1======================== 2ftrace - Function Tracer 3======================== 4 5Copyright 2008 Red Hat Inc. 6 7:Author: Steven Rostedt <srostedt@redhat.com> 8:License: The GNU Free Documentation License, Version 1.2 9 (dual licensed under the GPL v2) 10:Original Reviewers: Elias Oltmanns, Randy Dunlap, Andrew Morton, 11 John Kacur, and David Teigland. 12 13- Written for: 2.6.28-rc2 14- Updated for: 3.10 15- Updated for: 4.13 - Copyright 2017 VMware Inc. Steven Rostedt 16- Converted to rst format - Changbin Du <changbin.du@intel.com> 17 18Introduction 19------------ 20 21Ftrace is an internal tracer designed to help out developers and 22designers of systems to find what is going on inside the kernel. 23It can be used for debugging or analyzing latencies and 24performance issues that take place outside of user-space. 25 26Although ftrace is typically considered the function tracer, it 27is really a framework of several assorted tracing utilities. 28There's latency tracing to examine what occurs between interrupts 29disabled and enabled, as well as for preemption and from a time 30a task is woken to the task is actually scheduled in. 31 32One of the most common uses of ftrace is the event tracing. 33Throughout the kernel is hundreds of static event points that 34can be enabled via the tracefs file system to see what is 35going on in certain parts of the kernel. 36 37See events.rst for more information. 38 39 40Implementation Details 41---------------------- 42 43See Documentation/trace/ftrace-design.rst for details for arch porters and such. 44 45 46The File System 47--------------- 48 49Ftrace uses the tracefs file system to hold the control files as 50well as the files to display output. 51 52When tracefs is configured into the kernel (which selecting any ftrace 53option will do) the directory /sys/kernel/tracing will be created. To mount 54this directory, you can add to your /etc/fstab file:: 55 56 tracefs /sys/kernel/tracing tracefs defaults 0 0 57 58Or you can mount it at run time with:: 59 60 mount -t tracefs nodev /sys/kernel/tracing 61 62For quicker access to that directory you may want to make a soft link to 63it:: 64 65 ln -s /sys/kernel/tracing /tracing 66 67.. attention:: 68 69 Before 4.1, all ftrace tracing control files were within the debugfs 70 file system, which is typically located at /sys/kernel/debug/tracing. 71 For backward compatibility, when mounting the debugfs file system, 72 the tracefs file system will be automatically mounted at: 73 74 /sys/kernel/debug/tracing 75 76 All files located in the tracefs file system will be located in that 77 debugfs file system directory as well. 78 79.. attention:: 80 81 Any selected ftrace option will also create the tracefs file system. 82 The rest of the document will assume that you are in the ftrace directory 83 (cd /sys/kernel/tracing) and will only concentrate on the files within that 84 directory and not distract from the content with the extended 85 "/sys/kernel/tracing" path name. 86 87That's it! (assuming that you have ftrace configured into your kernel) 88 89After mounting tracefs you will have access to the control and output files 90of ftrace. Here is a list of some of the key files: 91 92 93 Note: all time values are in microseconds. 94 95 current_tracer: 96 97 This is used to set or display the current tracer 98 that is configured. Changing the current tracer clears 99 the ring buffer content as well as the "snapshot" buffer. 100 101 available_tracers: 102 103 This holds the different types of tracers that 104 have been compiled into the kernel. The 105 tracers listed here can be configured by 106 echoing their name into current_tracer. 107 108 tracing_on: 109 110 This sets or displays whether writing to the trace 111 ring buffer is enabled. Echo 0 into this file to disable 112 the tracer or 1 to enable it. Note, this only disables 113 writing to the ring buffer, the tracing overhead may 114 still be occurring. 115 116 The kernel function tracing_off() can be used within the 117 kernel to disable writing to the ring buffer, which will 118 set this file to "0". User space can re-enable tracing by 119 echoing "1" into the file. 120 121 Note, the function and event trigger "traceoff" will also 122 set this file to zero and stop tracing. Which can also 123 be re-enabled by user space using this file. 124 125 trace: 126 127 This file holds the output of the trace in a human 128 readable format (described below). Opening this file for 129 writing with the O_TRUNC flag clears the ring buffer content. 130 Note, this file is not a consumer. If tracing is off 131 (no tracer running, or tracing_on is zero), it will produce 132 the same output each time it is read. When tracing is on, 133 it may produce inconsistent results as it tries to read 134 the entire buffer without consuming it. 135 136 trace_pipe: 137 138 The output is the same as the "trace" file but this 139 file is meant to be streamed with live tracing. 140 Reads from this file will block until new data is 141 retrieved. Unlike the "trace" file, this file is a 142 consumer. This means reading from this file causes 143 sequential reads to display more current data. Once 144 data is read from this file, it is consumed, and 145 will not be read again with a sequential read. The 146 "trace" file is static, and if the tracer is not 147 adding more data, it will display the same 148 information every time it is read. 149 150 trace_options: 151 152 This file lets the user control the amount of data 153 that is displayed in one of the above output 154 files. Options also exist to modify how a tracer 155 or events work (stack traces, timestamps, etc). 156 157 options: 158 159 This is a directory that has a file for every available 160 trace option (also in trace_options). Options may also be set 161 or cleared by writing a "1" or "0" respectively into the 162 corresponding file with the option name. 163 164 tracing_max_latency: 165 166 Some of the tracers record the max latency. 167 For example, the maximum time that interrupts are disabled. 168 The maximum time is saved in this file. The max trace will also be 169 stored, and displayed by "trace". A new max trace will only be 170 recorded if the latency is greater than the value in this file 171 (in microseconds). 172 173 By echoing in a time into this file, no latency will be recorded 174 unless it is greater than the time in this file. 175 176 tracing_thresh: 177 178 Some latency tracers will record a trace whenever the 179 latency is greater than the number in this file. 180 Only active when the file contains a number greater than 0. 181 (in microseconds) 182 183 buffer_percent: 184 185 This is the watermark for how much the ring buffer needs to be filled 186 before a waiter is woken up. That is, if an application calls a 187 blocking read syscall on one of the per_cpu trace_pipe_raw files, it 188 will block until the given amount of data specified by buffer_percent 189 is in the ring buffer before it wakes the reader up. This also 190 controls how the splice system calls are blocked on this file:: 191 192 0 - means to wake up as soon as there is any data in the ring buffer. 193 50 - means to wake up when roughly half of the ring buffer sub-buffers 194 are full. 195 100 - means to block until the ring buffer is totally full and is 196 about to start overwriting the older data. 197 198 buffer_size_kb: 199 200 This sets or displays the number of kilobytes each CPU 201 buffer holds. By default, the trace buffers are the same size 202 for each CPU. The displayed number is the size of the 203 CPU buffer and not total size of all buffers. The 204 trace buffers are allocated in pages (blocks of memory 205 that the kernel uses for allocation, usually 4 KB in size). 206 A few extra pages may be allocated to accommodate buffer management 207 meta-data. If the last page allocated has room for more bytes 208 than requested, the rest of the page will be used, 209 making the actual allocation bigger than requested or shown. 210 ( Note, the size may not be a multiple of the page size 211 due to buffer management meta-data. ) 212 213 Buffer sizes for individual CPUs may vary 214 (see "per_cpu/cpu0/buffer_size_kb" below), and if they do 215 this file will show "X". 216 217 buffer_total_size_kb: 218 219 This displays the total combined size of all the trace buffers. 220 221 buffer_subbuf_size_kb: 222 223 This sets or displays the sub buffer size. The ring buffer is broken up 224 into several same size "sub buffers". An event can not be bigger than 225 the size of the sub buffer. Normally, the sub buffer is the size of the 226 architecture's page (4K on x86). The sub buffer also contains meta data 227 at the start which also limits the size of an event. That means when 228 the sub buffer is a page size, no event can be larger than the page 229 size minus the sub buffer meta data. 230 231 Note, the buffer_subbuf_size_kb is a way for the user to specify the 232 minimum size of the subbuffer. The kernel may make it bigger due to the 233 implementation details, or simply fail the operation if the kernel can 234 not handle the request. 235 236 Changing the sub buffer size allows for events to be larger than the 237 page size. 238 239 Note: When changing the sub-buffer size, tracing is stopped and any 240 data in the ring buffer and the snapshot buffer will be discarded. 241 242 free_buffer: 243 244 If a process is performing tracing, and the ring buffer should be 245 shrunk "freed" when the process is finished, even if it were to be 246 killed by a signal, this file can be used for that purpose. On close 247 of this file, the ring buffer will be resized to its minimum size. 248 Having a process that is tracing also open this file, when the process 249 exits its file descriptor for this file will be closed, and in doing so, 250 the ring buffer will be "freed". 