and thus
1118 not permitted as field names. It is recommended that field names
1119 clashing with CTF and C99 reserved keywords use an underscore prefix to
1120 eliminate the risk of generating a description containing an invalid
1121 field name. Consequently, fields starting with an underscore should have
1122 their leading underscore removed by the CTF trace readers.
1125 The information available in the dynamic scopes can be thought of as the
1126 current tracing context. At trace production, information about the
1127 current context is saved into the specified scope field levels. At trace
1128 consumption, for each event, the current trace context is therefore
1129 readable by accessing the upper dynamic scopes.
1134 The grammar representing the TSDL meta-data is presented in Appendix C.
1135 TSDL Grammar. This section presents a rather lighter reading that
1136 consists in examples of TSDL meta-data, with template values.
1138 The stream "id" can be left out if there is only one stream in the
1139 trace. The event "id" field can be left out if there is only one event
1143 major = value; /* CTF spec version major number */
1144 minor = value; /* CTF spec version minor number */
1145 uuid = "aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa"; /* Trace UUID */
1146 byte_order = be OR le; /* Endianness (required) */
1147 packet.header := struct {
1155 * The "env" (environment) scope contains assignment expressions. The
1156 * field names and content are implementation-defined.
1159 pid = value; /* example */
1160 proc_name = "name"; /* example */
1166 /* Type 1 - Few event IDs; Type 2 - Many event IDs. See section 6.1. */
1167 event.header := event_header_1 OR event_header_2;
1168 event.context := struct {
1171 packet.context := struct {
1177 name = "event_name";
1178 id = value; /* Numeric identifier within the stream */
1179 stream_id = stream_id;
1181 model.emf.uri = "string";
1191 name = "event_name";
1198 /* More detail on types in section 4. Types */
1203 * Type declarations behave similarly to the C standard.
1206 typedef aliased_type_specifiers new_type_declarators;
1208 /* e.g.: typedef struct example new_type_name[10]; */
1213 * The "typealias" declaration can be used to give a name (including
1214 * pointer declarator specifier) to a type. It should also be used to
1215 * map basic C types (float, int, unsigned long, ...) to a CTF type.
1216 * Typealias is a superset of "typedef": it also allows assignment of a
1217 * simple variable identifier to a type.
1220 typealias type_class {
1222 } := type_specifiers type_declarator;
1226 * typealias integer {
1230 * } := struct page *;
1232 * typealias integer {
1247 enum name : integer_type {
1253 * Unnamed types, contained within compound type fields, typedef or typealias.
1268 enum : integer_type {
1272 typedef type new_type[length];
1275 type field_name[length];
1278 typedef type new_type[length_type];
1281 type field_name[length_type];
1293 integer_type field_name:size; /* GNU/C bitfield */
1303 Clock metadata allows to describe the clock topology of the system, as
1304 well as to detail each clock parameter. In absence of clock description,
1305 it is assumed that all fields named "timestamp" use the same clock
1306 source, which increments once per nanosecond.
1308 Describing a clock and how it is used by streams is threefold: first,
1309 the clock and clock topology should be described in a "clock"
1310 description block, e.g.:
1313 name = cycle_counter_sync;
1314 uuid = "62189bee-96dc-11e0-91a8-cfa3d89f3923";
1315 description = "Cycle counter synchronized across CPUs";
1316 freq = 1000000000; /* frequency, in Hz */
1317 /* precision in seconds is: 1000 * (1/freq) */
1320 * clock value offset from Epoch is:
1321 * offset_s + (offset * (1/freq))
1323 offset_s = 1326476837;
1328 The mandatory "name" field specifies the name of the clock identifier,
1329 which can later be used as a reference. The optional field "uuid" is the
1330 unique identifier of the clock. It can be used to correlate different
1331 traces that use the same clock. An optional textual description string
1332 can be added with the "description" field. The "freq" field is the
1333 initial frequency of the clock, in Hz. If the "freq" field is not
1334 present, the frequency is assumed to be 1000000000 (providing clock
1335 increment of 1 ns). The optional "precision" field details the
1336 uncertainty on the clock measurements, in (1/freq) units. The "offset_s"
1337 and "offset" fields indicate the offset from POSIX.1 Epoch, 1970-01-01
1338 00:00:00 +0000 (UTC), to the zero of value of the clock. The "offset_s"
1339 field is in seconds. The "offset" field is in (1/freq) units. If any of
1340 the "offset_s" or "offset" field is not present, it is assigned the 0
1341 value. The field "absolute" is TRUE if the clock is a global reference
1342 across different clock uuid (e.g. NTP time). Otherwise, "absolute" is
1343 FALSE, and the clock can be considered as synchronized only with other
1344 clocks that have the same uuid.
