5 This document describes the on-disk perf.data format, generated by perf record
6 or perf inject and consumed by the other perf tools.
8 On a high level perf.data contains the events generated by the PMUs, plus metadata.
10 All fields are in native-endian of the machine that generated the perf.data.
12 When perf is writing to a pipe it uses a special version of the file
13 format that does not rely on seeking to adjust data offsets. This
14 format is described in "Pipe-mode data" section. The pipe data version can be
15 augmented with additional events using perf inject.
17 The file starts with a perf_header:
20 char magic[8]; /* PERFILE2 */
21 uint64_t size; /* size of the header */
22 uint64_t attr_size; /* size of an attribute in attrs */
23 struct perf_file_section attrs;
24 struct perf_file_section data;
25 struct perf_file_section event_types;
30 The magic number identifies the perf file and the version. Current perf versions
31 use PERFILE2. Old perf versions generated a version 1 format (PERFFILE). Version 1
32 is not described here. The magic number also identifies the endian. When the
33 magic value is 64bit byte swapped compared the file is in non-native
36 A perf_file_section contains a pointer to another section of the perf file.
37 The header contains three such pointers: for attributes, data and event types.
39 struct perf_file_section {
40 uint64_t offset; /* offset from start of file */
41 uint64_t size; /* size of the section */
46 The header is followed by different optional headers, described by the bits set
47 in flags. Only headers for which the bit is set are included. Each header
48 consists of a perf_file_section located after the initial header.
49 The respective perf_file_section points to the data of the additional
50 header and defines its size.
52 Some headers consist of strings, which are defined like this:
54 struct perf_header_string {
56 char string[len]; /* zero terminated */
59 Some headers consist of a sequence of strings, which start with a
61 struct perf_header_string_list {
63 struct perf_header_string strings[nr]; /* variable length records */
66 The bits are the flags bits in a 256 bit bitmap starting with
67 flags. These define the valid bits:
69 HEADER_RESERVED = 0, /* always cleared */
70 HEADER_FIRST_FEATURE = 1,
71 HEADER_TRACING_DATA = 1,
77 The header consists of an sequence of build_id_event. The size of each record
78 is defined by header.size (see perf_event.h). Each event defines a ELF build id
79 for a executable file name for a pid. An ELF build id is a unique identifier
80 assigned by the linker to an executable.
82 struct build_id_event {
83 struct perf_event_header header;
86 char filename[header.size - offsetof(struct build_id_event, filename)];
91 A perf_header_string with the hostname where the data was collected
96 A perf_header_string with the os release where the data was collected
101 A perf_header_string with the perf user tool version where the
102 data was collected. This is the same as the version of the source tree
103 the perf tool was built from.
107 A perf_header_string with the CPU architecture (uname -m)
111 A structure defining the number of CPUs.
114 uint32_t nr_cpus_online;
115 uint32_t nr_cpus_available; /* CPUs not yet onlined */
120 A perf_header_string with description of the CPU. On x86 this is the model name
125 A perf_header_string with the exact CPU type. On x86 this is
126 vendor,family,model,stepping. For example: GenuineIntel,6,69,1
128 HEADER_TOTAL_MEM = 10,
130 An uint64_t with the total memory in bytes.
134 A perf_header_string with the perf command line used to collect the data.
136 HEADER_EVENT_DESC = 12,
138 Another description of the perf_event_attrs, more detailed than header.attrs
139 including IDs and names. See perf_event.h or the man page for a description
140 of a struct perf_event_attr.
