xref: /linux/Documentation/userspace-api/perf_ring_buffer.rst (revision 79790b6818e96c58fe2bffee1b418c16e64e7b80)
1.. SPDX-License-Identifier: GPL-2.0
2
3================
4Perf ring buffer
5================
6
7.. CONTENTS
8
9    1. Introduction
10
11    2. Ring buffer implementation
12    2.1  Basic algorithm
13    2.2  Ring buffer for different tracing modes
14    2.2.1       Default mode
15    2.2.2       Per-thread mode
16    2.2.3       Per-CPU mode
17    2.2.4       System wide mode
18    2.3  Accessing buffer
19    2.3.1       Producer-consumer model
20    2.3.2       Properties of the ring buffers
21    2.3.3       Writing samples into buffer
22    2.3.4       Reading samples from buffer
23    2.3.5       Memory synchronization
24
25    3. The mechanism of AUX ring buffer
26    3.1  The relationship between AUX and regular ring buffers
27    3.2  AUX events
28    3.3  Snapshot mode
29
30
311. Introduction
32===============
33
34The ring buffer is a fundamental mechanism for data transfer.  perf uses
35ring buffers to transfer event data from kernel to user space, another
36kind of ring buffer which is so called auxiliary (AUX) ring buffer also
37plays an important role for hardware tracing with Intel PT, Arm
38CoreSight, etc.
39
40The ring buffer implementation is critical but it's also a very
41challenging work.  On the one hand, the kernel and perf tool in the user
42space use the ring buffer to exchange data and stores data into data
43file, thus the ring buffer needs to transfer data with high throughput;
44on the other hand, the ring buffer management should avoid significant
45overload to distract profiling results.
46
47This documentation dives into the details for perf ring buffer with two
48parts: firstly it explains the perf ring buffer implementation, then the
49second part discusses the AUX ring buffer mechanism.
50
512. Ring buffer implementation
52=============================
53
542.1 Basic algorithm
55-------------------
56
57That said, a typical ring buffer is managed by a head pointer and a tail
58pointer; the head pointer is manipulated by a writer and the tail
59pointer is updated by a reader respectively.
60
61::
62
63        +---------------------------+
64        |   |   |***|***|***|   |   |
65        +---------------------------+
66                `-> Tail    `-> Head
67
68        * : the data is filled by the writer.
69
70                Figure 1. Ring buffer
71
72Perf uses the same way to manage its ring buffer.  In the implementation
73there are two key data structures held together in a set of consecutive
74pages, the control structure and then the ring buffer itself.  The page
75with the control structure in is known as the "user page".  Being held
76in continuous virtual addresses simplifies locating the ring buffer
77address, it is in the pages after the page with the user page.
78
79The control structure is named as ``perf_event_mmap_page``, it contains a
80head pointer ``data_head`` and a tail pointer ``data_tail``.  When the
81kernel starts to fill records into the ring buffer, it updates the head
82pointer to reserve the memory so later it can safely store events into
83the buffer.  On the other side, when the user page is a writable mapping,
84the perf tool has the permission to update the tail pointer after consuming
85data from the ring buffer.  Yet another case is for the user page's
86read-only mapping, which is to be addressed in the section
87:ref:`writing_samples_into_buffer`.
88
89::
90
91          user page                          ring buffer
92    +---------+---------+   +---------------------------------------+
93    |data_head|data_tail|...|   |   |***|***|***|***|***|   |   |   |
94    +---------+---------+   +---------------------------------------+
95        `          `----------------^                   ^
96         `----------------------------------------------|
97
98              * : the data is filled by the writer.
99
100                Figure 2. Perf ring buffer
101
102When using the ``perf record`` tool, we can specify the ring buffer size
103with option ``-m`` or ``--mmap-pages=``, the given size will be rounded up
104to a power of two that is a multiple of a page size.  Though the kernel
105allocates at once for all memory pages, it's deferred to map the pages
106to VMA area until the perf tool accesses the buffer from the user space.