1347 Secondly, a reference to this clock should be added within an integer
1351 size = 64; align = 1; signed = false;
1352 map = clock.cycle_counter_sync.value;
1355 Thirdly, stream declarations can reference the clock they use as a
1358 struct packet_context {
1359 uint64_ccnt_t ccnt_begin;
1360 uint64_ccnt_t ccnt_end;
1366 event.header := struct {
1367 uint64_ccnt_t timestamp;
1370 packet.context := struct packet_context;
1373 For a N-bit integer type referring to a clock, if the integer overflows
1374 compared to the N low order bits of the clock prior value, then it is
1375 assumed that one, and only one, overflow occurred. It is therefore
1376 important that events encoding time on a small number of bits happen
1377 frequently enough to detect when more than one N-bit overflow occurs.
1379 In a packet context, clock field names ending with "_begin" and "_end"
1380 have a special meaning: this refers to the time-stamps at, respectively,
1381 the beginning and the end of each packet.
1386 The two following macros keep track of the size of a GNU/C structure without
1387 padding at the end by placing HEADER_END as the last field. A one byte end field
1388 is used for C90 compatibility (C99 flexible arrays could be used here). Note
1389 that this does not affect the effective structure size, which should always be
1390 calculated with the header_sizeof() helper.
1392 #define HEADER_END char end_field
1393 #define header_sizeof(type) offsetof(typeof(type), end_field)
1396 B. Stream Header Rationale
1398 An event stream is divided in contiguous event packets of variable size. These
1399 subdivisions allow the trace analyzer to perform a fast binary search by time
1400 within the stream (typically requiring to index only the event packet headers)
1401 without reading the whole stream. These subdivisions have a variable size to
1402 eliminate the need to transfer the event packet padding when partially filled
1403 event packets must be sent when streaming a trace for live viewing/analysis.
1404 An event packet can contain a certain amount of padding at the end. Dividing
1405 streams into event packets is also useful for network streaming over UDP and
1406 flight recorder mode tracing (a whole event packet can be swapped out of the
1407 buffer atomically for reading).
1409 The stream header is repeated at the beginning of each event packet to allow
1410 flexibility in terms of:
1412 - streaming support,
1413 - allowing arbitrary buffers to be discarded without making the trace
1415 - allow UDP packet loss handling by either dealing with missing event packet
1416 or asking for re-transmission.
1417 - transparently support flight recorder mode,
1418 - transparently support crash dump.
1424 * Common Trace Format (CTF) Trace Stream Description Language (TSDL) Grammar.
1426 * Inspired from the C99 grammar:
1427 * http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1124.pdf (Annex A)
1428 * and c++1x grammar (draft)
1429 * http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3291.pdf (Annex A)
1431 * Specialized for CTF needs by including only constant and declarations from
1432 * C99 (excluding function declarations), and by adding support for variants,
1433 * sequences and CTF-specific specifiers. Enumeration container types
1434 * semantic is inspired from c++1x enum-base.
1439 1.1) Lexical elements
1486 identifier identifier-nondigit
1489 identifier-nondigit:
1491 universal-character-name
1492 any other implementation-defined characters
1496 [a-zA-Z] /* regular expression */
1499 [0-9] /* regular expression */
1501 1.4) Universal character names
1503 universal-character-name:
1505 \U hex-quad hex-quad
1508 hexadecimal-digit hexadecimal-digit hexadecimal-digit hexadecimal-digit
1514 enumeration-constant
1518 decimal-constant integer-suffix-opt
1519 octal-constant integer-suffix-opt
1520 hexadecimal-constant integer-suffix-opt
1524 decimal-constant digit
1528 octal-constant octal-digit
1530 hexadecimal-constant:
1531 hexadecimal-prefix hexadecimal-digit
1532 hexadecimal-constant hexadecimal-digit
1542 unsigned-suffix long-suffix-opt
1543 unsigned-suffix long-long-suffix
1544 long-suffix unsigned-suffix-opt
1545 long-long-suffix unsigned-suffix-opt
1559 enumeration-constant:
1565 L' c-char-sequence '
1569 c-char-sequence c-char
1572 any member of source charset except single-quote ('), backslash
1573 (\), or new-line character.
1577 simple-escape-sequence
1578 octal-escape-sequence
1579 hexadecimal-escape-sequence
1580 universal-character-name
1582 simple-escape-sequence: one of
1583 \' \" \? \\ \a \b \f \n \r \t \v
1585 octal-escape-sequence:
1587 \ octal-digit octal-digit
1588 \ octal-digit octal-digit octal-digit
1590 hexadecimal-escape-sequence:
1591 \x hexadecimal-digit
1592 hexadecimal-escape-sequence hexadecimal-digit
1594 1.6) String literals
1597 " s-char-sequence-opt "
1598 L" s-char-sequence-opt "
1602 s-char-sequence s-char
1605 any member of source charset except double-quote ("), backslash
1606 (\), or new-line character.
