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
29 1. Preliminary definitions
30 2. High-level representation of a trace
34 4.1.1 Type inheritance
44 4.2.2 Variants (Discriminated/Tagged Unions)
48 5. Event Packet Header
49 5.1 Event Packet Header Description
50 5.2 Event Packet Context Description
53 6.1.1 Type 1 - Few event IDs
54 6.1.2 Type 2 - Many event IDs
59 7. Trace Stream Description Language (TSDL)
61 7.2 Declaration vs Definition
68 1. Preliminary definitions
70 - Event Trace: An ordered sequence of events.
71 - Event Stream: An ordered sequence of events, containing a subset of the
73 - Event Packet: A sequence of physically contiguous events within an event
75 - Event: This is the basic entry in a trace. (aka: a trace record).
76 - An event identifier (ID) relates to the class (a type) of event within
78 e.g. event: irq_entry.
79 - An event (or event record) relates to a specific instance of an event
81 e.g. event: irq_entry, at time X, on CPU Y
82 - Source Architecture: Architecture writing the trace.
83 - Reader Architecture: Architecture reading the trace.
86 2. High-level representation of a trace
88 A trace is divided into multiple event streams. Each event stream contains a
89 subset of the trace event types.
91 The final output of the trace, after its generation and optional transport over
92 the network, is expected to be either on permanent or temporary storage in a
93 virtual file system. Because each event stream is appended to while a trace is
94 being recorded, each is associated with a separate file for output. Therefore,
95 a stored trace can be represented as a directory containing one file per stream.
97 Meta-data description associated with the trace contains information on
98 trace event types expressed in the Trace Stream Description Language
99 (TSDL). This language describes:
103 - Per-trace event header description.
104 - Per-stream event header description.
105 - Per-stream event context description.
107 - Event type to stream mapping.
108 - Event type to name mapping.
109 - Event type to ID mapping.
110 - Event context description.
111 - Event fields description.
116 An event stream can be divided into contiguous event packets of variable
117 size. These subdivisions have a variable size. An event packet can
118 contain a certain amount of padding at the end. The stream header is
119 repeated at the beginning of each event packet. The rationale for the
120 event stream design choices is explained in Appendix B. Stream Header
123 The event stream header will therefore be referred to as the "event packet
124 header" throughout the rest of this document.
129 Types are organized as type classes. Each type class belong to either of two
130 kind of types: basic types or compound types.
134 A basic type is a scalar type, as described in this section. It includes
135 integers, GNU/C bitfields, enumerations, and floating point values.
137 4.1.1 Type inheritance
139 Type specifications can be inherited to allow deriving types from a
140 type class. For example, see the uint32_t named type derived from the "integer"
141 type class below ("Integers" section). Types have a precise binary
142 representation in the trace. A type class has methods to read and write these
143 types, but must be derived into a type to be usable in an event field.
147 We define "byte-packed" types as aligned on the byte size, namely 8-bit.
148 We define "bit-packed" types as following on the next bit, as defined by the
151 Each basic type must specify its alignment, in bits. Examples of
152 possible alignments are: bit-packed (align = 1), byte-packed (align =
153 8), or word-aligned (e.g. align = 32 or align = 64). The choice depends
154 on the architecture preference and compactness vs performance trade-offs
155 of the implementation. Architectures providing fast unaligned write
156 byte-packed basic types to save space, aligning each type on byte
157 boundaries (8-bit). Architectures with slow unaligned writes align types
158 on specific alignment values. If no specific alignment is declared for a
159 type, it is assumed to be bit-packed for integers with size not multiple
160 of 8 bits and for gcc bitfields. All other basic types are byte-packed
161 by default. It is however recommended to always specify the alignment
162 explicitly. Alignment values must be power of two. Compound types are
163 aligned as specified in their individual specification.
165 TSDL meta-data attribute representation of a specific alignment:
167 align = value; /* value in bits */
171 By default, the native endianness of the source architecture the trace is used.
172 Byte order can be overridden for a basic type by specifying a "byte_order"
173 attribute. Typical use-case is to specify the network byte order (big endian:
174 "be") to save data captured from the network into the trace without conversion.
175 If not specified, the byte order is native.
177 TSDL meta-data representation:
179 byte_order = native OR network OR be OR le; /* network and be are aliases */
183 Type size, in bits, for integers and floats is that returned by "sizeof()" in C
184 multiplied by CHAR_BIT.
185 We require the size of "char" and "unsigned char" types (CHAR_BIT) to be fixed
186 to 8 bits for cross-endianness compatibility.
188 TSDL meta-data representation:
190 size = value; (value is in bits)
194 Signed integers are represented in two-complement. Integer alignment,
195 size, signedness and byte ordering are defined in the TSDL meta-data.