251 252 It may also stop tracing if disable_on_free option is set. 253 254 tracing_cpumask: 255 256 This is a mask that lets the user only trace on specified CPUs. 257 The format is a hex string representing the CPUs. 258 259 set_ftrace_filter: 260 261 When dynamic ftrace is configured in (see the 262 section below "dynamic ftrace"), the code is dynamically 263 modified (code text rewrite) to disable calling of the 264 function profiler (mcount). This lets tracing be configured 265 in with practically no overhead in performance. This also 266 has a side effect of enabling or disabling specific functions 267 to be traced. Echoing names of functions into this file 268 will limit the trace to only those functions. 269 This influences the tracers "function" and "function_graph" 270 and thus also function profiling (see "function_profile_enabled"). 271 272 The functions listed in "available_filter_functions" are what 273 can be written into this file. 274 275 This interface also allows for commands to be used. See the 276 "Filter commands" section for more details. 277 278 As a speed up, since processing strings can be quite expensive 279 and requires a check of all functions registered to tracing, instead 280 an index can be written into this file. A number (starting with "1") 281 written will instead select the same corresponding at the line position 282 of the "available_filter_functions" file. 283 284 set_ftrace_notrace: 285 286 This has an effect opposite to that of 287 set_ftrace_filter. Any function that is added here will not 288 be traced. If a function exists in both set_ftrace_filter 289 and set_ftrace_notrace, the function will _not_ be traced. 290 291 set_ftrace_pid: 292 293 Have the function tracer only trace the threads whose PID are 294 listed in this file. 295 296 If the "function-fork" option is set, then when a task whose 297 PID is listed in this file forks, the child's PID will 298 automatically be added to this file, and the child will be 299 traced by the function tracer as well. This option will also 300 cause PIDs of tasks that exit to be removed from the file. 301 302 set_ftrace_notrace_pid: 303 304 Have the function tracer ignore threads whose PID are listed in 305 this file. 306 307 If the "function-fork" option is set, then when a task whose 308 PID is listed in this file forks, the child's PID will 309 automatically be added to this file, and the child will not be 310 traced by the function tracer as well. This option will also 311 cause PIDs of tasks that exit to be removed from the file. 312 313 If a PID is in both this file and "set_ftrace_pid", then this 314 file takes precedence, and the thread will not be traced. 315 316 set_event_pid: 317 318 Have the events only trace a task with a PID listed in this file. 319 Note, sched_switch and sched_wake_up will also trace events 320 listed in this file. 321 322 To have the PIDs of children of tasks with their PID in this file 323 added on fork, enable the "event-fork" option. That option will also 324 cause the PIDs of tasks to be removed from this file when the task 325 exits. 326 327 set_event_notrace_pid: 328 329 Have the events not trace a task with a PID listed in this file. 330 Note, sched_switch and sched_wakeup will trace threads not listed 331 in this file, even if a thread's PID is in the file if the 332 sched_switch or sched_wakeup events also trace a thread that should 333 be traced. 334 335 To have the PIDs of children of tasks with their PID in this file 336 added on fork, enable the "event-fork" option. That option will also 337 cause the PIDs of tasks to be removed from this file when the task 338 exits. 339 340 set_graph_function: 341 342 Functions listed in this file will cause the function graph 343 tracer to only trace these functions and the functions that 344 they call. (See the section "dynamic ftrace" for more details). 345 Note, set_ftrace_filter and set_ftrace_notrace still affects 346 what functions are being traced. 347 348 set_graph_notrace: 349 350 Similar to set_graph_function, but will disable function graph 351 tracing when the function is hit until it exits the function. 352 This makes it possible to ignore tracing functions that are called 353 by a specific function. 354 355 available_filter_functions: 356 357 This lists the functions that ftrace has processed and can trace. 358 These are the function names that you can pass to 359 "set_ftrace_filter", "set_ftrace_notrace", 360 "set_graph_function", or "set_graph_notrace". 361 (See the section "dynamic ftrace" below for more details.) 362 363 available_filter_functions_addrs: 364 365 Similar to available_filter_functions, but with address displayed 366 for each function. The displayed address is the patch-site address 367 and can differ from /proc/kallsyms address. 368 369 dyn_ftrace_total_info: 370 371 This file is for debugging purposes. The number of functions that 372 have been converted to nops and are available to be traced. 373 374 enabled_functions: 375 376 This file is more for debugging ftrace, but can also be useful 377 in seeing if any function has a callback attached to it. 378 Not only does the trace infrastructure use ftrace function 379 trace utility, but other subsystems might too. This file 380 displays all functions that have a callback attached to them 381 as well as the number of callbacks that have been attached. 382 Note, a callback may also call multiple functions which will 383 not be listed in this count. 384 385 If the callback registered to be traced by a function with 386 the "save regs" attribute (thus even more overhead), a 'R' 387 will be displayed on the same line as the function that 388 is returning registers. 389 390 If the callback registered to be traced by a function with 391 the "ip modify" attribute (thus the regs->ip can be changed), 392 an 'I' will be displayed on the same line as the function that 393 can be overridden. 394 395 If a non ftrace trampoline is attached (BPF) a 'D' will be displayed. 396 Note, normal ftrace trampolines can also be attached, but only one 397 "direct" trampoline can be attached to a given function at a time. 398 399 Some architectures can not call direct trampolines, but instead have 400 the ftrace ops function located above the function entry point. In 401 such cases an 'O' will be displayed. 402 403 If a function had either the "ip modify" or a "direct" call attached to 404 it in the past, a 'M' will be shown. This flag is never cleared. It is 405 used to know if a function was every modified by the ftrace infrastructure, 406 and can be used for debugging. 407 408 If the architecture supports it, it will also show what callback 409 is being directly called by the function. If the count is greater 410 than 1 it most likely will be ftrace_ops_list_func(). 411 412 If the callback of a function jumps to a trampoline that is 413 specific to the callback and which is not the standard trampoline, 414 its address will be printed as well as the function that the 415 trampoline calls. 416 417 touched_functions: 418 419 This file contains all the functions that ever had a function callback 420 to it via the ftrace infrastructure. It has the same format as 421 enabled_functions but shows all functions that have every been 422 traced. 423 424 To see any function that has every been modified by "ip modify" or a 425 direct trampoline, one can perform the following command: 426 427 grep ' M ' /sys/kernel/tracing/touched_functions 428 429 function_profile_enabled: 430 431 When set it will enable all functions with either the function 432 tracer, or if configured, the function graph tracer. It will 433 keep a histogram of the number of functions that were called 434 and if the function graph tracer was configured, it will also keep 435 track of the time spent in those functions. The histogram 436 content can be displayed in the files: 437 438 trace_stat/function<cpu> ( function0, function1, etc). 439 440 trace_stat: 441 442 A directory that holds different tracing stats. 443 444 kprobe_events: 445 446 Enable dynamic trace points. See kprobetrace.rst. 447 448 kprobe_profile: 449 450 Dynamic trace points stats. See kprobetrace.rst. 451 452 max_graph_depth: 453 454 Used with the function graph tracer. This is the max depth 455 it will trace into a function. Setting this to a value of 456 one will show only the first kernel function that is called 457 from user space. 458 459 printk_formats: 460 461 This is for tools that read the raw format files. If an event in 462 the ring buffer references a string, only a pointer to the string 463 is recorded into the buffer and not the string itself. This prevents 464 tools from knowing what that string was. This file displays the string 465 and address for the string allowing tools to map the pointers to what 466 the strings were. 467 468 saved_cmdlines: 469 470 Only the pid of the task is recorded in a trace event unless 471 the event specifically saves the task comm as well. Ftrace 472 makes a cache of pid mappings to comms to try to display 473 comms for events. If a pid for a comm is not listed, then 474 "<...>" is displayed in the output. 475 476 If the option "record-cmd" is set to "0", then comms of tasks 477 will not be saved during recording. By default, it is enabled. 478 479 saved_cmdlines_size: 480 481 By default, 128 comms are saved (see "saved_cmdlines" above). To 482 increase or decrease the amount of comms that are cached, echo 483 the number of comms to cache into this file. 484 485 saved_tgids: 486 487 If the option "record-tgid" is set, on each scheduling context switch 488 the Task Group ID of a task is saved in a table mapping the PID of 489 the thread to its TGID. By default, the "record-tgid" option is 490 disabled. 491 492 snapshot: 493 494 This displays the "snapshot" buffer and also lets the user 495 take a snapshot of the current running trace. 496 See the "Snapshot" section below for more details. 497 498 stack_max_size: 499 500 When the stack tracer is activated, this will display the 501 maximum stack size it has encountered. 502 See the "Stack Trace" section below. 503 504 stack_trace: 505 506 This displays the stack back trace of the largest stack 507 that was encountered when the stack tracer is activated. 508 See the "Stack Trace" section below. 509 510 stack_trace_filter: 511 512 This is similar to "set_ftrace_filter" but it limits what 513 functions the stack tracer will check. 