1612 [ ] ( ) { } . -> * + - < > : ; ... = ,
1615 2) Phrase structure grammar
1621 ( unary-expression )
1625 postfix-expression [ unary-expression ]
1626 postfix-expression . identifier
1627 postfix-expressoin -> identifier
1631 unary-operator postfix-expression
1633 unary-operator: one of
1636 assignment-operator:
1639 type-assignment-operator:
1642 constant-expression-range:
1643 unary-expression ... unary-expression
1648 declaration-specifiers declarator-list-opt ;
1651 declaration-specifiers:
1652 storage-class-specifier declaration-specifiers-opt
1653 type-specifier declaration-specifiers-opt
1654 type-qualifier declaration-specifiers-opt
1658 declarator-list , declarator
1660 abstract-declarator-list:
1662 abstract-declarator-list , abstract-declarator
1664 storage-class-specifier:
1687 align ( unary-expression )
1690 struct identifier-opt { struct-or-variant-declaration-list-opt } align-attribute-opt
1691 struct identifier align-attribute-opt
1693 struct-or-variant-declaration-list:
1694 struct-or-variant-declaration
1695 struct-or-variant-declaration-list struct-or-variant-declaration
1697 struct-or-variant-declaration:
1698 specifier-qualifier-list struct-or-variant-declarator-list ;
1699 declaration-specifiers-opt storage-class-specifier declaration-specifiers-opt declarator-list ;
1700 typealias declaration-specifiers abstract-declarator-list type-assignment-operator declaration-specifiers abstract-declarator-list ;
1701 typealias declaration-specifiers abstract-declarator-list type-assignment-operator declarator-list ;
1703 specifier-qualifier-list:
1704 type-specifier specifier-qualifier-list-opt
1705 type-qualifier specifier-qualifier-list-opt
1707 struct-or-variant-declarator-list:
1708 struct-or-variant-declarator
1709 struct-or-variant-declarator-list , struct-or-variant-declarator
1711 struct-or-variant-declarator:
1713 declarator-opt : unary-expression
1716 variant identifier-opt variant-tag-opt { struct-or-variant-declaration-list }
1717 variant identifier variant-tag
1720 < unary-expression >
1723 enum identifier-opt { enumerator-list }
1724 enum identifier-opt { enumerator-list , }
1726 enum identifier-opt : declaration-specifiers { enumerator-list }
1727 enum identifier-opt : declaration-specifiers { enumerator-list , }
1731 enumerator-list , enumerator
1734 enumeration-constant
1735 enumeration-constant assignment-operator unary-expression
1736 enumeration-constant assignment-operator constant-expression-range
1742 pointer-opt direct-declarator
1747 direct-declarator [ unary-expression ]
1749 abstract-declarator:
1750 pointer-opt direct-abstract-declarator
1752 direct-abstract-declarator:
1754 ( abstract-declarator )
1755 direct-abstract-declarator [ unary-expression ]
1756 direct-abstract-declarator [ ]
1759 * type-qualifier-list-opt
1760 * type-qualifier-list-opt pointer
1762 type-qualifier-list:
1764 type-qualifier-list type-qualifier
1769 2.3) CTF-specific declarations
1772 clock { ctf-assignment-expression-list-opt }
1773 event { ctf-assignment-expression-list-opt }
1774 stream { ctf-assignment-expression-list-opt }
1775 env { ctf-assignment-expression-list-opt }
1776 trace { ctf-assignment-expression-list-opt }
1777 callsite { ctf-assignment-expression-list-opt }
1778 typealias declaration-specifiers abstract-declarator-list type-assignment-operator declaration-specifiers abstract-declarator-list
1779 typealias declaration-specifiers abstract-declarator-list type-assignment-operator declarator-list
1782 floating_point { ctf-assignment-expression-list-opt }
1783 integer { ctf-assignment-expression-list-opt }
1784 string { ctf-assignment-expression-list-opt }
1787 ctf-assignment-expression-list:
1788 ctf-assignment-expression ;
1789 ctf-assignment-expression-list ctf-assignment-expression ;
1791 ctf-assignment-expression:
1792 unary-expression assignment-operator unary-expression
1793 unary-expression type-assignment-operator type-specifier
1794 declaration-specifiers-opt storage-class-specifier declaration-specifiers-opt declarator-list
1795 typealias declaration-specifiers abstract-declarator-list type-assignment-operator declaration-specifiers abstract-declarator-list
1796 typealias declaration-specifiers abstract-declarator-list type-assignment-operator declarator-list