196 Integers aligned on byte size (8-bit) and with length multiple of byte
197 size (8-bit) correspond to the C99 standard integers. In addition,
198 integers with alignment and/or size that are _not_ a multiple of the
199 byte size are permitted; these correspond to the C99 standard bitfields,
200 with the added specification that the CTF integer bitfields have a fixed
201 binary representation. A MIT-licensed reference implementation of the
202 CTF portable bitfields is available at:
204 http://git.efficios.com/?p=babeltrace.git;a=blob;f=include/babeltrace/bitfield.h
206 Binary representation of integers:
208 - On little and big endian:
209 - Within a byte, high bits correspond to an integer high bits, and low bits
210 correspond to low bits.
212 - Integer across multiple bytes are placed from the less significant to the
214 - Consecutive integers are placed from lower bits to higher bits (even within
217 - Integer across multiple bytes are placed from the most significant to the
219 - Consecutive integers are placed from higher bits to lower bits (even within
222 This binary representation is derived from the bitfield implementation in GCC
223 for little and big endian. However, contrary to what GCC does, integers can
224 cross units boundaries (no padding is required). Padding can be explicitly
225 added (see 4.1.6 GNU/C bitfields) to follow the GCC layout if needed.
227 TSDL meta-data representation:
230 signed = true OR false; /* default false */
231 byte_order = native OR network OR be OR le; /* default native */
232 size = value; /* value in bits, no default */
233 align = value; /* value in bits */
234 /* based used for pretty-printing output, default: decimal. */
235 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
236 OR octal OR oct OR o OR 8 OR binary OR b OR 2;
237 /* character encoding, default: none */
238 encoding = none or UTF8 or ASCII;
241 Example of type inheritance (creation of a uint32_t named type):
249 Definition of a named 5-bit signed bitfield:
257 The character encoding field can be used to specify that the integer
258 must be printed as a text character when read. e.g.:
268 4.1.6 GNU/C bitfields
270 The GNU/C bitfields follow closely the integer representation, with a
271 particularity on alignment: if a bitfield cannot fit in the current unit, the
272 unit is padded and the bitfield starts at the following unit. The unit size is
273 defined by the size of the type "unit_type".
275 TSDL meta-data representation:
279 As an example, the following structure declared in C compiled by GCC:
286 The example structure is aligned on the largest element (short). The second
287 bitfield would be aligned on the next unit boundary, because it would not fit in
292 The floating point values byte ordering is defined in the TSDL meta-data.
294 Floating point values follow the IEEE 754-2008 standard interchange formats.
295 Description of the floating point values include the exponent and mantissa size
296 in bits. Some requirements are imposed on the floating point values:
298 - FLT_RADIX must be 2.
299 - mant_dig is the number of digits represented in the mantissa. It is specified
300 by the ISO C99 standard, section 5.2.4, as FLT_MANT_DIG, DBL_MANT_DIG and
301 LDBL_MANT_DIG as defined by <float.h>.
302 - exp_dig is the number of digits represented in the exponent. Given that
303 mant_dig is one bit more than its actual size in bits (leading 1 is not
304 needed) and also given that the sign bit always takes one bit, exp_dig can be
307 - sizeof(float) * CHAR_BIT - FLT_MANT_DIG
308 - sizeof(double) * CHAR_BIT - DBL_MANT_DIG
309 - sizeof(long double) * CHAR_BIT - LDBL_MANT_DIG
311 TSDL meta-data representation:
316 byte_order = native OR network OR be OR le;
320 Example of type inheritance:
322 typealias floating_point {
323 exp_dig = 8; /* sizeof(float) * CHAR_BIT - FLT_MANT_DIG */
324 mant_dig = 24; /* FLT_MANT_DIG */
329 TODO: define NaN, +inf, -inf behavior.
331 Bit-packed, byte-packed or larger alignments can be used for floating
332 point values, similarly to integers.
336 Enumerations are a mapping between an integer type and a table of strings. The
337 numerical representation of the enumeration follows the integer type specified
338 by the meta-data. The enumeration mapping table is detailed in the enumeration
339 description within the meta-data. The mapping table maps inclusive value
340 ranges (or single values) to strings. Instead of being limited to simple
341 "value -> string" mappings, these enumerations map
342 "[ start_value ... end_value ] -> string", which map inclusive ranges of
343 values to strings. An enumeration from the C language can be represented in
344 this format by having the same start_value and end_value for each element, which
345 is in fact a range of size 1. This single-value range is supported without
346 repeating the start and end values with the value = string declaration.
348 enum name : integer_type {
349 somestring = start_value1 ... end_value1,
350 "other string" = start_value2 ... end_value2,
351 yet_another_string, /* will be assigned to end_value2 + 1 */
352 "some other string" = value,
356 If the values are omitted, the enumeration starts at 0 and increment of 1 for
359 enum name : unsigned int {
367 Overlapping ranges within a single enumeration are implementation defined.
369 A nameless enumeration can be declared as a field type or as part of a typedef:
371 enum : integer_type {
375 Enumerations omitting the container type ": integer_type" use the "int"
376 type (for compatibility with C99). The "int" type must be previously
379 typealias integer { size = 32; align = 32; signed = true } := int;
388 Compound are aggregation of type declarations. Compound types include
389 structures, variant, arrays, sequences, and strings.