514 515 trace_clock: 516 517 Whenever an event is recorded into the ring buffer, a 518 "timestamp" is added. This stamp comes from a specified 519 clock. By default, ftrace uses the "local" clock. This 520 clock is very fast and strictly per cpu, but on some 521 systems it may not be monotonic with respect to other 522 CPUs. In other words, the local clocks may not be in sync 523 with local clocks on other CPUs. 524 525 Usual clocks for tracing:: 526 527 # cat trace_clock 528 [local] global counter x86-tsc 529 530 The clock with the square brackets around it is the one in effect. 531 532 local: 533 Default clock, but may not be in sync across CPUs 534 535 global: 536 This clock is in sync with all CPUs but may 537 be a bit slower than the local clock. 538 539 counter: 540 This is not a clock at all, but literally an atomic 541 counter. It counts up one by one, but is in sync 542 with all CPUs. This is useful when you need to 543 know exactly the order events occurred with respect to 544 each other on different CPUs. 545 546 uptime: 547 This uses the jiffies counter and the time stamp 548 is relative to the time since boot up. 549 550 perf: 551 This makes ftrace use the same clock that perf uses. 552 Eventually perf will be able to read ftrace buffers 553 and this will help out in interleaving the data. 554 555 x86-tsc: 556 Architectures may define their own clocks. For 557 example, x86 uses its own TSC cycle clock here. 558 559 ppc-tb: 560 This uses the powerpc timebase register value. 561 This is in sync across CPUs and can also be used 562 to correlate events across hypervisor/guest if 563 tb_offset is known. 564 565 mono: 566 This uses the fast monotonic clock (CLOCK_MONOTONIC) 567 which is monotonic and is subject to NTP rate adjustments. 568 569 mono_raw: 570 This is the raw monotonic clock (CLOCK_MONOTONIC_RAW) 571 which is monotonic but is not subject to any rate adjustments 572 and ticks at the same rate as the hardware clocksource. 573 574 boot: 575 This is the boot clock (CLOCK_BOOTTIME) and is based on the 576 fast monotonic clock, but also accounts for time spent in 577 suspend. Since the clock access is designed for use in 578 tracing in the suspend path, some side effects are possible 579 if clock is accessed after the suspend time is accounted before 580 the fast mono clock is updated. In this case, the clock update 581 appears to happen slightly sooner than it normally would have. 582 Also on 32-bit systems, it's possible that the 64-bit boot offset 583 sees a partial update. These effects are rare and post 584 processing should be able to handle them. See comments in the 585 ktime_get_boot_fast_ns() function for more information. 586 587 tai: 588 This is the tai clock (CLOCK_TAI) and is derived from the wall- 589 clock time. However, this clock does not experience 590 discontinuities and backwards jumps caused by NTP inserting leap 591 seconds. Since the clock access is designed for use in tracing, 592 side effects are possible. The clock access may yield wrong 593 readouts in case the internal TAI offset is updated e.g., caused 594 by setting the system time or using adjtimex() with an offset. 595 These effects are rare and post processing should be able to 596 handle them. See comments in the ktime_get_tai_fast_ns() 597 function for more information. 598 599 To set a clock, simply echo the clock name into this file:: 600 601 # echo global > trace_clock 602 603 Setting a clock clears the ring buffer content as well as the 604 "snapshot" buffer. 605 606 trace_marker: 607 608 This is a very useful file for synchronizing user space 609 with events happening in the kernel. Writing strings into 610 this file will be written into the ftrace buffer. 611 612 It is useful in applications to open this file at the start 613 of the application and just reference the file descriptor 614 for the file:: 615 616 void trace_write(const char *fmt, ...) 617 { 618 va_list ap; 619 char buf[256]; 620 int n; 621 622 if (trace_fd < 0) 623 return; 624 625 va_start(ap, fmt); 626 n = vsnprintf(buf, 256, fmt, ap); 627 va_end(ap); 628 629 write(trace_fd, buf, n); 630 } 631 632 start:: 633 634 trace_fd = open("trace_marker", O_WRONLY); 635 636 Note: Writing into the trace_marker file can also initiate triggers 637 that are written into /sys/kernel/tracing/events/ftrace/print/trigger 638 See "Event triggers" in Documentation/trace/events.rst and an 639 example in Documentation/trace/histogram.rst (Section 3.) 640 641 trace_marker_raw: 642 643 This is similar to trace_marker above, but is meant for binary data 644 to be written to it, where a tool can be used to parse the data 645 from trace_pipe_raw. 646 647 uprobe_events: 648 649 Add dynamic tracepoints in programs. 650 See uprobetracer.rst 651 652 uprobe_profile: 653 654 Uprobe statistics. See uprobetrace.txt 655 656 instances: 657 658 This is a way to make multiple trace buffers where different 659 events can be recorded in different buffers. 660 See "Instances" section below. 661 662 events: 663 664 This is the trace event directory. It holds event tracepoints 665 (also known as static tracepoints) that have been compiled 666 into the kernel. It shows what event tracepoints exist 667 and how they are grouped by system. There are "enable" 668 files at various levels that can enable the tracepoints 669 when a "1" is written to them. 670 671 See events.rst for more information. 672 673 set_event: 674 675 By echoing in the event into this file, will enable that event. 676 677 See events.rst for more information. 678 679 available_events: 680 681 A list of events that can be enabled in tracing. 682 683 See events.rst for more information. 684 685 timestamp_mode: 686 687 Certain tracers may change the timestamp mode used when 688 logging trace events into the event buffer. Events with 689 different modes can coexist within a buffer but the mode in 690 effect when an event is logged determines which timestamp mode 691 is used for that event. The default timestamp mode is 692 'delta'. 693 694 Usual timestamp modes for tracing: 695 696 # cat timestamp_mode 697 [delta] absolute 698 699 The timestamp mode with the square brackets around it is the 700 one in effect. 701 702 delta: Default timestamp mode - timestamp is a delta against 703 a per-buffer timestamp. 704 705 absolute: The timestamp is a full timestamp, not a delta 706 against some other value. As such it takes up more 707 space and is less efficient. 708 709 hwlat_detector: 710 711 Directory for the Hardware Latency Detector. 712 See "Hardware Latency Detector" section below. 713 714 per_cpu: 715 716 This is a directory that contains the trace per_cpu information. 717 718 per_cpu/cpu0/buffer_size_kb: 719 720 The ftrace buffer is defined per_cpu. That is, there's a separate 721 buffer for each CPU to allow writes to be done atomically, 722 and free from cache bouncing. These buffers may have different 723 size buffers. This file is similar to the buffer_size_kb 724 file, but it only displays or sets the buffer size for the 725 specific CPU. (here cpu0). 726 727 per_cpu/cpu0/trace: 728 729 This is similar to the "trace" file, but it will only display 730 the data specific for the CPU. If written to, it only clears 731 the specific CPU buffer. 732 733 per_cpu/cpu0/trace_pipe 734 735 This is similar to the "trace_pipe" file, and is a consuming 736 read, but it will only display (and consume) the data specific 737 for the CPU. 738 739 per_cpu/cpu0/trace_pipe_raw 740 741 For tools that can parse the ftrace ring buffer binary format, 742 the trace_pipe_raw file can be used to extract the data 743 from the ring buffer directly. With the use of the splice() 744 system call, the buffer data can be quickly transferred to 745 a file or to the network where a server is collecting the 746 data. 747 748 Like trace_pipe, this is a consuming reader, where multiple 749 reads will always produce different data. 750 751 per_cpu/cpu0/snapshot: 752 753 This is similar to the main "snapshot" file, but will only 754 snapshot the current CPU (if supported). It only displays 755 the content of the snapshot for a given CPU, and if 756 written to, only clears this CPU buffer. 757 758 per_cpu/cpu0/snapshot_raw: 759 760 Similar to the trace_pipe_raw, but will read the binary format 761 from the snapshot buffer for the given CPU. 762 763 per_cpu/cpu0/stats: 764 765 This displays certain stats about the ring buffer: 766 767 entries: 768 The number of events that are still in the buffer. 769 770 overrun: 771 The number of lost events due to overwriting when 772 the buffer was full. 773 774 commit overrun: 775 Should always be zero. 776 This gets set if so many events happened within a nested 777 event (ring buffer is re-entrant), that it fills the 778 buffer and starts dropping events. 779 780 bytes: 781 Bytes actually read (not overwritten). 782 783 oldest event ts: 784 The oldest timestamp in the buffer 785 786 now ts: 787 The current timestamp 788 789 dropped events: 790 Events lost due to overwrite option being off. 791 792 read events: 793 The number of events read. 794 795The Tracers 796----------- 797 798Here is the list of current tracers that may be configured. 799 800 "function" 801 802 Function call tracer to trace all kernel functions. 803 804 "function_graph" 805 806 Similar to the function tracer except that the 807 function tracer probes the functions on their entry 808 whereas the function graph tracer traces on both entry 809 and exit of the functions. It then provides the ability 810 to draw a graph of function calls similar to C code 811 source. 812 813 "blk" 814 815 The block tracer. The tracer used by the blktrace user 816 application. 817 818 "hwlat" 819 820 The Hardware Latency tracer is used to detect if the hardware 821 produces any latency. See "Hardware Latency Detector" section 822 below. 823 824 "irqsoff" 825 826 Traces the areas that disable interrupts and saves 827 the trace with the longest max latency. 828 See tracing_max_latency. When a new max is recorded, 829 it replaces the old trace. It is best to view this 830 trace with the latency-format option enabled, which 831 happens automatically when the tracer is selected. 832 833 "preemptoff" 834 835 Similar to irqsoff but traces and records the amount of 836 time for which preemption is disabled. 