107In other words, at the first time accesses the buffer's page from user
108space in the perf tool, a data abort exception for page fault is taken
109and the kernel uses this occasion to map the page into process VMA
110(see ``perf_mmap_fault()``), thus the perf tool can continue to access
111the page after returning from the exception.
112
1132.2 Ring buffer for different tracing modes
114-------------------------------------------
115
116The perf profiles programs with different modes: default mode, per thread
117mode, per cpu mode, and system wide mode.  This section describes these
118modes and how the ring buffer meets requirements for them.  At last we
119will review the race conditions caused by these modes.
120
1212.2.1 Default mode
122^^^^^^^^^^^^^^^^^^
123
124Usually we execute ``perf record`` command followed by a profiling program
125name, like below command::
126
127        perf record test_program
128
129This command doesn't specify any options for CPU and thread modes, the
130perf tool applies the default mode on the perf event.  It maps all the
131CPUs in the system and the profiled program's PID on the perf event, and
132it enables inheritance mode on the event so that child tasks inherits
133the events.  As a result, the perf event is attributed as::
134
135    evsel::cpus::map[]    = { 0 .. _SC_NPROCESSORS_ONLN-1 }
136    evsel::threads::map[] = { pid }
137    evsel::attr::inherit  = 1
138
139These attributions finally will be reflected on the deployment of ring
140buffers.  As shown below, the perf tool allocates individual ring buffer
141for each CPU, but it only enables events for the profiled program rather
142than for all threads in the system.  The *T1* thread represents the
143thread context of the 'test_program', whereas *T2* and *T3* are irrelevant
144threads in the system.   The perf samples are exclusively collected for
145the *T1* thread and stored in the ring buffer associated with the CPU on
146which the *T1* thread is running.
147
148::
149
150              T1                      T2                 T1
151            +----+              +-----------+          +----+
152    CPU0    |xxxx|              |xxxxxxxxxxx|          |xxxx|
153            +----+--------------+-----------+----------+----+-------->
154              |                                          |
155              v                                          v
156            +-----------------------------------------------------+
157            |                  Ring buffer 0                      |
158            +-----------------------------------------------------+
159
160                   T1
161                 +-----+
162    CPU1         |xxxxx|
163            -----+-----+--------------------------------------------->
164                    |
165                    v
166            +-----------------------------------------------------+
167            |                  Ring buffer 1                      |
168            +-----------------------------------------------------+
169
170                                        T1              T3
171                                      +----+        +-------+
172    CPU2                              |xxxx|        |xxxxxxx|
173            --------------------------+----+--------+-------+-------->
174                                        |
175                                        v
176            +-----------------------------------------------------+
177            |                  Ring buffer 2                      |
178            +-----------------------------------------------------+
179
180                              T1
181                       +--------------+
182    CPU3               |xxxxxxxxxxxxxx|
183            -----------+--------------+------------------------------>
184                              |
185                              v
186            +-----------------------------------------------------+
187            |                  Ring buffer 3                      |
188            +-----------------------------------------------------+
189
190	    T1: Thread 1; T2: Thread 2; T3: Thread 3
191	    x: Thread is in running state
192
193                Figure 3. Ring buffer for default mode
194
1952.2.2 Per-thread mode
196^^^^^^^^^^^^^^^^^^^^^
197
198By specifying option ``--per-thread`` in perf command, e.g.
199
200::
201
202        perf record --per-thread test_program
203
204The perf event doesn't map to any CPUs and is only bound to the
205profiled process, thus, the perf event's attributions are::
206
207    evsel::cpus::map[0]   = { -1 }
208    evsel::threads::map[] = { pid }
209    evsel::attr::inherit  = 0
210
211In this mode, a single ring buffer is allocated for the profiled thread;
212if the thread is scheduled on a CPU, the events on that CPU will be
213enabled; and if the thread is scheduled out from the CPU, the events on
214the CPU will be disabled.  When the thread is migrated from one CPU to
215another, the events are to be disabled on the previous CPU and enabled
216on the next CPU correspondingly.