393 Structures are aligned on the largest alignment required by basic types
394 contained within the structure. (This follows the ISO/C standard for structures)
396 TSDL meta-data representation of a named structure:
399 field_type field_name;
400 field_type field_name;
407 integer { /* Nameless type */
412 uint64_t second_field_name; /* Named type declared in the meta-data */
415 The fields are placed in a sequence next to each other. They each possess a
416 field name, which is a unique identifier within the structure.
418 A nameless structure can be declared as a field type or as part of a typedef:
424 Alignment for a structure compound type can be forced to a minimum value
425 by adding an "align" specifier after the declaration of a structure
426 body. This attribute is read as: align(value). The value is specified in
427 bits. The structure will be aligned on the maximum value between this
428 attribute and the alignment required by the basic types contained within
435 4.2.2 Variants (Discriminated/Tagged Unions)
437 A CTF variant is a selection between different types. A CTF variant must
438 always be defined within the scope of a structure or within fields
439 contained within a structure (defined recursively). A "tag" enumeration
440 field must appear in either the same lexical scope, prior to the variant
441 field (in field declaration order), in an upper lexical scope (see
442 Section 7.3.1), or in an upper dynamic scope (see Section 7.3.2). The
443 type selection is indicated by the mapping from the enumeration value to
444 the string used as variant type selector. The field to use as tag is
445 specified by the "tag_field", specified between "< >" after the
446 "variant" keyword for unnamed variants, and after "variant name" for
449 The alignment of the variant is the alignment of the type as selected by the tag
450 value for the specific instance of the variant. The alignment of the type
451 containing the variant is independent of the variant alignment. The size of the
452 variant is the size as selected by the tag value for the specific instance of
455 A named variant declaration followed by its definition within a structure
466 enum : integer_type { sel1, sel2, sel3, ... } tag_field;
468 variant name <tag_field> v;
471 An unnamed variant definition within a structure is expressed by the following
475 enum : integer_type { sel1, sel2, sel3, ... } tag_field;
477 variant <tag_field> {
485 Example of a named variant within a sequence that refers to a single tag field:
494 enum : uint2_t { a, b, c } choice;
496 variant example <choice> v[seqlen];
499 Example of an unnamed variant:
502 enum : uint2_t { a, b, c, d } choice;
503 /* Unrelated fields can be added between the variant and its tag */
516 Example of an unnamed variant within an array:
519 enum : uint2_t { a, b, c } choice;
527 Example of a variant type definition within a structure, where the defined type
528 is then declared within an array of structures. This variant refers to a tag
529 located in an upper lexical scope. This example clearly shows that a variant
530 type definition referring to the tag "x" uses the closest preceding field from
531 the lexical scope of the type definition.
534 enum : uint2_t { a, b, c, d } x;
536 typedef variant <x> { /*
537 * "x" refers to the preceding "x" enumeration in the
538 * lexical scope of the type definition.
546 enum : int { x, y, z } x; /* This enumeration is not used by "v". */
547 example_variant v; /*
548 * "v" uses the "enum : uint2_t { a, b, c, d }"
556 Arrays are fixed-length. Their length is declared in the type
557 declaration within the meta-data. They contain an array of "inner type"
558 elements, which can refer to any type not containing the type of the
559 array being declared (no circular dependency). The length is the number
560 of elements in an array.
562 TSDL meta-data representation of a named array:
564 typedef elem_type name[length];
566 A nameless array can be declared as a field type within a structure, e.g.:
568 uint8_t field_name[10];
570 Arrays are always aligned on their element alignment requirement.
574 Sequences are dynamically-sized arrays. They refer to a a "length"
575 unsigned integer field, which must appear in either the same lexical scope,
576 prior to the sequence field (in field declaration order), in an upper
577 lexical scope (see Section 7.3.1), or in an upper dynamic scope (see
578 Section 7.3.2). This length field represents the number of elements in
579 the sequence. The sequence per se is an array of "inner type" elements.