837 838 "preemptirqsoff" 839 840 Similar to irqsoff and preemptoff, but traces and 841 records the largest time for which irqs and/or preemption 842 is disabled. 843 844 "wakeup" 845 846 Traces and records the max latency that it takes for 847 the highest priority task to get scheduled after 848 it has been woken up. 849 Traces all tasks as an average developer would expect. 850 851 "wakeup_rt" 852 853 Traces and records the max latency that it takes for just 854 RT tasks (as the current "wakeup" does). This is useful 855 for those interested in wake up timings of RT tasks. 856 857 "wakeup_dl" 858 859 Traces and records the max latency that it takes for 860 a SCHED_DEADLINE task to be woken (as the "wakeup" and 861 "wakeup_rt" does). 862 863 "mmiotrace" 864 865 A special tracer that is used to trace binary module. 866 It will trace all the calls that a module makes to the 867 hardware. Everything it writes and reads from the I/O 868 as well. 869 870 "branch" 871 872 This tracer can be configured when tracing likely/unlikely 873 calls within the kernel. It will trace when a likely and 874 unlikely branch is hit and if it was correct in its prediction 875 of being correct. 876 877 "nop" 878 879 This is the "trace nothing" tracer. To remove all 880 tracers from tracing simply echo "nop" into 881 current_tracer. 882 883Error conditions 884---------------- 885 886 For most ftrace commands, failure modes are obvious and communicated 887 using standard return codes. 888 889 For other more involved commands, extended error information may be 890 available via the tracing/error_log file. For the commands that 891 support it, reading the tracing/error_log file after an error will 892 display more detailed information about what went wrong, if 893 information is available. The tracing/error_log file is a circular 894 error log displaying a small number (currently, 8) of ftrace errors 895 for the last (8) failed commands. 896 897 The extended error information and usage takes the form shown in 898 this example:: 899 900 # echo xxx > /sys/kernel/tracing/events/sched/sched_wakeup/trigger 901 echo: write error: Invalid argument 902 903 # cat /sys/kernel/tracing/error_log 904 [ 5348.887237] location: error: Couldn't yyy: zzz 905 Command: xxx 906 ^ 907 [ 7517.023364] location: error: Bad rrr: sss 908 Command: ppp qqq 909 ^ 910 911 To clear the error log, echo the empty string into it:: 912 913 # echo > /sys/kernel/tracing/error_log 914 915Examples of using the tracer 916---------------------------- 917 918Here are typical examples of using the tracers when controlling 919them only with the tracefs interface (without using any 920user-land utilities). 921 922Output format: 923-------------- 924 925Here is an example of the output format of the file "trace":: 926 927 # tracer: function 928 # 929 # entries-in-buffer/entries-written: 140080/250280 #P:4 930 # 931 # _-----=> irqs-off 932 # / _----=> need-resched 933 # | / _---=> hardirq/softirq 934 # || / _--=> preempt-depth 935 # ||| / delay 936 # TASK-PID CPU# |||| TIMESTAMP FUNCTION 937 # | | | |||| | | 938 bash-1977 [000] .... 17284.993652: sys_close <-system_call_fastpath 939 bash-1977 [000] .... 17284.993653: __close_fd <-sys_close 940 bash-1977 [000] .... 17284.993653: _raw_spin_lock <-__close_fd 941 sshd-1974 [003] .... 17284.993653: __srcu_read_unlock <-fsnotify 942 bash-1977 [000] .... 17284.993654: add_preempt_count <-_raw_spin_lock 943 bash-1977 [000] ...1 17284.993655: _raw_spin_unlock <-__close_fd 944 bash-1977 [000] ...1 17284.993656: sub_preempt_count <-_raw_spin_unlock 945 bash-1977 [000] .... 17284.993657: filp_close <-__close_fd 946 bash-1977 [000] .... 17284.993657: dnotify_flush <-filp_close 947 sshd-1974 [003] .... 17284.993658: sys_select <-system_call_fastpath 948 .... 949 950A header is printed with the tracer name that is represented by 951the trace. In this case the tracer is "function". Then it shows the 952number of events in the buffer as well as the total number of entries 953that were written. The difference is the number of entries that were 954lost due to the buffer filling up (250280 - 140080 = 110200 events 955lost). 956 957The header explains the content of the events. Task name "bash", the task 958PID "1977", the CPU that it was running on "000", the latency format 959(explained below), the timestamp in <secs>.<usecs> format, the 960function name that was traced "sys_close" and the parent function that 961called this function "system_call_fastpath". The timestamp is the time 962at which the function was entered. 963 964Latency trace format 965-------------------- 966 967When the latency-format option is enabled or when one of the latency 968tracers is set, the trace file gives somewhat more information to see 969why a latency happened. Here is a typical trace:: 970 971 # tracer: irqsoff 972 # 973 # irqsoff latency trace v1.1.5 on 3.8.0-test+ 974 # -------------------------------------------------------------------- 975 # latency: 259 us, #4/4, CPU#2 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4) 976 # ----------------- 977 # | task: ps-6143 (uid:0 nice:0 policy:0 rt_prio:0) 978 # ----------------- 979 # => started at: __lock_task_sighand 980 # => ended at: _raw_spin_unlock_irqrestore 981 # 982 # 983 # _------=> CPU# 984 # / _-----=> irqs-off 985 # | / _----=> need-resched 986 # || / _---=> hardirq/softirq 987 # ||| / _--=> preempt-depth 988 # |||| / delay 989 # cmd pid ||||| time | caller 990 # \ / ||||| \ | / 991 ps-6143 2d... 0us!: trace_hardirqs_off <-__lock_task_sighand 992 ps-6143 2d..1 259us+: trace_hardirqs_on <-_raw_spin_unlock_irqrestore 993 ps-6143 2d..1 263us+: time_hardirqs_on <-_raw_spin_unlock_irqrestore 994 ps-6143 2d..1 306us : <stack trace> 995 => trace_hardirqs_on_caller 996 => trace_hardirqs_on 997 => _raw_spin_unlock_irqrestore 998 => do_task_stat 999 => proc_tgid_stat 1000 => proc_single_show 1001 => seq_read 1002 => vfs_read 1003 => sys_read 1004 => system_call_fastpath 1005 1006 1007This shows that the current tracer is "irqsoff" tracing the time 1008for which interrupts were disabled. It gives the trace version (which 1009never changes) and the version of the kernel upon which this was executed on 1010(3.8). Then it displays the max latency in microseconds (259 us). The number 1011of trace entries displayed and the total number (both are four: #4/4). 1012VP, KP, SP, and HP are always zero and are reserved for later use. 1013#P is the number of online CPUs (#P:4). 1014 1015The task is the process that was running when the latency 1016occurred. (ps pid: 6143). 1017 1018The start and stop (the functions in which the interrupts were 1019disabled and enabled respectively) that caused the latencies: 1020 1021 - __lock_task_sighand is where the interrupts were disabled. 1022 - _raw_spin_unlock_irqrestore is where they were enabled again. 1023 1024The next lines after the header are the trace itself. The header 1025explains which is which. 1026 1027 cmd: The name of the process in the trace. 1028 1029 pid: The PID of that process. 1030 1031 CPU#: The CPU which the process was running on. 1032 1033 irqs-off: 'd' interrupts are disabled. '.' otherwise. 1034 .. caution:: If the architecture does not support a way to 1035 read the irq flags variable, an 'X' will always 1036 be printed here. 1037 1038 need-resched: 1039 - 'N' both TIF_NEED_RESCHED and PREEMPT_NEED_RESCHED is set, 1040 - 'n' only TIF_NEED_RESCHED is set, 1041 - 'p' only PREEMPT_NEED_RESCHED is set, 1042 - '.' otherwise. 1043 1044 hardirq/softirq: 1045 - 'Z' - NMI occurred inside a hardirq 1046 - 'z' - NMI is running 1047 - 'H' - hard irq occurred inside a softirq. 1048 - 'h' - hard irq is running 1049 - 's' - soft irq is running 1050 - '.' - normal context. 1051 1052 preempt-depth: The level of preempt_disabled 1053 1054The above is mostly meaningful for kernel developers. 1055 1056 time: 1057 When the latency-format option is enabled, the trace file 1058 output includes a timestamp relative to the start of the 1059 trace. This differs from the output when latency-format 1060 is disabled, which includes an absolute timestamp. 1061 1062 delay: 1063 This is just to help catch your eye a bit better. And 1064 needs to be fixed to be only relative to the same CPU. 1065 The marks are determined by the difference between this 1066 current trace and the next trace. 1067 1068 - '$' - greater than 1 second 1069 - '@' - greater than 100 millisecond 1070 - '*' - greater than 10 millisecond 1071 - '#' - greater than 1000 microsecond 1072 - '!' - greater than 100 microsecond 1073 - '+' - greater than 10 microsecond 1074 - ' ' - less than or equal to 10 microsecond. 1075 1076 The rest is the same as the 'trace' file. 1077 1078 Note, the latency tracers will usually end with a back trace 1079 to easily find where the latency occurred. 1080 1081trace_options 1082------------- 1083 1084The trace_options file (or the options directory) is used to control 1085what gets printed in the trace output, or manipulate the tracers. 1086To see what is available, simply cat the file:: 1087 1088 cat trace_options 1089 print-parent 1090 nosym-offset 1091 nosym-addr 1092 noverbose 1093 noraw 1094 nohex 1095 nobin 1096 noblock 1097 nofields 1098 trace_printk 1099 annotate 1100 nouserstacktrace 1101 nosym-userobj 1102 noprintk-msg-only 1103 context-info 1104 nolatency-format 1105 record-cmd 1106 norecord-tgid 1107 overwrite 1108 nodisable_on_free 1109 irq-info 1110 markers 1111 noevent-fork 1112 function-trace 1113 nofunction-fork 1114 nodisplay-graph 1115 nostacktrace 1116 nobranch 1117 1118To disable one of the options, echo in the option prepended with 1119"no":: 1120 1121 echo noprint-parent > trace_options 1122 1123To enable an option, leave off the "no":: 1124 1125 echo sym-offset > trace_options 1126 1127Here are the available options: 1128 1129 print-parent 1130 On function traces, display the calling (parent) 1131 function as well as the function being traced. 1132 :: 1133 1134 print-parent: 1135 bash-4000 [01] 1477.606694: simple_strtoul <-kstrtoul 1136 1137 noprint-parent: 1138 bash-4000 [01] 1477.606694: simple_strtoul 1139 1140 1141 sym-offset 1142 Display not only the function name, but also the 1143 offset in the function. For example, instead of 1144 seeing just "ktime_get", you will see 1145 "ktime_get+0xb/0x20". 