217
218::
219
220              T1                      T2                 T1
221            +----+              +-----------+          +----+
222    CPU0    |xxxx|              |xxxxxxxxxxx|          |xxxx|
223            +----+--------------+-----------+----------+----+-------->
224              |                                           |
225              |    T1                                     |
226              |  +-----+                                  |
227    CPU1      |  |xxxxx|                                  |
228            --|--+-----+----------------------------------|---------->
229              |     |                                     |
230              |     |                   T1            T3  |
231              |     |                 +----+        +---+ |
232    CPU2      |     |                 |xxxx|        |xxx| |
233            --|-----|-----------------+----+--------+---+-|---------->
234              |     |                   |                 |
235              |     |         T1        |                 |
236              |     |  +--------------+ |                 |
237    CPU3      |     |  |xxxxxxxxxxxxxx| |                 |
238            --|-----|--+--------------+-|-----------------|---------->
239              |     |         |         |                 |
240              v     v         v         v                 v
241            +-----------------------------------------------------+
242            |                  Ring buffer                        |
243            +-----------------------------------------------------+
244
245            T1: Thread 1
246            x: Thread is in running state
247
248                Figure 4. Ring buffer for per-thread mode
249
250When perf runs in per-thread mode, a ring buffer is allocated for the
251profiled thread *T1*.  The ring buffer is dedicated for thread *T1*, if the
252thread *T1* is running, the perf events will be recorded into the ring
253buffer; when the thread is sleeping, all associated events will be
254disabled, thus no trace data will be recorded into the ring buffer.
255
2562.2.3 Per-CPU mode
257^^^^^^^^^^^^^^^^^^
258
259The option ``-C`` is used to collect samples on the list of CPUs, for
260example the below perf command receives option ``-C 0,2``::
261
262	perf record -C 0,2 test_program
263
264It maps the perf event to CPUs 0 and 2, and the event is not associated to any
265PID.  Thus the perf event attributions are set as::
266
267    evsel::cpus::map[0]   = { 0, 2 }
268    evsel::threads::map[] = { -1 }
269    evsel::attr::inherit  = 0
270
271This results in the session of ``perf record`` will sample all threads on CPU0
272and CPU2, and be terminated until test_program exits.  Even there have tasks
273running on CPU1 and CPU3, since the ring buffer is absent for them, any
274activities on these two CPUs will be ignored.  A usage case is to combine the
275options for per-thread mode and per-CPU mode, e.g. the options ``–C 0,2`` and
276``––per–thread`` are specified together, the samples are recorded only when
277the profiled thread is scheduled on any of the listed CPUs.
278
279::
280
281              T1                      T2                 T1
282            +----+              +-----------+          +----+
283    CPU0    |xxxx|              |xxxxxxxxxxx|          |xxxx|
284            +----+--------------+-----------+----------+----+-------->
285              |                       |                  |
286              v                       v                  v
287            +-----------------------------------------------------+
288            |                  Ring buffer 0                      |
289            +-----------------------------------------------------+
290
291                   T1
292                 +-----+
293    CPU1         |xxxxx|
294            -----+-----+--------------------------------------------->
295
296                                        T1              T3
297                                      +----+        +-------+
298    CPU2                              |xxxx|        |xxxxxxx|
299            --------------------------+----+--------+-------+-------->
300                                        |               |
301                                        v               v
302            +-----------------------------------------------------+
303            |                  Ring buffer 1                      |
304            +-----------------------------------------------------+
305
306                              T1
307                       +--------------+
308    CPU3               |xxxxxxxxxxxxxx|
309            -----------+--------------+------------------------------>
310
311            T1: Thread 1; T2: Thread 2; T3: Thread 3
312            x: Thread is in running state
313
314                Figure 5. Ring buffer for per-CPU mode
315
3162.2.4 System wide mode
317^^^^^^^^^^^^^^^^^^^^^^
318
319By using option ``–a`` or ``––all–cpus``, perf collects samples on all CPUs
320for all tasks, we call it as the system wide mode, the command is::
321
322        perf record -a test_program
323
324Similar to the per-CPU mode, the perf event doesn't bind to any PID, and
325it maps to all CPUs in the system::
326
327   evsel::cpus::map[]    = { 0 .. _SC_NPROCESSORS_ONLN-1 }
328   evsel::threads::map[] = { -1 }
329   evsel::attr::inherit  = 0
330
331In the system wide mode, every CPU has its own ring buffer, all threads
332are monitored during the running state and the samples are recorded into
333the ring buffer belonging to the CPU which the events occurred on.