581 TSDL meta-data representation for a sequence type definition:
584 unsigned int length_field;
585 typedef elem_type typename[length_field];
586 typename seq_field_name;
589 A sequence can also be declared as a field type, e.g.:
592 unsigned int length_field;
593 long seq_field_name[length_field];
596 Multiple sequences can refer to the same length field, and these length
597 fields can be in a different upper dynamic scope:
599 e.g., assuming the stream.event.header defines:
604 event.header := struct {
613 long seq_a[stream.event.header.seq_len];
614 char seq_b[stream.event.header.seq_len];
618 The sequence elements follow the "array" specifications.
622 Strings are an array of bytes of variable size and are terminated by a '\0'
623 "NULL" character. Their encoding is described in the TSDL meta-data. In
624 absence of encoding attribute information, the default encoding is
627 TSDL meta-data representation of a named string type:
630 encoding = UTF8 OR ASCII;
633 A nameless string type can be declared as a field type:
635 string field_name; /* Use default UTF8 encoding */
637 Strings are always aligned on byte size.
639 5. Event Packet Header
641 The event packet header consists of two parts: the "event packet header"
642 is the same for all streams of a trace. The second part, the "event
643 packet context", is described on a per-stream basis. Both are described
644 in the TSDL meta-data. The packets are aligned on architecture-page-sized
647 Event packet header (all fields are optional, specified by TSDL meta-data):
649 - Magic number (CTF magic number: 0xC1FC1FC1) specifies that this is a
650 CTF packet. This magic number is optional, but when present, it should
651 come at the very beginning of the packet.
652 - Trace UUID, used to ensure the event packet match the meta-data used.
653 (note: we cannot use a meta-data checksum in every cases instead of a
654 UUID because meta-data can be appended to while tracing is active)
655 This field is optional.
656 - Stream ID, used as reference to stream description in meta-data.
657 This field is optional if there is only one stream description in the
658 meta-data, but becomes required if there are more than one stream in
659 the TSDL meta-data description.
661 Event packet context (all fields are optional, specified by TSDL meta-data):
663 - Event packet content size (in bytes).
664 - Event packet size (in bytes, includes padding).
665 - Event packet content checksum (optional). Checksum excludes the event packet
667 - Per-stream event packet sequence count (to deal with UDP packet loss). The
668 number of significant sequence counter bits should also be present, so
669 wrap-arounds are dealt with correctly.
670 - Time-stamp at the beginning and time-stamp at the end of the event packet.
671 Both timestamps are written in the packet header, but sampled respectively
672 while (or before) writing the first event and while (or after) writing the
673 last event in the packet. The inclusive range between these timestamps should
674 include all event timestamps assigned to events contained within the packet.
675 - Events discarded count
676 - Snapshot of a per-stream free-running counter, counting the number of
677 events discarded that were supposed to be written in the stream prior to
678 the first event in the event packet.
679 * Note: producer-consumer buffer full condition should fill the current
680 event packet with padding so we know exactly where events have been
682 - Lossless compression scheme used for the event packet content. Applied
683 directly to raw data. New types of compression can be added in following
684 versions of the format.
685 0: no compression scheme
689 - Cypher used for the event packet content. Applied after compression.
692 - Checksum scheme used for the event packet content. Applied after encryption.
698 5.1 Event Packet Header Description
700 The event packet header layout is indicated by the trace packet.header
701 field. Here is a recommended structure type for the packet header with
702 the fields typically expected (although these fields are each optional):
704 struct event_packet_header {
712 packet.header := struct event_packet_header;
715 If the magic number is not present, tools such as "file" will have no
716 mean to discover the file type.
718 If the uuid is not present, no validation that the meta-data actually
719 corresponds to the stream is performed.
721 If the stream_id packet header field is missing, the trace can only
722 contain a single stream. Its "id" field can be left out, and its events
723 don't need to declare a "stream_id" field.
726 5.2 Event Packet Context Description
728 Event packet context example. These are declared within the stream declaration
729 in the meta-data. All these fields are optional. If the packet size field is
730 missing, the whole stream only contains a single packet. If the content
731 size field is missing, the packet is filled (no padding). The content
732 and packet sizes include all headers.
734 An example event packet context type:
736 struct event_packet_context {
737 uint64_t timestamp_begin;
738 uint64_t timestamp_end;
740 uint32_t stream_packet_count;
741 uint32_t events_discarded;
743 uint32_t/uint16_t content_size;
744 uint32_t/uint16_t packet_size;
745 uint8_t stream_packet_count_bits; /* Significant counter bits */
746 uint8_t compression_scheme;
747 uint8_t encryption_scheme;
748 uint8_t checksum_scheme;
754 The overall structure of an event is:
756 1 - Stream Packet Context (as specified by the stream meta-data)
757 2 - Event Header (as specified by the stream meta-data)
758 3 - Stream Event Context (as specified by the stream meta-data)
759 4 - Event Context (as specified by the event meta-data)
760 5 - Event Payload (as specified by the event meta-data)
762 This structure defines an implicit dynamic scoping, where variants
763 located in inner structures (those with a higher number in the listing
764 above) can refer to the fields of outer structures (with lower number in
765 the listing above). See Section 7.3 TSDL Scopes for more detail.
769 Event headers can be described within the meta-data. We hereby propose, as an
770 example, two types of events headers. Type 1 accommodates streams with less than
771 31 event IDs. Type 2 accommodates streams with 31 or more event IDs.