1146 :: 1147 1148 sym-offset: 1149 bash-4000 [01] 1477.606694: simple_strtoul+0x6/0xa0 1150 1151 sym-addr 1152 This will also display the function address as well 1153 as the function name. 1154 :: 1155 1156 sym-addr: 1157 bash-4000 [01] 1477.606694: simple_strtoul <c0339346> 1158 1159 verbose 1160 This deals with the trace file when the 1161 latency-format option is enabled. 1162 :: 1163 1164 bash 4000 1 0 00000000 00010a95 [58127d26] 1720.415ms \ 1165 (+0.000ms): simple_strtoul (kstrtoul) 1166 1167 raw 1168 This will display raw numbers. This option is best for 1169 use with user applications that can translate the raw 1170 numbers better than having it done in the kernel. 1171 1172 hex 1173 Similar to raw, but the numbers will be in a hexadecimal format. 1174 1175 bin 1176 This will print out the formats in raw binary. 1177 1178 block 1179 When set, reading trace_pipe will not block when polled. 1180 1181 fields 1182 Print the fields as described by their types. This is a better 1183 option than using hex, bin or raw, as it gives a better parsing 1184 of the content of the event. 1185 1186 trace_printk 1187 Can disable trace_printk() from writing into the buffer. 1188 1189 trace_printk_dest 1190 Set to have trace_printk() and similar internal tracing functions 1191 write into this instance. Note, only one trace instance can have 1192 this set. By setting this flag, it clears the trace_printk_dest flag 1193 of the instance that had it set previously. By default, the top 1194 level trace has this set, and will get it set again if another 1195 instance has it set then clears it. 1196 1197 This flag cannot be cleared by the top level instance, as it is the 1198 default instance. The only way the top level instance has this flag 1199 cleared, is by it being set in another instance. 1200 1201 annotate 1202 It is sometimes confusing when the CPU buffers are full 1203 and one CPU buffer had a lot of events recently, thus 1204 a shorter time frame, were another CPU may have only had 1205 a few events, which lets it have older events. When 1206 the trace is reported, it shows the oldest events first, 1207 and it may look like only one CPU ran (the one with the 1208 oldest events). When the annotate option is set, it will 1209 display when a new CPU buffer started:: 1210 1211 <idle>-0 [001] dNs4 21169.031481: wake_up_idle_cpu <-add_timer_on 1212 <idle>-0 [001] dNs4 21169.031482: _raw_spin_unlock_irqrestore <-add_timer_on 1213 <idle>-0 [001] .Ns4 21169.031484: sub_preempt_count <-_raw_spin_unlock_irqrestore 1214 ##### CPU 2 buffer started #### 1215 <idle>-0 [002] .N.1 21169.031484: rcu_idle_exit <-cpu_idle 1216 <idle>-0 [001] .Ns3 21169.031484: _raw_spin_unlock <-clocksource_watchdog 1217 <idle>-0 [001] .Ns3 21169.031485: sub_preempt_count <-_raw_spin_unlock 1218 1219 userstacktrace 1220 This option changes the trace. It records a 1221 stacktrace of the current user space thread after 1222 each trace event. 1223 1224 sym-userobj 1225 when user stacktrace are enabled, look up which 1226 object the address belongs to, and print a 1227 relative address. This is especially useful when 1228 ASLR is on, otherwise you don't get a chance to 1229 resolve the address to object/file/line after 1230 the app is no longer running 1231 1232 The lookup is performed when you read 1233 trace,trace_pipe. Example:: 1234 1235 a.out-1623 [000] 40874.465068: /root/a.out[+0x480] <-/root/a.out[+0 1236 x494] <- /root/a.out[+0x4a8] <- /lib/libc-2.7.so[+0x1e1a6] 1237 1238 1239 printk-msg-only 1240 When set, trace_printk()s will only show the format 1241 and not their parameters (if trace_bprintk() or 1242 trace_bputs() was used to save the trace_printk()). 1243 1244 context-info 1245 Show only the event data. Hides the comm, PID, 1246 timestamp, CPU, and other useful data. 1247 1248 latency-format 1249 This option changes the trace output. When it is enabled, 1250 the trace displays additional information about the 1251 latency, as described in "Latency trace format". 1252 1253 pause-on-trace 1254 When set, opening the trace file for read, will pause 1255 writing to the ring buffer (as if tracing_on was set to zero). 1256 This simulates the original behavior of the trace file. 1257 When the file is closed, tracing will be enabled again. 1258 1259 hash-ptr 1260 When set, "%p" in the event printk format displays the 1261 hashed pointer value instead of real address. 1262 This will be useful if you want to find out which hashed 1263 value is corresponding to the real value in trace log. 1264 1265 record-cmd 1266 When any event or tracer is enabled, a hook is enabled 1267 in the sched_switch trace point to fill comm cache 1268 with mapped pids and comms. But this may cause some 1269 overhead, and if you only care about pids, and not the 1270 name of the task, disabling this option can lower the 1271 impact of tracing. See "saved_cmdlines". 1272 1273 record-tgid 1274 When any event or tracer is enabled, a hook is enabled 1275 in the sched_switch trace point to fill the cache of 1276 mapped Thread Group IDs (TGID) mapping to pids. See 1277 "saved_tgids". 1278 1279 overwrite 1280 This controls what happens when the trace buffer is 1281 full. If "1" (default), the oldest events are 1282 discarded and overwritten. If "0", then the newest 1283 events are discarded. 1284 (see per_cpu/cpu0/stats for overrun and dropped) 1285 1286 disable_on_free 1287 When the free_buffer is closed, tracing will 1288 stop (tracing_on set to 0). 1289 1290 irq-info 1291 Shows the interrupt, preempt count, need resched data. 1292 When disabled, the trace looks like:: 1293 1294 # tracer: function 1295 # 1296 # entries-in-buffer/entries-written: 144405/9452052 #P:4 1297 # 1298 # TASK-PID CPU# TIMESTAMP FUNCTION 1299 # | | | | | 1300 <idle>-0 [002] 23636.756054: ttwu_do_activate.constprop.89 <-try_to_wake_up 1301 <idle>-0 [002] 23636.756054: activate_task <-ttwu_do_activate.constprop.89 1302 <idle>-0 [002] 23636.756055: enqueue_task <-activate_task 1303 1304 1305 markers 1306 When set, the trace_marker is writable (only by root). 1307 When disabled, the trace_marker will error with EINVAL 1308 on write. 1309 1310 event-fork 1311 When set, tasks with PIDs listed in set_event_pid will have 1312 the PIDs of their children added to set_event_pid when those 1313 tasks fork. Also, when tasks with PIDs in set_event_pid exit, 1314 their PIDs will be removed from the file. 1315 1316 This affects PIDs listed in set_event_notrace_pid as well. 1317 1318 function-trace 1319 The latency tracers will enable function tracing 1320 if this option is enabled (default it is). When 1321 it is disabled, the latency tracers do not trace 1322 functions. This keeps the overhead of the tracer down 1323 when performing latency tests. 1324 1325 function-fork 1326 When set, tasks with PIDs listed in set_ftrace_pid will 1327 have the PIDs of their children added to set_ftrace_pid 1328 when those tasks fork. Also, when tasks with PIDs in 1329 set_ftrace_pid exit, their PIDs will be removed from the 1330 file. 1331 1332 This affects PIDs in set_ftrace_notrace_pid as well. 1333 1334 display-graph 1335 When set, the latency tracers (irqsoff, wakeup, etc) will 1336 use function graph tracing instead of function tracing. 1337 1338 stacktrace 1339 When set, a stack trace is recorded after any trace event 1340 is recorded. 1341 1342 branch 1343 Enable branch tracing with the tracer. This enables branch 1344 tracer along with the currently set tracer. Enabling this 1345 with the "nop" tracer is the same as just enabling the 1346 "branch" tracer. 1347 1348.. tip:: Some tracers have their own options. They only appear in this 1349 file when the tracer is active. They always appear in the 1350 options directory. 1351 1352 1353Here are the per tracer options: 1354 1355Options for function tracer: 1356 1357 func_stack_trace 1358 When set, a stack trace is recorded after every 1359 function that is recorded. NOTE! Limit the functions 1360 that are recorded before enabling this, with 1361 "set_ftrace_filter" otherwise the system performance 1362 will be critically degraded. Remember to disable 1363 this option before clearing the function filter. 1364 1365Options for function_graph tracer: 1366 1367 Since the function_graph tracer has a slightly different output 1368 it has its own options to control what is displayed. 1369 1370 funcgraph-overrun 1371 When set, the "overrun" of the graph stack is 1372 displayed after each function traced. The 1373 overrun, is when the stack depth of the calls 1374 is greater than what is reserved for each task. 1375 Each task has a fixed array of functions to 1376 trace in the call graph. If the depth of the 1377 calls exceeds that, the function is not traced. 1378 The overrun is the number of functions missed 1379 due to exceeding this array. 1380 1381 funcgraph-cpu 1382 When set, the CPU number of the CPU where the trace 1383 occurred is displayed. 1384 1385 funcgraph-overhead 1386 When set, if the function takes longer than 1387 A certain amount, then a delay marker is 1388 displayed. See "delay" above, under the 1389 header description. 1390 1391 funcgraph-proc 1392 Unlike other tracers, the process' command line 1393 is not displayed by default, but instead only 1394 when a task is traced in and out during a context 1395 switch. Enabling this options has the command 1396 of each process displayed at every line. 1397 1398 funcgraph-duration 1399 At the end of each function (the return) 1400 the duration of the amount of time in the 1401 function is displayed in microseconds. 1402 1403 funcgraph-abstime 1404 When set, the timestamp is displayed at each line. 1405 1406 funcgraph-irqs 1407 When disabled, functions that happen inside an 1408 interrupt will not be traced. 1409 1410 funcgraph-tail 1411 When set, the return event will include the function 1412 that it represents. By default this is off, and 1413 only a closing curly bracket "}" is displayed for 1414 the return of a function. 