334
335::
336
337              T1                      T2                 T1
338            +----+              +-----------+          +----+
339    CPU0    |xxxx|              |xxxxxxxxxxx|          |xxxx|
340            +----+--------------+-----------+----------+----+-------->
341              |                       |                  |
342              v                       v                  v
343            +-----------------------------------------------------+
344            |                  Ring buffer 0                      |
345            +-----------------------------------------------------+
346
347                   T1
348                 +-----+
349    CPU1         |xxxxx|
350            -----+-----+--------------------------------------------->
351                    |
352                    v
353            +-----------------------------------------------------+
354            |                  Ring buffer 1                      |
355            +-----------------------------------------------------+
356
357                                        T1              T3
358                                      +----+        +-------+
359    CPU2                              |xxxx|        |xxxxxxx|
360            --------------------------+----+--------+-------+-------->
361                                        |               |
362                                        v               v
363            +-----------------------------------------------------+
364            |                  Ring buffer 2                      |
365            +-----------------------------------------------------+
366
367                              T1
368                       +--------------+
369    CPU3               |xxxxxxxxxxxxxx|
370            -----------+--------------+------------------------------>
371                              |
372                              v
373            +-----------------------------------------------------+
374            |                  Ring buffer 3                      |
375            +-----------------------------------------------------+
376
377            T1: Thread 1; T2: Thread 2; T3: Thread 3
378            x: Thread is in running state
379
380                Figure 6. Ring buffer for system wide mode
381
3822.3 Accessing buffer
383--------------------
384
385Based on the understanding of how the ring buffer is allocated in
386various modes, this section explains access the ring buffer.
387
3882.3.1 Producer-consumer model
389^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
390
391In the Linux kernel, the PMU events can produce samples which are stored
392into the ring buffer; the perf command in user space consumes the
393samples by reading out data from the ring buffer and finally saves the
394data into the file for post analysis.  It’s a typical producer-consumer
395model for using the ring buffer.
396
397The perf process polls on the PMU events and sleeps when no events are
398incoming.  To prevent frequent exchanges between the kernel and user
399space, the kernel event core layer introduces a watermark, which is
400stored in the ``perf_buffer::watermark``.  When a sample is recorded into
401the ring buffer, and if the used buffer exceeds the watermark, the
402kernel wakes up the perf process to read samples from the ring buffer.
403
404::
405
406                       Perf
407                       / | Read samples
408             Polling  /  `--------------|               Ring buffer
409                     v                  v    ;---------------------v
410    +----------------+     +---------+---------+   +-------------------+
411    |Event wait queue|     |data_head|data_tail|   |***|***|   |   |***|
412    +----------------+     +---------+---------+   +-------------------+
413             ^                  ^ `------------------------^
414             | Wake up tasks    | Store samples
415          +-----------------------------+
416          |  Kernel event core layer    |
417          +-----------------------------+
418
419              * : the data is filled by the writer.
420
421                Figure 7. Writing and reading the ring buffer
422
423When the kernel event core layer notifies the user space, because
424multiple events might share the same ring buffer for recording samples,
425the core layer iterates every event associated with the ring buffer and
426wakes up tasks waiting on the event.  This is fulfilled by the kernel
427function ``ring_buffer_wakeup()``.
428
429After the perf process is woken up, it starts to check the ring buffers
430one by one, if it finds any ring buffer containing samples it will read
431out the samples for statistics or saving into the data file.  Given the
432perf process is able to run on any CPU, this leads to the ring buffer
433potentially being accessed from multiple CPUs simultaneously, which
434causes race conditions.  The race condition handling is described in the
435section :ref:`memory_synchronization`.
436
4372.3.2 Properties of the ring buffers
438^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
439
440Linux kernel supports two write directions for the ring buffer: forward and
441backward.  The forward writing saves samples from the beginning of the ring
442buffer, the backward writing stores data from the end of the ring buffer with
443the reversed direction.  The perf tool determines the writing direction.