773 One major factor can vary between streams: the number of event IDs assigned to
774 a stream. Luckily, this information tends to stay relatively constant (modulo
775 event registration while trace is being recorded), so we can specify different
776 representations for streams containing few event IDs and streams containing
777 many event IDs, so we end up representing the event ID and time-stamp as
778 densely as possible in each case.
780 The header is extended in the rare occasions where the information cannot be
781 represented in the ranges available in the standard event header. They are also
782 used in the rare occasions where the data required for a field could not be
783 collected: the flag corresponding to the missing field within the missing_fields
784 array is then set to 1.
786 Types uintX_t represent an X-bit unsigned integer, as declared with
789 typealias integer { size = X; align = X; signed = false } := uintX_t;
793 typealias integer { size = X; align = 1; signed = false } := uintX_t;
795 6.1.1 Type 1 - Few event IDs
797 - Aligned on 32-bit (or 8-bit if byte-packed, depending on the architecture
799 - Native architecture byte ordering.
800 - For "compact" selection
801 - Fixed size: 32 bits.
802 - For "extended" selection
803 - Size depends on the architecture and variant alignment.
805 struct event_header_1 {
808 * id 31 is reserved to indicate an extended header.
810 enum : uint5_t { compact = 0 ... 30, extended = 31 } id;
816 uint32_t id; /* 32-bit event IDs */
817 uint64_t timestamp; /* 64-bit timestamps */
820 } align(32); /* or align(8) */
823 6.1.2 Type 2 - Many event IDs
825 - Aligned on 16-bit (or 8-bit if byte-packed, depending on the architecture
827 - Native architecture byte ordering.
828 - For "compact" selection
829 - Size depends on the architecture and variant alignment.
830 - For "extended" selection
831 - Size depends on the architecture and variant alignment.
833 struct event_header_2 {
835 * id: range: 0 - 65534.
836 * id 65535 is reserved to indicate an extended header.
838 enum : uint16_t { compact = 0 ... 65534, extended = 65535 } id;
844 uint32_t id; /* 32-bit event IDs */
845 uint64_t timestamp; /* 64-bit timestamps */
848 } align(16); /* or align(8) */
853 The event context contains information relative to the current event.
854 The choice and meaning of this information is specified by the TSDL
855 stream and event meta-data descriptions. The stream context is applied
856 to all events within the stream. The stream context structure follows
857 the event header. The event context is applied to specific events. Its
858 structure follows the stream context structure.
860 An example of stream-level event context is to save the event payload size with
861 each event, or to save the current PID with each event. These are declared
862 within the stream declaration within the meta-data:
866 event.context := struct {
868 uint16_t payload_size;
872 An example of event-specific event context is to declare a bitmap of missing
873 fields, only appended after the stream event context if the extended event
874 header is selected. NR_FIELDS is the number of fields within the event (a
882 uint1_t missing_fields[NR_FIELDS]; /* missing event fields bitmap */
891 An event payload contains fields specific to a given event type. The fields
892 belonging to an event type are described in the event-specific meta-data
893 within a structure type.
897 No padding at the end of the event payload. This differs from the ISO/C standard
898 for structures, but follows the CTF standard for structures. In a trace, even
899 though it makes sense to align the beginning of a structure, it really makes no
900 sense to add padding at the end of the structure, because structures are usually
901 not followed by a structure of the same type.
903 This trick can be done by adding a zero-length "end" field at the end of the C
904 structures, and by using the offset of this field rather than using sizeof()
905 when calculating the size of a structure (see Appendix "A. Helper macros").
909 The event payload is aligned on the largest alignment required by types
910 contained within the payload. (This follows the ISO/C standard for structures)
913 7. Trace Stream Description Language (TSDL)
915 The Trace Stream Description Language (TSDL) allows expression of the
916 binary trace streams layout in a C99-like Domain Specific Language
922 The trace stream layout description is located in the trace meta-data.
923 The meta-data is itself located in a stream identified by its name:
926 The meta-data description can be expressed in two different formats:
927 text-only and packet-based. The text-only description facilitates
928 generation of meta-data and provides a convenient way to enter the
929 meta-data information by hand. The packet-based meta-data provides the
930 CTF stream packet facilities (checksumming, compression, encryption,
931 network-readiness) for meta-data stream generated and transported by a
934 The text-only meta-data file is a plain text TSDL description.
936 The packet-based meta-data is made of "meta-data packets", which each
937 start with a meta-data packet header. The packet-based meta-data
938 description is detected by reading the magic number "0x75D11D57" at the
939 beginning of the file. This magic number is also used to detect the
940 endianness of the architecture by trying to read the CTF magic number
941 and its counterpart in reversed endianness. The events within the
942 meta-data stream have no event header nor event context. Each event only
943 contains a "sequence" payload, which is a sequence of bits using the
944 "trace.packet.header.content_size" field as a placeholder for its length
945 (the packet header size should be substracted). The formatting of this
946 sequence of bits is a plain-text representation of the TSDL description.