143 uint32_t nr; /* number of events */
144 uint32_t attr_size; /* size of each perf_event_attr */
146 struct perf_event_attr attr; /* size of attr_size */
148 struct perf_header_string event_string;
149 uint64_t ids[nr_ids];
150 } events[nr]; /* Variable length records */
153 HEADER_CPU_TOPOLOGY = 13,
155 String lists defining the core and CPU threads topology.
158 struct perf_header_string_list cores; /* Variable length */
159 struct perf_header_string_list threads; /* Variable length */
164 sibling threads : 0-1
165 sibling threads : 2-3
167 HEADER_NUMA_TOPOLOGY = 14,
169 A list of NUMA node descriptions
177 struct perf_header_string cpus;
178 } nodes[nr]; /* Variable length records */
181 HEADER_BRANCH_STACK = 15,
183 Not implemented in perf.
185 HEADER_PMU_MAPPINGS = 16,
187 A list of PMU structures, defining the different PMUs supported by perf.
193 struct perf_header_string pmu_name;
194 } [nr]; /* Variable length records */
197 HEADER_GROUP_DESC = 17,
199 Description of counter groups ({...} in perf syntax)
204 struct perf_header_string string;
207 } [nr]; /* Variable length records */
210 HEADER_AUXTRACE = 18,
212 Define additional auxtrace areas in the perf.data. auxtrace is used to store
213 undecoded hardware tracing information, such as Intel Processor Trace data.
216 * struct auxtrace_index_entry - indexes a AUX area tracing event within a
218 * @file_offset: offset within the perf.data file
219 * @sz: size of the event
221 struct auxtrace_index_entry {
226 #define PERF_AUXTRACE_INDEX_ENTRY_COUNT 256
229 * struct auxtrace_index - index of AUX area tracing events within a perf.data
231 * @list: linking a number of arrays of entries
232 * @nr: number of entries
233 * @entries: array of entries
235 struct auxtrace_index {
236 struct list_head list;
238 struct auxtrace_index_entry entries[PERF_AUXTRACE_INDEX_ENTRY_COUNT];
243 This is merely a flag signifying that the data section contains data
244 recorded from perf stat record.
248 Description of the cache hierarchy. Based on the Linux sysfs format
249 in /sys/devices/system/cpu/cpu*/cache/
251 u32 version Currently always 1
252 u32 number_of_cache_levels
259 struct perf_header_string type;
260 struct perf_header_string size;
261 struct perf_header_string map;
262 }[number_of_cache_levels];
264 HEADER_SAMPLE_TIME = 21,
266 Two uint64_t for the time of first sample and the time of last sample.
268 other bits are reserved and should ignored for now
269 HEADER_FEAT_BITS = 256,
273 This is an array of perf_event_attrs, each attr_size bytes long, which defines
274 each event collected. See perf_event.h or the man page for a detailed
279 This section is the bulk of the file. It consist of a stream of perf_events
280 describing events. This matches the format generated by the kernel.
281 See perf_event.h or the manpage for a detailed description.
283 Some notes on parsing:
287 The events are not necessarily in time stamp order, as they can be
288 collected in parallel on different CPUs. If the events should be
289 processed in time order they need to be sorted first. It is possible
290 to only do a partial sort using the FINISHED_ROUND event header (see
291 below). perf record guarantees that there is no reordering over a
296 When the event stream contains multiple events each event is identified
297 by an ID. This can be either through the PERF_SAMPLE_ID or the
298 PERF_SAMPLE_IDENTIFIER header. The PERF_SAMPLE_IDENTIFIER header is
299 at a fixed offset from the event header, which allows reliable
300 parsing of the header. Relying on ID may be ambiguous.
301 IDENTIFIER is only supported by newer Linux kernels.
303 Perf record specific events:
305 In addition to the kernel generated event types perf record adds its
306 own event types (in addition it also synthesizes some kernel events,
307 for example MMAP events)
309 PERF_RECORD_USER_TYPE_START = 64,
310 PERF_RECORD_HEADER_ATTR = 64,
313 struct perf_event_header header;
314 struct perf_event_attr attr;
318 PERF_RECORD_HEADER_EVENT_TYPE = 65, /* deprecated */
320 #define MAX_EVENT_NAME 64
322 struct perf_trace_event_type {
324 char name[MAX_EVENT_NAME];
327 struct event_type_event {
328 struct perf_event_header header;
329 struct perf_trace_event_type event_type;
333 PERF_RECORD_HEADER_TRACING_DATA = 66,
337 struct tracing_data_event {
338 struct perf_event_header header;