1415 1416 funcgraph-retval 1417 When set, the return value of each traced function 1418 will be printed after an equal sign "=". By default 1419 this is off. 1420 1421 funcgraph-retval-hex 1422 When set, the return value will always be printed 1423 in hexadecimal format. If the option is not set and 1424 the return value is an error code, it will be printed 1425 in signed decimal format; otherwise it will also be 1426 printed in hexadecimal format. By default, this option 1427 is off. 1428 1429 sleep-time 1430 When running function graph tracer, to include 1431 the time a task schedules out in its function. 1432 When enabled, it will account time the task has been 1433 scheduled out as part of the function call. 1434 1435 graph-time 1436 When running function profiler with function graph tracer, 1437 to include the time to call nested functions. When this is 1438 not set, the time reported for the function will only 1439 include the time the function itself executed for, not the 1440 time for functions that it called. 1441 1442Options for blk tracer: 1443 1444 blk_classic 1445 Shows a more minimalistic output. 1446 1447 1448irqsoff 1449------- 1450 1451When interrupts are disabled, the CPU can not react to any other 1452external event (besides NMIs and SMIs). This prevents the timer 1453interrupt from triggering or the mouse interrupt from letting 1454the kernel know of a new mouse event. The result is a latency 1455with the reaction time. 1456 1457The irqsoff tracer tracks the time for which interrupts are 1458disabled. When a new maximum latency is hit, the tracer saves 1459the trace leading up to that latency point so that every time a 1460new maximum is reached, the old saved trace is discarded and the 1461new trace is saved. 1462 1463To reset the maximum, echo 0 into tracing_max_latency. Here is 1464an example:: 1465 1466 # echo 0 > options/function-trace 1467 # echo irqsoff > current_tracer 1468 # echo 1 > tracing_on 1469 # echo 0 > tracing_max_latency 1470 # ls -ltr 1471 [...] 1472 # echo 0 > tracing_on 1473 # cat trace 1474 # tracer: irqsoff 1475 # 1476 # irqsoff latency trace v1.1.5 on 3.8.0-test+ 1477 # -------------------------------------------------------------------- 1478 # latency: 16 us, #4/4, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4) 1479 # ----------------- 1480 # | task: swapper/0-0 (uid:0 nice:0 policy:0 rt_prio:0) 1481 # ----------------- 1482 # => started at: run_timer_softirq 1483 # => ended at: run_timer_softirq 1484 # 1485 # 1486 # _------=> CPU# 1487 # / _-----=> irqs-off 1488 # | / _----=> need-resched 1489 # || / _---=> hardirq/softirq 1490 # ||| / _--=> preempt-depth 1491 # |||| / delay 1492 # cmd pid ||||| time | caller 1493 # \ / ||||| \ | / 1494 <idle>-0 0d.s2 0us+: _raw_spin_lock_irq <-run_timer_softirq 1495 <idle>-0 0dNs3 17us : _raw_spin_unlock_irq <-run_timer_softirq 1496 <idle>-0 0dNs3 17us+: trace_hardirqs_on <-run_timer_softirq 1497 <idle>-0 0dNs3 25us : <stack trace> 1498 => _raw_spin_unlock_irq 1499 => run_timer_softirq 1500 => __do_softirq 1501 => call_softirq 1502 => do_softirq 1503 => irq_exit 1504 => smp_apic_timer_interrupt 1505 => apic_timer_interrupt 1506 => rcu_idle_exit 1507 => cpu_idle 1508 => rest_init 1509 => start_kernel 1510 => x86_64_start_reservations 1511 => x86_64_start_kernel 1512 1513Here we see that we had a latency of 16 microseconds (which is 1514very good). The _raw_spin_lock_irq in run_timer_softirq disabled 1515interrupts. The difference between the 16 and the displayed 1516timestamp 25us occurred because the clock was incremented 1517between the time of recording the max latency and the time of 1518recording the function that had that latency. 1519 1520Note the above example had function-trace not set. If we set 1521function-trace, we get a much larger output:: 1522 1523 with echo 1 > options/function-trace 1524 1525 # tracer: irqsoff 1526 # 1527 # irqsoff latency trace v1.1.5 on 3.8.0-test+ 1528 # -------------------------------------------------------------------- 1529 # latency: 71 us, #168/168, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4) 1530 # ----------------- 1531 # | task: bash-2042 (uid:0 nice:0 policy:0 rt_prio:0) 1532 # ----------------- 1533 # => started at: ata_scsi_queuecmd 1534 # => ended at: ata_scsi_queuecmd 1535 # 1536 # 1537 # _------=> CPU# 1538 # / _-----=> irqs-off 1539 # | / _----=> need-resched 1540 # || / _---=> hardirq/softirq 1541 # ||| / _--=> preempt-depth 1542 # |||| / delay 1543 # cmd pid ||||| time | caller 1544 # \ / ||||| \ | / 1545 bash-2042 3d... 0us : _raw_spin_lock_irqsave <-ata_scsi_queuecmd 1546 bash-2042 3d... 0us : add_preempt_count <-_raw_spin_lock_irqsave 1547 bash-2042 3d..1 1us : ata_scsi_find_dev <-ata_scsi_queuecmd 1548 bash-2042 3d..1 1us : __ata_scsi_find_dev <-ata_scsi_find_dev 1549 bash-2042 3d..1 2us : ata_find_dev.part.14 <-__ata_scsi_find_dev 1550 bash-2042 3d..1 2us : ata_qc_new_init <-__ata_scsi_queuecmd 1551 bash-2042 3d..1 3us : ata_sg_init <-__ata_scsi_queuecmd 1552 bash-2042 3d..1 4us : ata_scsi_rw_xlat <-__ata_scsi_queuecmd 1553 bash-2042 3d..1 4us : ata_build_rw_tf <-ata_scsi_rw_xlat 1554 [...] 1555 bash-2042 3d..1 67us : delay_tsc <-__delay 1556 bash-2042 3d..1 67us : add_preempt_count <-delay_tsc 1557 bash-2042 3d..2 67us : sub_preempt_count <-delay_tsc 1558 bash-2042 3d..1 67us : add_preempt_count <-delay_tsc 1559 bash-2042 3d..2 68us : sub_preempt_count <-delay_tsc 1560 bash-2042 3d..1 68us+: ata_bmdma_start <-ata_bmdma_qc_issue 1561 bash-2042 3d..1 71us : _raw_spin_unlock_irqrestore <-ata_scsi_queuecmd 1562 bash-2042 3d..1 71us : _raw_spin_unlock_irqrestore <-ata_scsi_queuecmd 1563 bash-2042 3d..1 72us+: trace_hardirqs_on <-ata_scsi_queuecmd 1564 bash-2042 3d..1 120us : <stack trace> 1565 => _raw_spin_unlock_irqrestore 1566 => ata_scsi_queuecmd 1567 => scsi_dispatch_cmd 1568 => scsi_request_fn 1569 => __blk_run_queue_uncond 1570 => __blk_run_queue 1571 => blk_queue_bio 1572 => submit_bio_noacct 1573 => submit_bio 1574 => submit_bh 1575 => __ext3_get_inode_loc 1576 => ext3_iget 1577 => ext3_lookup 1578 => lookup_real 1579 => __lookup_hash 1580 => walk_component 1581 => lookup_last 1582 => path_lookupat 1583 => filename_lookup 1584 => user_path_at_empty 1585 => user_path_at 1586 => vfs_fstatat 1587 => vfs_stat 1588 => sys_newstat 1589 => system_call_fastpath 1590 1591 1592Here we traced a 71 microsecond latency. But we also see all the 1593functions that were called during that time. Note that by 1594enabling function tracing, we incur an added overhead. This 1595overhead may extend the latency times. But nevertheless, this 1596trace has provided some very helpful debugging information. 1597 1598If we prefer function graph output instead of function, we can set 1599display-graph option:: 1600 1601 with echo 1 > options/display-graph 1602 1603 # tracer: irqsoff 1604 # 1605 # irqsoff latency trace v1.1.5 on 4.20.0-rc6+ 1606 # -------------------------------------------------------------------- 1607 # latency: 3751 us, #274/274, CPU#0 | (M:desktop VP:0, KP:0, SP:0 HP:0 #P:4) 1608 # ----------------- 1609 # | task: bash-1507 (uid:0 nice:0 policy:0 rt_prio:0) 1610 # ----------------- 1611 # => started at: free_debug_processing 1612 # => ended at: return_to_handler 1613 # 1614 # 1615 # _-----=> irqs-off 1616 # / _----=> need-resched 1617 # | / _---=> hardirq/softirq 1618 # || / _--=> preempt-depth 1619 # ||| / 1620 # REL TIME CPU TASK/PID |||| DURATION FUNCTION CALLS 1621 # | | | | |||| | | | | | | 1622 0 us | 0) bash-1507 | d... | 0.000 us | _raw_spin_lock_irqsave(); 1623 0 us | 0) bash-1507 | d..1 | 0.378 us | do_raw_spin_trylock(); 1624 1 us | 0) bash-1507 | d..2 | | set_track() { 1625 2 us | 0) bash-1507 | d..2 | | save_stack_trace() { 1626 2 us | 0) bash-1507 | d..2 | | __save_stack_trace() { 1627 3 us | 0) bash-1507 | d..2 | | __unwind_start() { 1628 3 us | 0) bash-1507 | d..2 | | get_stack_info() { 1629 3 us | 0) bash-1507 | d..2 | 0.351 us | in_task_stack(); 1630 4 us | 0) bash-1507 | d..2 | 1.107 us | } 1631 [...] 1632 3750 us | 0) bash-1507 | d..1 | 0.516 us | do_raw_spin_unlock(); 1633 3750 us | 0) bash-1507 | d..1 | 0.000 us | _raw_spin_unlock_irqrestore(); 1634 3764 us | 0) bash-1507 | d..1 | 0.000 us | tracer_hardirqs_on(); 1635 bash-1507 0d..1 3792us : <stack trace> 1636 => free_debug_processing 1637 => __slab_free 1638 => kmem_cache_free 1639 => vm_area_free 1640 => remove_vma 1641 => exit_mmap 1642 => mmput 1643 => begin_new_exec 1644 => load_elf_binary 1645 => search_binary_handler 1646 => __do_execve_file.isra.32 1647 => __x64_sys_execve 1648 => do_syscall_64 1649 => entry_SYSCALL_64_after_hwframe 1650 1651preemptoff 1652---------- 1653 1654When preemption is disabled, we may be able to receive 1655interrupts but the task cannot be preempted and a higher 1656priority task must wait for preemption to be enabled again 1657before it can preempt a lower priority task. 1658 1659The preemptoff tracer traces the places that disable preemption. 1660Like the irqsoff tracer, it records the maximum latency for 1661which preemption was disabled. The control of preemptoff tracer 1662is much like the irqsoff tracer. 1663:: 1664 1665 # echo 0 > options/function-trace 1666 # echo preemptoff > current_tracer 1667 # echo 1 > tracing_on 1668 # echo 0 > tracing_max_latency 1669 # ls -ltr 1670 [...] 