947 Each meta-data packet start with a special packet header, specific to
948 the meta-data stream, which contains, exactly:
950 struct metadata_packet_header {
951 uint32_t magic; /* 0x75D11D57 */
952 uint8_t uuid[16]; /* Unique Universal Identifier */
953 uint32_t checksum; /* 0 if unused */
954 uint32_t content_size; /* in bits */
955 uint32_t packet_size; /* in bits */
956 uint8_t compression_scheme; /* 0 if unused */
957 uint8_t encryption_scheme; /* 0 if unused */
958 uint8_t checksum_scheme; /* 0 if unused */
961 The packet-based meta-data can be converted to a text-only meta-data by
962 concatenating all the strings in contains.
964 In the textual representation of the meta-data, the text contained
965 within "/*" and "*/", as well as within "//" and end of line, are
966 treated as comments. Boolean values can be represented as true, TRUE,
967 or 1 for true, and false, FALSE, or 0 for false. Within the string-based
968 meta-data description, the trace UUID is represented as a string of
969 hexadecimal digits and dashes "-". In the event packet header, the trace
970 UUID is represented as an array of bytes.
973 7.2 Declaration vs Definition
975 A declaration associates a layout to a type, without specifying where
976 this type is located in the event structure hierarchy (see Section 6).
977 This therefore includes typedef, typealias, as well as all type
978 specifiers. In certain circumstances (typedef, structure field and
979 variant field), a declaration is followed by a declarator, which specify
980 the newly defined type name (for typedef), or the field name (for
981 declarations located within structure and variants). Array and sequence,
982 declared with square brackets ("[" "]"), are part of the declarator,
983 similarly to C99. The enumeration base type is specified by
984 ": enum_base", which is part of the type specifier. The variant tag
985 name, specified between "<" ">", is also part of the type specifier.
987 A definition associates a type to a location in the event structure
988 hierarchy (see Section 6). This association is denoted by ":=", as shown
994 TSDL uses two different types of scoping: a lexical scope is used for
995 declarations and type definitions, and a dynamic scope is used for
996 variants references to tag fields and for sequence references to length
1001 Each of "trace", "stream", "event", "struct" and "variant" have their own
1002 nestable declaration scope, within which types can be declared using "typedef"
1003 and "typealias". A root declaration scope also contains all declarations
1004 located outside of any of the aforementioned declarations. An inner
1005 declaration scope can refer to type declared within its container
1006 lexical scope prior to the inner declaration scope. Redefinition of a
1007 typedef or typealias is not valid, although hiding an upper scope
1008 typedef or typealias is allowed within a sub-scope.
1012 A dynamic scope consists in the lexical scope augmented with the
1013 implicit event structure definition hierarchy presented at Section 6.
1014 The dynamic scope is used for variant tag and sequence length
1015 definitions. It is used at definition time to look up the location of
1016 the tag field associated with a variant, and to lookup up the location
1017 of the length field associated with a sequence.
1019 Therefore, variants (or sequences) in lower levels in the dynamic scope
1020 (e.g. event context) can refer to a tag (or length) field located in
1021 upper levels (e.g. in the event header) by specifying, in this case, the
1022 associated tag with <header.field_name>. This allows, for instance, the
1023 event context to define a variant referring to the "id" field of the
1024 event header as selector.
1026 The target dynamic scope must be specified explicitly when referring to
1027 a field outside of the local static scope. The dynamic scope prefixes
1030 - Trace Packet Header: <trace.packet.header. >,
1031 - Stream Packet Context: <stream.packet.context. >,
1032 - Event Header: <stream.event.header. >,
1033 - Stream Event Context: <stream.event.context. >,
1034 - Event Context: <event.context. >,
1035 - Event Payload: <event.fields. >.
1037 Multiple declarations of the same field name within a single scope is
1038 not valid. It is however valid to re-use the same field name in
1039 different scopes. There is no possible conflict, because the dynamic
1040 scope must be specified when a variant refers to a tag field located in
1041 a different dynamic scope.
1043 The information available in the dynamic scopes can be thought of as the
1044 current tracing context. At trace production, information about the
1045 current context is saved into the specified scope field levels. At trace
1046 consumption, for each event, the current trace context is therefore
1047 readable by accessing the upper dynamic scopes.
1052 The grammar representing the TSDL meta-data is presented in Appendix C.
1053 TSDL Grammar. This section presents a rather lighter reading that
1054 consists in examples of TSDL meta-data, with template values.