342 PERF_RECORD_HEADER_BUILD_ID = 67,
344 Define a ELF build ID for a referenced executable.
346 struct build_id_event; /* See above */
348 PERF_RECORD_FINISHED_ROUND = 68,
350 No event reordering over this header. No payload.
352 PERF_RECORD_ID_INDEX = 69,
354 Map event ids to CPUs and TIDs.
356 struct id_index_entry {
363 struct id_index_event {
364 struct perf_event_header header;
366 struct id_index_entry entries[nr];
369 PERF_RECORD_AUXTRACE_INFO = 70,
371 Auxtrace type specific information. Describe me
373 struct auxtrace_info_event {
374 struct perf_event_header header;
376 uint32_t reserved__; /* For alignment */
380 PERF_RECORD_AUXTRACE = 71,
382 Defines auxtrace data. Followed by the actual data. The contents of
383 the auxtrace data is dependent on the event and the CPU. For example
384 for Intel Processor Trace it contains Processor Trace data generated
387 struct auxtrace_event {
388 struct perf_event_header header;
395 uint32_t reserved__; /* For alignment */
399 struct perf_event_header header;
405 PERF_RECORD_AUXTRACE_ERROR = 72,
407 Describes an error in hardware tracing
409 enum auxtrace_error_type {
410 PERF_AUXTRACE_ERROR_ITRACE = 1,
411 PERF_AUXTRACE_ERROR_MAX
414 #define MAX_AUXTRACE_ERROR_MSG 64
416 struct auxtrace_error_event {
417 struct perf_event_header header;
423 uint32_t reserved__; /* For alignment */
425 char msg[MAX_AUXTRACE_ERROR_MSG];
428 PERF_RECORD_HEADER_FEATURE = 80,
430 Describes a header feature. These are records used in pipe-mode that
431 contain information that otherwise would be in perf.data file's header.
435 Define the event attributes with their IDs.
437 An array bound by the perf_file_section size.
440 struct perf_event_attr attr; /* Size defined by header.attr_size */
441 struct perf_file_section ids;
444 ids points to a array of uint64_t defining the ids for event attr attr.
448 Pipe-mode avoid seeks in the file by removing the perf_file_section and flags
449 from the struct perf_header. The trimmed header is:
451 struct perf_pipe_file_header {
456 The information about attrs, data, and event_types is instead in the
457 synthesized events PERF_RECORD_ATTR, PERF_RECORD_HEADER_TRACING_DATA,
458 PERF_RECORD_HEADER_EVENT_TYPE, and PERF_RECORD_HEADER_FEATURE
459 that are generated by perf record in pipe-mode.
464 include/uapi/linux/perf_event.h
466 This is the canonical description of the kernel generated perf_events
467 and the perf_event_attrs.
471 A manpage describing perf_event and perf_event_attr is here:
472 http://web.eece.maine.edu/~vweaver/projects/perf_events/programming.html
473 This tends to be slightly behind the kernel include, but has better
474 descriptions. An (typically older) version of the man page may be
475 included with the standard Linux man pages, available with "man
480 https://github.com/andikleen/pmu-tools/tree/master/parser
482 A definition of the perf.data format in python "construct" format is available
483 in pmu-tools parser. This allows to read perf.data from python and dump it.
487 The quipper C++ parser is available at
488 https://chromium.googlesource.com/chromiumos/platform2
490 It is under the chromiumos-wide-profiling/ subdirectory. This library can
491 convert a perf data file to a protobuf and vice versa.
493 Unfortunately this parser tends to be many versions behind and may not be able
494 to parse data files generated by recent perf.