444
445Additionally, the tool can map buffers in either read-write mode or read-only
446mode to the user space.
447
448The ring buffer in the read-write mode is mapped with the property
449``PROT_READ | PROT_WRITE``.  With the write permission, the perf tool
450updates the ``data_tail`` to indicate the data start position.  Combining
451with the head pointer ``data_head``, which works as the end position of
452the current data, the perf tool can easily know where read out the data
453from.
454
455Alternatively, in the read-only mode, only the kernel keeps to update
456the ``data_head`` while the user space cannot access the ``data_tail`` due
457to the mapping property ``PROT_READ``.
458
459As a result, the matrix below illustrates the various combinations of
460direction and mapping characteristics.  The perf tool employs two of these
461combinations to support buffer types: the non-overwrite buffer and the
462overwritable buffer.
463
464.. list-table::
465   :widths: 1 1 1
466   :header-rows: 1
467
468   * - Mapping mode
469     - Forward
470     - Backward
471   * - read-write
472     - Non-overwrite ring buffer
473     - Not used
474   * - read-only
475     - Not used
476     - Overwritable ring buffer
477
478The non-overwrite ring buffer uses the read-write mapping with forward
479writing.  It starts to save data from the beginning of the ring buffer
480and wrap around when overflow, which is used with the read-write mode in
481the normal ring buffer.  When the consumer doesn't keep up with the
482producer, it would lose some data, the kernel keeps how many records it
483lost and generates the ``PERF_RECORD_LOST`` records in the next time
484when it finds a space in the ring buffer.
485
486The overwritable ring buffer uses the backward writing with the
487read-only mode.  It saves the data from the end of the ring buffer and
488the ``data_head`` keeps the position of current data, the perf always
489knows where it starts to read and until the end of the ring buffer, thus
490it don't need the ``data_tail``.  In this mode, it will not generate the
491``PERF_RECORD_LOST`` records.
492
493.. _writing_samples_into_buffer:
494
4952.3.3 Writing samples into buffer
496^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
497
498When a sample is taken and saved into the ring buffer, the kernel
499prepares sample fields based on the sample type; then it prepares the
500info for writing ring buffer which is stored in the structure
501``perf_output_handle``.  In the end, the kernel outputs the sample into
502the ring buffer and updates the head pointer in the user page so the
503perf tool can see the latest value.
504
505The structure ``perf_output_handle`` serves as a temporary context for
506tracking the information related to the buffer.  The advantages of it is
507that it enables concurrent writing to the buffer by different events.
508For example, a software event and a hardware PMU event both are enabled
509for profiling, two instances of ``perf_output_handle`` serve as separate
510contexts for the software event and the hardware event respectively.
511This allows each event to reserve its own memory space for populating
512the record data.
513
5142.3.4 Reading samples from buffer
515^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
516
517In the user space, the perf tool utilizes the ``perf_event_mmap_page``
518structure to handle the head and tail of the buffer.  It also uses
519``perf_mmap`` structure to keep track of a context for the ring buffer, this
520context includes information about the buffer's starting and ending
521addresses.  Additionally, the mask value can be utilized to compute the
522circular buffer pointer even for an overflow.
523
524Similar to the kernel, the perf tool in the user space first reads out
525the recorded data from the ring buffer, and then updates the buffer's
526tail pointer ``perf_event_mmap_page::data_tail``.