1056 The stream "id" can be left out if there is only one stream in the
1057 trace. The event "id" field can be left out if there is only one event
1061 major = value; /* Trace format version */
1063 uuid = "aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa"; /* Trace UUID */
1064 byte_order = be OR le; /* Endianness (required) */
1065 packet.header := struct {
1074 /* Type 1 - Few event IDs; Type 2 - Many event IDs. See section 6.1. */
1075 event.header := event_header_1 OR event_header_2;
1076 event.context := struct {
1079 packet.context := struct {
1086 id = value; /* Numeric identifier within the stream */
1087 stream_id = stream_id;
1096 /* More detail on types in section 4. Types */
1101 * Type declarations behave similarly to the C standard.
1104 typedef aliased_type_specifiers new_type_declarators;
1106 /* e.g.: typedef struct example new_type_name[10]; */
1111 * The "typealias" declaration can be used to give a name (including
1112 * pointer declarator specifier) to a type. It should also be used to
1113 * map basic C types (float, int, unsigned long, ...) to a CTF type.
1114 * Typealias is a superset of "typedef": it also allows assignment of a
1115 * simple variable identifier to a type.
1118 typealias type_class {
1120 } := type_specifiers type_declarator;
1124 * typealias integer {
1128 * } := struct page *;
1130 * typealias integer {
1145 enum name : integer_type {
1151 * Unnamed types, contained within compound type fields, typedef or typealias.
1166 enum : integer_type {
1170 typedef type new_type[length];
1173 type field_name[length];
1176 typedef type new_type[length_type];
1179 type field_name[length_type];
1191 integer_type field_name:size; /* GNU/C bitfield */
1201 The two following macros keep track of the size of a GNU/C structure without
1202 padding at the end by placing HEADER_END as the last field. A one byte end field
1203 is used for C90 compatibility (C99 flexible arrays could be used here). Note
1204 that this does not affect the effective structure size, which should always be
1205 calculated with the header_sizeof() helper.
1207 #define HEADER_END char end_field
1208 #define header_sizeof(type) offsetof(typeof(type), end_field)
1211 B. Stream Header Rationale
1213 An event stream is divided in contiguous event packets of variable size. These
1214 subdivisions allow the trace analyzer to perform a fast binary search by time
1215 within the stream (typically requiring to index only the event packet headers)
1216 without reading the whole stream. These subdivisions have a variable size to
1217 eliminate the need to transfer the event packet padding when partially filled
1218 event packets must be sent when streaming a trace for live viewing/analysis.
1219 An event packet can contain a certain amount of padding at the end. Dividing
1220 streams into event packets is also useful for network streaming over UDP and
1221 flight recorder mode tracing (a whole event packet can be swapped out of the
1222 buffer atomically for reading).
1224 The stream header is repeated at the beginning of each event packet to allow
1225 flexibility in terms of:
1227 - streaming support,
1228 - allowing arbitrary buffers to be discarded without making the trace
1230 - allow UDP packet loss handling by either dealing with missing event packet
1231 or asking for re-transmission.
1232 - transparently support flight recorder mode,
1233 - transparently support crash dump.
1239 * Common Trace Format (CTF) Trace Stream Description Language (TSDL) Grammar.
1241 * Inspired from the C99 grammar:
1242 * http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1124.pdf (Annex A)
1243 * and c++1x grammar (draft)
1244 * http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3291.pdf (Annex A)
1246 * Specialized for CTF needs by including only constant and declarations from
1247 * C99 (excluding function declarations), and by adding support for variants,
1248 * sequences and CTF-specific specifiers. Enumeration container types
1249 * semantic is inspired from c++1x enum-base.
1254 1.1) Lexical elements
1298 identifier identifier-nondigit
1301 identifier-nondigit:
1303 universal-character-name
1304 any other implementation-defined characters
1308 [a-zA-Z] /* regular expression */
1311 [0-9] /* regular expression */
1313 1.4) Universal character names
1315 universal-character-name:
1317 \U hex-quad hex-quad
1320 hexadecimal-digit hexadecimal-digit hexadecimal-digit hexadecimal-digit
1326 enumeration-constant
1330 decimal-constant integer-suffix-opt
1331 octal-constant integer-suffix-opt
1332 hexadecimal-constant integer-suffix-opt
1336 decimal-constant digit
1340 octal-constant octal-digit
1342 hexadecimal-constant:
1343 hexadecimal-prefix hexadecimal-digit
1344 hexadecimal-constant hexadecimal-digit
1354 unsigned-suffix long-suffix-opt
1355 unsigned-suffix long-long-suffix
1356 long-suffix unsigned-suffix-opt
1357 long-long-suffix unsigned-suffix-opt
1371 enumeration-constant:
1377 L' c-char-sequence '
1381 c-char-sequence c-char
1384 any member of source charset except single-quote ('), backslash
1385 (\), or new-line character.