1671 # echo 0 > tracing_on 1672 # cat trace 1673 # tracer: preemptoff 1674 # 1675 # preemptoff latency trace v1.1.5 on 3.8.0-test+ 1676 # -------------------------------------------------------------------- 1677 # latency: 46 us, #4/4, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4) 1678 # ----------------- 1679 # | task: sshd-1991 (uid:0 nice:0 policy:0 rt_prio:0) 1680 # ----------------- 1681 # => started at: do_IRQ 1682 # => ended at: do_IRQ 1683 # 1684 # 1685 # _------=> CPU# 1686 # / _-----=> irqs-off 1687 # | / _----=> need-resched 1688 # || / _---=> hardirq/softirq 1689 # ||| / _--=> preempt-depth 1690 # |||| / delay 1691 # cmd pid ||||| time | caller 1692 # \ / ||||| \ | / 1693 sshd-1991 1d.h. 0us+: irq_enter <-do_IRQ 1694 sshd-1991 1d..1 46us : irq_exit <-do_IRQ 1695 sshd-1991 1d..1 47us+: trace_preempt_on <-do_IRQ 1696 sshd-1991 1d..1 52us : <stack trace> 1697 => sub_preempt_count 1698 => irq_exit 1699 => do_IRQ 1700 => ret_from_intr 1701 1702 1703This has some more changes. Preemption was disabled when an 1704interrupt came in (notice the 'h'), and was enabled on exit. 1705But we also see that interrupts have been disabled when entering 1706the preempt off section and leaving it (the 'd'). We do not know if 1707interrupts were enabled in the mean time or shortly after this 1708was over. 1709:: 1710 1711 # tracer: preemptoff 1712 # 1713 # preemptoff latency trace v1.1.5 on 3.8.0-test+ 1714 # -------------------------------------------------------------------- 1715 # latency: 83 us, #241/241, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4) 1716 # ----------------- 1717 # | task: bash-1994 (uid:0 nice:0 policy:0 rt_prio:0) 1718 # ----------------- 1719 # => started at: wake_up_new_task 1720 # => ended at: task_rq_unlock 1721 # 1722 # 1723 # _------=> CPU# 1724 # / _-----=> irqs-off 1725 # | / _----=> need-resched 1726 # || / _---=> hardirq/softirq 1727 # ||| / _--=> preempt-depth 1728 # |||| / delay 1729 # cmd pid ||||| time | caller 1730 # \ / ||||| \ | / 1731 bash-1994 1d..1 0us : _raw_spin_lock_irqsave <-wake_up_new_task 1732 bash-1994 1d..1 0us : select_task_rq_fair <-select_task_rq 1733 bash-1994 1d..1 1us : __rcu_read_lock <-select_task_rq_fair 1734 bash-1994 1d..1 1us : source_load <-select_task_rq_fair 1735 bash-1994 1d..1 1us : source_load <-select_task_rq_fair 1736 [...] 1737 bash-1994 1d..1 12us : irq_enter <-smp_apic_timer_interrupt 1738 bash-1994 1d..1 12us : rcu_irq_enter <-irq_enter 1739 bash-1994 1d..1 13us : add_preempt_count <-irq_enter 1740 bash-1994 1d.h1 13us : exit_idle <-smp_apic_timer_interrupt 1741 bash-1994 1d.h1 13us : hrtimer_interrupt <-smp_apic_timer_interrupt 1742 bash-1994 1d.h1 13us : _raw_spin_lock <-hrtimer_interrupt 1743 bash-1994 1d.h1 14us : add_preempt_count <-_raw_spin_lock 1744 bash-1994 1d.h2 14us : ktime_get_update_offsets <-hrtimer_interrupt 1745 [...] 1746 bash-1994 1d.h1 35us : lapic_next_event <-clockevents_program_event 1747 bash-1994 1d.h1 35us : irq_exit <-smp_apic_timer_interrupt 1748 bash-1994 1d.h1 36us : sub_preempt_count <-irq_exit 1749 bash-1994 1d..2 36us : do_softirq <-irq_exit 1750 bash-1994 1d..2 36us : __do_softirq <-call_softirq 1751 bash-1994 1d..2 36us : __local_bh_disable <-__do_softirq 1752 bash-1994 1d.s2 37us : add_preempt_count <-_raw_spin_lock_irq 1753 bash-1994 1d.s3 38us : _raw_spin_unlock <-run_timer_softirq 1754 bash-1994 1d.s3 39us : sub_preempt_count <-_raw_spin_unlock 1755 bash-1994 1d.s2 39us : call_timer_fn <-run_timer_softirq 1756 [...] 1757 bash-1994 1dNs2 81us : cpu_needs_another_gp <-rcu_process_callbacks 1758 bash-1994 1dNs2 82us : __local_bh_enable <-__do_softirq 1759 bash-1994 1dNs2 82us : sub_preempt_count <-__local_bh_enable 1760 bash-1994 1dN.2 82us : idle_cpu <-irq_exit 1761 bash-1994 1dN.2 83us : rcu_irq_exit <-irq_exit 1762 bash-1994 1dN.2 83us : sub_preempt_count <-irq_exit 1763 bash-1994 1.N.1 84us : _raw_spin_unlock_irqrestore <-task_rq_unlock 1764 bash-1994 1.N.1 84us+: trace_preempt_on <-task_rq_unlock 1765 bash-1994 1.N.1 104us : <stack trace> 1766 => sub_preempt_count 1767 => _raw_spin_unlock_irqrestore 1768 => task_rq_unlock 1769 => wake_up_new_task 1770 => do_fork 1771 => sys_clone 1772 => stub_clone 1773 1774 1775The above is an example of the preemptoff trace with 1776function-trace set. Here we see that interrupts were not disabled 1777the entire time. The irq_enter code lets us know that we entered 1778an interrupt 'h'. Before that, the functions being traced still 1779show that it is not in an interrupt, but we can see from the 1780functions themselves that this is not the case. 1781 1782preemptirqsoff 1783-------------- 1784 1785Knowing the locations that have interrupts disabled or 1786preemption disabled for the longest times is helpful. But 1787sometimes we would like to know when either preemption and/or 1788interrupts are disabled. 1789 1790Consider the following code:: 1791 1792 local_irq_disable(); 1793 call_function_with_irqs_off(); 1794 preempt_disable(); 1795 call_function_with_irqs_and_preemption_off(); 1796 local_irq_enable(); 1797 call_function_with_preemption_off(); 1798 preempt_enable(); 1799 1800The irqsoff tracer will record the total length of 1801call_function_with_irqs_off() and 1802call_function_with_irqs_and_preemption_off(). 1803 1804The preemptoff tracer will record the total length of 1805call_function_with_irqs_and_preemption_off() and 1806call_function_with_preemption_off(). 1807 1808But neither will trace the time that interrupts and/or 1809preemption is disabled. This total time is the time that we can 1810not schedule. To record this time, use the preemptirqsoff 1811tracer. 1812 1813Again, using this trace is much like the irqsoff and preemptoff 1814tracers. 1815:: 1816 1817 # echo 0 > options/function-trace 1818 # echo preemptirqsoff > current_tracer 1819 # echo 1 > tracing_on 1820 # echo 0 > tracing_max_latency 1821 # ls -ltr 1822 [...] 1823 # echo 0 > tracing_on 1824 # cat trace 1825 # tracer: preemptirqsoff 1826 # 1827 # preemptirqsoff latency trace v1.1.5 on 3.8.0-test+ 1828 # -------------------------------------------------------------------- 1829 # latency: 100 us, #4/4, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4) 1830 # ----------------- 1831 # | task: ls-2230 (uid:0 nice:0 policy:0 rt_prio:0) 1832 # ----------------- 1833 # => started at: ata_scsi_queuecmd 1834 # => ended at: ata_scsi_queuecmd 1835 # 1836 # 1837 # _------=> CPU# 1838 # / _-----=> irqs-off 1839 # | / _----=> need-resched 1840 # || / _---=> hardirq/softirq 1841 # ||| / _--=> preempt-depth 1842 # |||| / delay 1843 # cmd pid ||||| time | caller 1844 # \ / ||||| \ | / 1845 ls-2230 3d... 0us+: _raw_spin_lock_irqsave <-ata_scsi_queuecmd 1846 ls-2230 3...1 100us : _raw_spin_unlock_irqrestore <-ata_scsi_queuecmd 1847 ls-2230 3...1 101us+: trace_preempt_on <-ata_scsi_queuecmd 1848 ls-2230 3...1 111us : <stack trace> 1849 => sub_preempt_count 1850 => _raw_spin_unlock_irqrestore 1851 => ata_scsi_queuecmd 1852 => scsi_dispatch_cmd 1853 => scsi_request_fn 1854 => __blk_run_queue_uncond 1855 => __blk_run_queue 1856 => blk_queue_bio 1857 => submit_bio_noacct 1858 => submit_bio 1859 => submit_bh 1860 => ext3_bread 1861 => ext3_dir_bread 1862 => htree_dirblock_to_tree 1863 => ext3_htree_fill_tree 1864 => ext3_readdir 1865 => vfs_readdir 1866 => sys_getdents 1867 => system_call_fastpath 1868 1869 1870The trace_hardirqs_off_thunk is called from assembly on x86 when 1871interrupts are disabled in the assembly code. Without the 1872function tracing, we do not know if interrupts were enabled 1873within the preemption points. We do see that it started with 1874preemption enabled. 1875 1876Here is a trace with function-trace set:: 1877 1878 # tracer: preemptirqsoff 1879 # 1880 # preemptirqsoff latency trace v1.1.5 on 3.8.0-test+ 1881 # -------------------------------------------------------------------- 1882 # latency: 161 us, #339/339, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4) 1883 # ----------------- 1884 # | task: ls-2269 (uid:0 nice:0 policy:0 rt_prio:0) 1885 # ----------------- 1886 # => started at: schedule 1887 # => ended at: mutex_unlock 1888 # 1889 # 1890 # _------=> CPU# 1891 # / _-----=> irqs-off 1892 # | / _----=> need-resched 1893 # || / _---=> hardirq/softirq 1894 # ||| / _--=> preempt-depth 1895 # |||| / delay 1896 # cmd pid ||||| time | caller 1897 # \ / ||||| \ | / 1898 kworker/-59 3...1 0us : __schedule <-schedule 1899 kworker/-59 3d..1 0us : rcu_preempt_qs <-rcu_note_context_switch 1900 kworker/-59 3d..1 1us : add_preempt_count <-_raw_spin_lock_irq 1901 kworker/-59 3d..2 1us : deactivate_task <-__schedule 1902 kworker/-59 3d..2 1us : dequeue_task <-deactivate_task 1903 kworker/-59 3d..2 2us : update_rq_clock <-dequeue_task 1904 kworker/-59 3d..2 2us : dequeue_task_fair <-dequeue_task 1905 kworker/-59 3d..2 2us : update_curr <-dequeue_task_fair 1906 kworker/-59 3d..2 2us : update_min_vruntime <-update_curr 1907 kworker/-59 3d..2 3us : cpuacct_charge <-update_curr 1908 kworker/-59 3d..2 3us : __rcu_read_lock <-cpuacct_charge 1909 kworker/-59 3d..2 3us : __rcu_read_unlock <-cpuacct_charge 1910 kworker/-59 3d..2 3us : update_cfs_rq_blocked_load <-dequeue_task_fair 1911 kworker/-59 3d..2 4us : clear_buddies <-dequeue_task_fair 1912 kworker/-59 3d..2 4us : account_entity_dequeue <-dequeue_task_fair 1913 kworker/-59 3d..2 4us : update_min_vruntime <-dequeue_task_fair 1914 kworker/-59 3d..2 4us : update_cfs_shares <-dequeue_task_fair 1915 kworker/-59 3d..2 5us : hrtick_update <-dequeue_task_fair 1916 kworker/-59 3d..2 5us : wq_worker_sleeping <-__schedule 1917 kworker/-59 3d..2 5us : kthread_data <-wq_worker_sleeping 1918 kworker/-59 3d..2 5us : put_prev_task_fair <-__schedule 1919 kworker/-59 3d..2 6us : pick_next_task_fair <-pick_next_task 1920 kworker/-59 3d..2 6us : clear_buddies <-pick_next_task_fair 1921 kworker/-59 3d..2 6us : set_next_entity <-pick_next_task_fair 1922 kworker/-59 3d..2 6us : update_stats_wait_end <-set_next_entity 1923 ls-2269 3d..2 7us : finish_task_switch <-__schedule 1924 ls-2269 3d..2 7us : _raw_spin_unlock_irq <-finish_task_switch 1925 ls-2269 3d..2 8us : do_IRQ <-ret_from_intr 1926 ls-2269 3d..2 8us : irq_enter <-do_IRQ 1927 ls-2269 3d..2 8us : rcu_irq_enter <-irq_enter 1928 ls-2269 3d..2 9us : add_preempt_count <-irq_enter 1929 ls-2269 3d.h2 9us : exit_idle <-do_IRQ 1930 [...] 1931 ls-2269 3d.h3 20us : sub_preempt_count <-_raw_spin_unlock 1932 ls-2269 3d.h2 20us : irq_exit <-do_IRQ 1933 ls-2269 3d.h2 21us : sub_preempt_count <-irq_exit 1934 ls-2269 3d..3 21us : do_softirq <-irq_exit 1935 ls-2269 3d..3 21us : __do_softirq <-call_softirq 1936 ls-2269 3d..3 21us+: __local_bh_disable <-__do_softirq 1937 ls-2269 3d.s4 29us : sub_preempt_count <-_local_bh_enable_ip 1938 ls-2269 3d.s5 29us : sub_preempt_count <-_local_bh_enable_ip 1939 ls-2269 3d.s5 31us : do_IRQ <-ret_from_intr 1940 ls-2269 3d.s5 31us : irq_enter <-do_IRQ 1941 ls-2269 3d.s5 31us : rcu_irq_enter <-irq_enter 1942 [...] 1943 ls-2269 3d.s5 31us : rcu_irq_enter <-irq_enter 1944 ls-2269 3d.s5 32us : add_preempt_count <-irq_enter 1945 ls-2269 3d.H5 32us : exit_idle <-do_IRQ 1946 ls-2269 3d.H5 32us : handle_irq <-do_IRQ 1947 ls-2269 3d.H5 32us : irq_to_desc <-handle_irq 1948 ls-2269 3d.H5 33us : handle_fasteoi_irq <-handle_irq 1949 [...] 