527
528.. _memory_synchronization:
529
5302.3.5 Memory synchronization
531^^^^^^^^^^^^^^^^^^^^^^^^^^^^
532
533The modern CPUs with relaxed memory model cannot promise the memory
534ordering, this means it’s possible to access the ring buffer and the
535``perf_event_mmap_page`` structure out of order.  To assure the specific
536sequence for memory accessing perf ring buffer, memory barriers are
537used to assure the data dependency.  The rationale for the memory
538synchronization is as below::
539
540  Kernel                          User space
541
542  if (LOAD ->data_tail) {         LOAD ->data_head
543                   (A)            smp_rmb()        (C)
544    STORE $data                   LOAD $data
545    smp_wmb()      (B)            smp_mb()         (D)
546    STORE ->data_head             STORE ->data_tail
547  }
548
549The comments in tools/include/linux/ring_buffer.h gives nice description
550for why and how to use memory barriers, here we will just provide an
551alternative explanation:
552
553(A) is a control dependency so that CPU assures order between checking
554pointer ``perf_event_mmap_page::data_tail`` and filling sample into ring
555buffer;
556
557(D) pairs with (A).  (D) separates the ring buffer data reading from
558writing the pointer ``data_tail``, perf tool first consumes samples and then
559tells the kernel that the data chunk has been released.  Since a reading
560operation is followed by a writing operation, thus (D) is a full memory
561barrier.
562
563(B) is a writing barrier in the middle of two writing operations, which
564makes sure that recording a sample must be prior to updating the head
565pointer.
566
567(C) pairs with (B).  (C) is a read memory barrier to ensure the head
568pointer is fetched before reading samples.
569
570To implement the above algorithm, the ``perf_output_put_handle()`` function
571in the kernel and two helpers ``ring_buffer_read_head()`` and
572``ring_buffer_write_tail()`` in the user space are introduced, they rely
573on memory barriers as described above to ensure the data dependency.
574
575Some architectures support one-way permeable barrier with load-acquire
576and store-release operations, these barriers are more relaxed with less
577performance penalty, so (C) and (D) can be optimized to use barriers
578``smp_load_acquire()`` and ``smp_store_release()`` respectively.
579
580If an architecture doesn’t support load-acquire and store-release in its
581memory model, it will roll back to the old fashion of memory barrier
582operations.  In this case, ``smp_load_acquire()`` encapsulates
583``READ_ONCE()`` + ``smp_mb()``, since ``smp_mb()`` is costly,
584``ring_buffer_read_head()`` doesn't invoke ``smp_load_acquire()`` and it uses
585the barriers ``READ_ONCE()`` + ``smp_rmb()`` instead.
586
5873. The mechanism of AUX ring buffer
588===================================
589
590In this chapter, we will explain the implementation of the AUX ring
591buffer.  In the first part it will discuss the connection between the
592AUX ring buffer and the regular ring buffer, then the second part will
593examine how the AUX ring buffer co-works with the regular ring buffer,
594as well as the additional features introduced by the AUX ring buffer for
595the sampling mechanism.
596
5973.1 The relationship between AUX and regular ring buffers
598---------------------------------------------------------
599
600Generally, the AUX ring buffer is an auxiliary for the regular ring
601buffer.  The regular ring buffer is primarily used to store the event
602samples and every event format complies with the definition in the
603union ``perf_event``; the AUX ring buffer is for recording the hardware
604trace data and the trace data format is hardware IP dependent.
605
606The general use and advantage of the AUX ring buffer is that it is
607written directly by hardware rather than by the kernel.  For example,
608regular profile samples that write to the regular ring buffer cause an
609interrupt.  Tracing execution requires a high number of samples and
610using interrupts would be overwhelming for the regular ring buffer
611mechanism.  Having an AUX buffer allows for a region of memory more
612decoupled from the kernel and written to directly by hardware tracing.
613
614The AUX ring buffer reuses the same algorithm with the regular ring
615buffer for the buffer management.  The control structure
616``perf_event_mmap_page`` extends the new fields ``aux_head`` and ``aux_tail``
617for the head and tail pointers of the AUX ring buffer.
618
619During the initialisation phase, besides the mmap()-ed regular ring
620buffer, the perf tool invokes a second syscall in the
621``auxtrace_mmap__mmap()`` function for the mmap of the AUX buffer with
622non-zero file offset; ``rb_alloc_aux()`` in the kernel allocates pages
623correspondingly, these pages will be deferred to map into VMA when
624handling the page fault, which is the same lazy mechanism with the
625regular ring buffer.