1389 simple-escape-sequence
1390 octal-escape-sequence
1391 hexadecimal-escape-sequence
1392 universal-character-name
1394 simple-escape-sequence: one of
1395 \' \" \? \\ \a \b \f \n \r \t \v
1397 octal-escape-sequence:
1399 \ octal-digit octal-digit
1400 \ octal-digit octal-digit octal-digit
1402 hexadecimal-escape-sequence:
1403 \x hexadecimal-digit
1404 hexadecimal-escape-sequence hexadecimal-digit
1406 1.6) String literals
1409 " s-char-sequence-opt "
1410 L" s-char-sequence-opt "
1414 s-char-sequence s-char
1417 any member of source charset except double-quote ("), backslash
1418 (\), or new-line character.
1424 [ ] ( ) { } . -> * + - < > : ; ... = ,
1427 2) Phrase structure grammar
1433 ( unary-expression )
1437 postfix-expression [ unary-expression ]
1438 postfix-expression . identifier
1439 postfix-expressoin -> identifier
1443 unary-operator postfix-expression
1445 unary-operator: one of
1448 assignment-operator:
1451 type-assignment-operator:
1454 constant-expression-range:
1455 unary-expression ... unary-expression
1460 declaration-specifiers declarator-list-opt ;
1463 declaration-specifiers:
1464 storage-class-specifier declaration-specifiers-opt
1465 type-specifier declaration-specifiers-opt
1466 type-qualifier declaration-specifiers-opt
1470 declarator-list , declarator
1472 abstract-declarator-list:
1474 abstract-declarator-list , abstract-declarator
1476 storage-class-specifier:
1499 align ( unary-expression )
1502 struct identifier-opt { struct-or-variant-declaration-list-opt } align-attribute-opt
1503 struct identifier align-attribute-opt
1505 struct-or-variant-declaration-list:
1506 struct-or-variant-declaration
1507 struct-or-variant-declaration-list struct-or-variant-declaration
1509 struct-or-variant-declaration:
1510 specifier-qualifier-list struct-or-variant-declarator-list ;
1511 declaration-specifiers-opt storage-class-specifier declaration-specifiers-opt declarator-list ;
1512 typealias declaration-specifiers abstract-declarator-list type-assignment-operator declaration-specifiers abstract-declarator-list ;
1513 typealias declaration-specifiers abstract-declarator-list type-assignment-operator declarator-list ;
1515 specifier-qualifier-list:
1516 type-specifier specifier-qualifier-list-opt
1517 type-qualifier specifier-qualifier-list-opt
1519 struct-or-variant-declarator-list:
1520 struct-or-variant-declarator
1521 struct-or-variant-declarator-list , struct-or-variant-declarator
1523 struct-or-variant-declarator:
1525 declarator-opt : unary-expression
1528 variant identifier-opt variant-tag-opt { struct-or-variant-declaration-list }
1529 variant identifier variant-tag
1535 enum identifier-opt { enumerator-list }
1536 enum identifier-opt { enumerator-list , }
1538 enum identifier-opt : declaration-specifiers { enumerator-list }
1539 enum identifier-opt : declaration-specifiers { enumerator-list , }
1543 enumerator-list , enumerator
1546 enumeration-constant
1547 enumeration-constant assignment-operator unary-expression
1548 enumeration-constant assignment-operator constant-expression-range
1554 pointer-opt direct-declarator
1559 direct-declarator [ unary-expression ]
1561 abstract-declarator:
1562 pointer-opt direct-abstract-declarator
1564 direct-abstract-declarator:
1566 ( abstract-declarator )
1567 direct-abstract-declarator [ unary-expression ]
1568 direct-abstract-declarator [ ]
1571 * type-qualifier-list-opt
1572 * type-qualifier-list-opt pointer
1574 type-qualifier-list:
1576 type-qualifier-list type-qualifier
1581 2.3) CTF-specific declarations
1584 event { ctf-assignment-expression-list-opt }
1585 stream { ctf-assignment-expression-list-opt }
1586 trace { ctf-assignment-expression-list-opt }
1587 typealias declaration-specifiers abstract-declarator-list type-assignment-operator declaration-specifiers abstract-declarator-list
1588 typealias declaration-specifiers abstract-declarator-list type-assignment-operator declarator-list
1591 floating_point { ctf-assignment-expression-list-opt }
1592 integer { ctf-assignment-expression-list-opt }
1593 string { ctf-assignment-expression-list-opt }
1596 ctf-assignment-expression-list:
1597 ctf-assignment-expression ;
1598 ctf-assignment-expression-list ctf-assignment-expression ;
1600 ctf-assignment-expression:
1601 unary-expression assignment-operator unary-expression
1602 unary-expression type-assignment-operator type-specifier
1603 declaration-specifiers-opt storage-class-specifier declaration-specifiers-opt declarator-list
1604 typealias declaration-specifiers abstract-declarator-list type-assignment-operator declaration-specifiers abstract-declarator-list
1605 typealias declaration-specifiers abstract-declarator-list type-assignment-operator declarator-list