626
627AUX events and AUX trace data are two different things.  Let's see an
628example::
629
630        perf record -a -e cycles -e cs_etm/@tmc_etr0/ -- sleep 2
631
632The above command enables two events: one is the event *cycles* from PMU
633and another is the AUX event *cs_etm* from Arm CoreSight, both are saved
634into the regular ring buffer while the CoreSight's AUX trace data is
635stored in the AUX ring buffer.
636
637As a result, we can see the regular ring buffer and the AUX ring buffer
638are allocated in pairs.  The perf in default mode allocates the regular
639ring buffer and the AUX ring buffer per CPU-wise, which is the same as
640the system wide mode, however, the default mode records samples only for
641the profiled program, whereas the latter mode profiles for all programs
642in the system.  For per-thread mode, the perf tool allocates only one
643regular ring buffer and one AUX ring buffer for the whole session.  For
644the per-CPU mode, the perf allocates two kinds of ring buffers for
645selected CPUs specified by the option ``-C``.
646
647The below figure demonstrates the buffers' layout in the system wide
648mode; if there are any activities on one CPU, the AUX event samples and
649the hardware trace data will be recorded into the dedicated buffers for
650the CPU.
651
652::
653
654              T1                      T2                 T1
655            +----+              +-----------+          +----+
656    CPU0    |xxxx|              |xxxxxxxxxxx|          |xxxx|
657            +----+--------------+-----------+----------+----+-------->
658              |                       |                  |
659              v                       v                  v
660            +-----------------------------------------------------+
661            |                  Ring buffer 0                      |
662            +-----------------------------------------------------+
663              |                       |                  |
664              v                       v                  v
665            +-----------------------------------------------------+
666            |               AUX Ring buffer 0                     |
667            +-----------------------------------------------------+
668
669                   T1
670                 +-----+
671    CPU1         |xxxxx|
672            -----+-----+--------------------------------------------->
673                    |
674                    v
675            +-----------------------------------------------------+
676            |                  Ring buffer 1                      |
677            +-----------------------------------------------------+
678                    |
679                    v
680            +-----------------------------------------------------+
681            |               AUX Ring buffer 1                     |
682            +-----------------------------------------------------+
683
684                                        T1              T3
685                                      +----+        +-------+
686    CPU2                              |xxxx|        |xxxxxxx|
687            --------------------------+----+--------+-------+-------->
688                                        |               |
689                                        v               v
690            +-----------------------------------------------------+
691            |                  Ring buffer 2                      |
692            +-----------------------------------------------------+
693                                        |               |
694                                        v               v
695            +-----------------------------------------------------+
696            |               AUX Ring buffer 2                     |
697            +-----------------------------------------------------+
698
699                              T1
700                       +--------------+
701    CPU3               |xxxxxxxxxxxxxx|
702            -----------+--------------+------------------------------>
703                              |
704                              v
705            +-----------------------------------------------------+
706            |                  Ring buffer 3                      |
707            +-----------------------------------------------------+
708                              |
709                              v
710            +-----------------------------------------------------+
711            |               AUX Ring buffer 3                     |
712            +-----------------------------------------------------+
713
714            T1: Thread 1; T2: Thread 2; T3: Thread 3
715            x: Thread is in running state
716
717                Figure 8. AUX ring buffer for system wide mode
718
7193.2 AUX events
720--------------
721
722Similar to ``perf_output_begin()`` and ``perf_output_end()``'s working for the
723regular ring buffer, ``perf_aux_output_begin()`` and ``perf_aux_output_end()``
724serve for the AUX ring buffer for processing the hardware trace data.
725
726Once the hardware trace data is stored into the AUX ring buffer, the PMU
727driver will stop hardware tracing by calling the ``pmu::stop()`` callback.
728Similar to the regular ring buffer, the AUX ring buffer needs to apply
729the memory synchronization mechanism as discussed in the section
730:ref:`memory_synchronization`.  Since the AUX ring buffer is managed by the
731PMU driver, the barrier (B), which is a writing barrier to ensure the trace
732data is externally visible prior to updating the head pointer, is asked
733to be implemented in the PMU driver.
734
735Then ``pmu::stop()`` can safely call the ``perf_aux_output_end()`` function to
736finish two things:
737
738- It fills an event ``PERF_RECORD_AUX`` into the regular ring buffer, this
739  event delivers the information of the start address and data size for a
740  chunk of hardware trace data has been stored into the AUX ring buffer;
741
742- Since the hardware trace driver has stored new trace data into the AUX
743  ring buffer, the argument *size* indicates how many bytes have been
744  consumed by the hardware tracing, thus ``perf_aux_output_end()`` updates the
745  header pointer ``perf_buffer::aux_head`` to reflect the latest buffer usage.
746
747At the end, the PMU driver will restart hardware tracing.  During this
748temporary suspending period, it will lose hardware trace data, which
749will introduce a discontinuity during decoding phase.
750
751The event ``PERF_RECORD_AUX`` presents an AUX event which is handled in the
752kernel, but it lacks the information for saving the AUX trace data in
753the perf file.  When the perf tool copies the trace data from AUX ring
754buffer to the perf data file, it synthesizes a ``PERF_RECORD_AUXTRACE``
755event which is not a kernel ABI, it's defined by the perf tool to describe
756which portion of data in the AUX ring buffer is saved.  Afterwards, the perf
757tool reads out the AUX trace data from the perf file based on the
758``PERF_RECORD_AUXTRACE`` events, and the ``PERF_RECORD_AUX`` event is used to
759decode a chunk of data by correlating with time order.
760
7613.3 Snapshot mode
762-----------------
763
764Perf supports snapshot mode for AUX ring buffer, in this mode, users
765only record AUX trace data at a specific time point which users are
766interested in.  E.g. below gives an example of how to take snapshots
767with 1 second interval with Arm CoreSight::
768
769  perf record -e cs_etm/@tmc_etr0/u -S -a program &
770  PERFPID=$!
771  while true; do
772      kill -USR2 $PERFPID
773      sleep 1
774  done
775
776The main flow for snapshot mode is:
777
778- Before a snapshot is taken, the AUX ring buffer acts in free run mode.
779  During free run mode the perf doesn't record any of the AUX events and
780  trace data;
781
782- Once the perf tool receives the *USR2* signal, it triggers the callback
783  function ``auxtrace_record::snapshot_start()`` to deactivate hardware
784  tracing.  The kernel driver then populates the AUX ring buffer with the
785  hardware trace data, and the event ``PERF_RECORD_AUX`` is stored in the
786  regular ring buffer;
787
788- Then perf tool takes a snapshot, ``record__read_auxtrace_snapshot()``
789  reads out the hardware trace data from the AUX ring buffer and saves it
790  into perf data file;
791
792- After the snapshot is finished, ``auxtrace_record::snapshot_finish()``
793  restarts the PMU event for AUX tracing.
794
795The perf only accesses the head pointer ``perf_event_mmap_page::aux_head``
796in snapshot mode and doesn’t touch tail pointer ``aux_tail``, this is
797because the AUX ring buffer can overflow in free run mode, the tail
798pointer is useless in this case.  Alternatively, the callback
799``auxtrace_record::find_snapshot()`` is introduced for making the decision
800of whether the AUX ring buffer has been wrapped around or not, at the
801end it fixes up the AUX buffer's head which are used to calculate the
802trace data size.
803
804As we know, the buffers' deployment can be per-thread mode, per-CPU
805mode, or system wide mode, and the snapshot can be applied to any of
806these modes.  Below is an example of taking snapshot with system wide
807mode.
808
809::
810
811                                         Snapshot is taken
812                                                 |
813                                                 v
814                        +------------------------+
815                        |  AUX Ring buffer 0     | <- aux_head
816                        +------------------------+
817                                                 v
818                +--------------------------------+
819                |          AUX Ring buffer 1     | <- aux_head
820                +--------------------------------+
821                                                 v
822    +--------------------------------------------+
823    |                      AUX Ring buffer 2     | <- aux_head
824    +--------------------------------------------+
825                                                 v
826         +---------------------------------------+
827         |                 AUX Ring buffer 3     | <- aux_head
828         +---------------------------------------+
829
830                Figure 9. Snapshot with system wide mode
831