1950 ls-2269 3d.s5 158us : _raw_spin_unlock_irqrestore <-rtl8139_poll 1951 ls-2269 3d.s3 158us : net_rps_action_and_irq_enable.isra.65 <-net_rx_action 1952 ls-2269 3d.s3 159us : __local_bh_enable <-__do_softirq 1953 ls-2269 3d.s3 159us : sub_preempt_count <-__local_bh_enable 1954 ls-2269 3d..3 159us : idle_cpu <-irq_exit 1955 ls-2269 3d..3 159us : rcu_irq_exit <-irq_exit 1956 ls-2269 3d..3 160us : sub_preempt_count <-irq_exit 1957 ls-2269 3d... 161us : __mutex_unlock_slowpath <-mutex_unlock 1958 ls-2269 3d... 162us+: trace_hardirqs_on <-mutex_unlock 1959 ls-2269 3d... 186us : <stack trace> 1960 => __mutex_unlock_slowpath 1961 => mutex_unlock 1962 => process_output 1963 => n_tty_write 1964 => tty_write 1965 => vfs_write 1966 => sys_write 1967 => system_call_fastpath 1968 1969This is an interesting trace. It started with kworker running and 1970scheduling out and ls taking over. But as soon as ls released the 1971rq lock and enabled interrupts (but not preemption) an interrupt 1972triggered. When the interrupt finished, it started running softirqs. 1973But while the softirq was running, another interrupt triggered. 1974When an interrupt is running inside a softirq, the annotation is 'H'. 1975 1976 1977wakeup 1978------ 1979 1980One common case that people are interested in tracing is the 1981time it takes for a task that is woken to actually wake up. 1982Now for non Real-Time tasks, this can be arbitrary. But tracing 1983it nonetheless can be interesting. 1984 1985Without function tracing:: 1986 1987 # echo 0 > options/function-trace 1988 # echo wakeup > current_tracer 1989 # echo 1 > tracing_on 1990 # echo 0 > tracing_max_latency 1991 # chrt -f 5 sleep 1 1992 # echo 0 > tracing_on 1993 # cat trace 1994 # tracer: wakeup 1995 # 1996 # wakeup latency trace v1.1.5 on 3.8.0-test+ 1997 # -------------------------------------------------------------------- 1998 # latency: 15 us, #4/4, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4) 1999 # ----------------- 2000 # | task: kworker/3:1H-312 (uid:0 nice:-20 policy:0 rt_prio:0) 2001 # ----------------- 2002 # 2003 # _------=> CPU# 2004 # / _-----=> irqs-off 2005 # | / _----=> need-resched 2006 # || / _---=> hardirq/softirq 2007 # ||| / _--=> preempt-depth 2008 # |||| / delay 2009 # cmd pid ||||| time | caller 2010 # \ / ||||| \ | / 2011 <idle>-0 3dNs7 0us : 0:120:R + [003] 312:100:R kworker/3:1H 2012 <idle>-0 3dNs7 1us+: ttwu_do_activate.constprop.87 <-try_to_wake_up 2013 <idle>-0 3d..3 15us : __schedule <-schedule 2014 <idle>-0 3d..3 15us : 0:120:R ==> [003] 312:100:R kworker/3:1H 2015 2016The tracer only traces the highest priority task in the system 2017to avoid tracing the normal circumstances. Here we see that 2018the kworker with a nice priority of -20 (not very nice), took 2019just 15 microseconds from the time it woke up, to the time it 2020ran. 2021 2022Non Real-Time tasks are not that interesting. A more interesting 2023trace is to concentrate only on Real-Time tasks. 2024 2025wakeup_rt 2026--------- 2027 2028In a Real-Time environment it is very important to know the 2029wakeup time it takes for the highest priority task that is woken 2030up to the time that it executes. This is also known as "schedule 2031latency". I stress the point that this is about RT tasks. It is 2032also important to know the scheduling latency of non-RT tasks, 2033but the average schedule latency is better for non-RT tasks. 2034Tools like LatencyTop are more appropriate for such 2035measurements. 2036 2037Real-Time environments are interested in the worst case latency. 2038That is the longest latency it takes for something to happen, 2039and not the average. We can have a very fast scheduler that may 2040only have a large latency once in a while, but that would not 2041work well with Real-Time tasks. The wakeup_rt tracer was designed 2042to record the worst case wakeups of RT tasks. Non-RT tasks are 2043not recorded because the tracer only records one worst case and 2044tracing non-RT tasks that are unpredictable will overwrite the 2045worst case latency of RT tasks (just run the normal wakeup 2046tracer for a while to see that effect). 2047 2048Since this tracer only deals with RT tasks, we will run this 2049slightly differently than we did with the previous tracers. 2050Instead of performing an 'ls', we will run 'sleep 1' under 2051'chrt' which changes the priority of the task. 2052:: 2053 2054 # echo 0 > options/function-trace 2055 # echo wakeup_rt > current_tracer 2056 # echo 1 > tracing_on 2057 # echo 0 > tracing_max_latency 2058 # chrt -f 5 sleep 1 2059 # echo 0 > tracing_on 2060 # cat trace 2061 # tracer: wakeup 2062 # 2063 # tracer: wakeup_rt 2064 # 2065 # wakeup_rt latency trace v1.1.5 on 3.8.0-test+ 2066 # -------------------------------------------------------------------- 2067 # latency: 5 us, #4/4, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4) 2068 # ----------------- 2069 # | task: sleep-2389 (uid:0 nice:0 policy:1 rt_prio:5) 2070 # ----------------- 2071 # 2072 # _------=> CPU# 2073 # / _-----=> irqs-off 2074 # | / _----=> need-resched 2075 # || / _---=> hardirq/softirq 2076 # ||| / _--=> preempt-depth 2077 # |||| / delay 2078 # cmd pid ||||| time | caller 2079 # \ / ||||| \ | / 2080 <idle>-0 3d.h4 0us : 0:120:R + [003] 2389: 94:R sleep 2081 <idle>-0 3d.h4 1us+: ttwu_do_activate.constprop.87 <-try_to_wake_up 2082 <idle>-0 3d..3 5us : __schedule <-schedule 2083 <idle>-0 3d..3 5us : 0:120:R ==> [003] 2389: 94:R sleep 2084 2085 2086Running this on an idle system, we see that it only took 5 microseconds 2087to perform the task switch. Note, since the trace point in the schedule 2088is before the actual "switch", we stop the tracing when the recorded task 2089is about to schedule in. This may change if we add a new marker at the 2090end of the scheduler. 2091 2092Notice that the recorded task is 'sleep' with the PID of 2389 2093and it has an rt_prio of 5. This priority is user-space priority 2094and not the internal kernel priority. The policy is 1 for 2095SCHED_FIFO and 2 for SCHED_RR. 2096 2097Note, that the trace data shows the internal priority (99 - rtprio). 2098:: 2099 2100 <idle>-0 3d..3 5us : 0:120:R ==> [003] 2389: 94:R sleep 2101 2102The 0:120:R means idle was running with a nice priority of 0 (120 - 120) 2103and in the running state 'R'. The sleep task was scheduled in with 21042389: 94:R. That is the priority is the kernel rtprio (99 - 5 = 94) 2105and it too is in the running state. 2106 2107Doing the same with chrt -r 5 and function-trace set. 2108:: 2109 2110 echo 1 > options/function-trace 2111 2112 # tracer: wakeup_rt 2113 # 2114 # wakeup_rt latency trace v1.1.5 on 3.8.0-test+ 2115 # -------------------------------------------------------------------- 2116 # latency: 29 us, #85/85, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4) 2117 # ----------------- 2118 # | task: sleep-2448 (uid:0 nice:0 policy:1 rt_prio:5) 2119 # ----------------- 2120 # 2121 # _------=> CPU# 2122 # / _-----=> irqs-off 2123 # | / _----=> need-resched 2124 # || / _---=> hardirq/softirq 2125 # ||| / _--=> preempt-depth 2126 # |||| / delay 2127 # cmd pid ||||| time | caller 2128 # \ / ||||| \ | / 2129 <idle>-0 3d.h4 1us+: 0:120:R + [003] 2448: 94:R sleep 2130 <idle>-0 3d.h4 2us : ttwu_do_activate.constprop.87 <-try_to_wake_up 2131 <idle>-0 3d.h3 3us : check_preempt_curr <-ttwu_do_wakeup 2132 <idle>-0 3d.h3 3us : resched_curr <-check_preempt_curr 2133 <idle>-0 3dNh3 4us : task_woken_rt <-ttwu_do_wakeup 2134 <idle>-0 3dNh3 4us : _raw_spin_unlock <-try_to_wake_up 2135 <idle>-0 3dNh3 4us : sub_preempt_count <-_raw_spin_unlock 2136 <idle>-0 3dNh2 5us : ttwu_stat <-try_to_wake_up 2137 <idle>-0 3dNh2 5us : _raw_spin_unlock_irqrestore <-try_to_wake_up 2138 <idle>-0 3dNh2 6us : sub_preempt_count <-_raw_spin_unlock_irqrestore 2139 <idle>-0 3dNh1 6us : _raw_spin_lock <-__run_hrtimer 2140 <idle>-0 3dNh1 6us : add_preempt_count <-_raw_spin_lock 2141 <idle>-0 3dNh2 7us : _raw_spin_unlock <-hrtimer_interrupt 2142 <idle>-0 3dNh2 7us : sub_preempt_count <-_raw_spin_unlock 2143 <idle>-0 3dNh1 7us : tick_program_event <-hrtimer_interrupt 2144 <idle>-0 3dNh1 7us : clockevents_program_event <-tick_program_event 2145 <idle>-0 3dNh1 8us : ktime_get <-clockevents_program_event 2146 <idle>-0 3dNh1 8us : lapic_next_event <-clockevents_program_event 2147 <idle>-0 3dNh1 8us : irq_exit <-smp_apic_timer_interrupt 2148 <idle>-0 3dNh1 9us : sub_preempt_count <-irq_exit 2149 <idle>-0 3dN.2 9us : idle_cpu <-irq_exit 2150 <idle>-0 3dN.2 9us : rcu_irq_exit <-irq_exit 2151 <idle>-0 3dN.2 10us : rcu_eqs_enter_common.isra.45 <-rcu_irq_exit 2152 <idle>-0 3dN.2 10us : sub_preempt_count <-irq_exit 2153 <idle>-0 3.N.1 11us : rcu_idle_exit <-cpu_idle 2154 <idle>-0 3dN.1 11us : rcu_eqs_exit_common.isra.43 <-rcu_idle_exit 2155 <idle>-0 3.N.1 11us : tick_nohz_idle_exit <-cpu_idle 2156 <idle>-0 3dN.1 12us : menu_hrtimer_cancel <-tick_nohz_idle_exit 2157 <idle>-0 3dN.1 12us : ktime_get <-tick_nohz_idle_exit 2158 <idle>-0 3dN.1 12us : tick_do_update_jiffies64 <-tick_nohz_idle_exit 2159 <idle>-0 3dN.1 13us : cpu_load_update_nohz <-tick_nohz_idle_exit 2160 <idle>-0 3dN.1 13us : _raw_spin_lock <-cpu_load_update_nohz 2161 <idle>-0 3dN.1 13us : add_preempt_count <-_raw_spin_lock 2162 <idle>-0 3dN.2 13us : __cpu_load_update <-cpu_load_update_nohz 2163 <idle>-0 3dN.2 14us : sched_avg_update <-__cpu_load_update 2164 <idle>-0 3dN.2 14us : _raw_spin_unlock <-cpu_load_update_nohz 2165 <idle>-0 3dN.2 14us : sub_preempt_count <-_raw_spin_unlock 2166 <idle>-0 3dN.1 15us : calc_load_nohz_stop <-tick_nohz_idle_exit 2167 <idle>-0 3dN.1 15us : touch_softlockup_watchdog <-tick_nohz_idle_exit 2168 <idle>-0 3dN.1 15us : hrtimer_cancel <-tick_nohz_idle_exit 2169 <idle>-0 3dN.1 15us : hrtimer_try_to_cancel <-hrtimer_cancel 2170 <idle>-0 3dN.1 16us : lock_hrtimer_base.isra.18 <-hrtimer_try_to_cancel 2171 <idle>-0 3dN.1 16us : _raw_spin_lock_irqsave <-lock_hrtimer_base.isra.18 2172 <idle>-0 3dN.1 16us : add_preempt_count <-_raw_spin_lock_irqsave 2173 <idle>-0 3dN.2 17us : __remove_hrtimer <-remove_hrtimer.part.16 2174 <idle>-0 3dN.2 17us : hrtimer_force_reprogram <-__remove_hrtimer 2175 <idle>-0 3dN.2 17us : tick_program_event <-hrtimer_force_reprogram 2176 <idle>-0 3dN.2 18us : clockevents_program_event <-tick_program_event 2177 <idle>-0 3dN.2 18us : ktime_get <-clockevents_program_event 2178 <idle>-0 3dN.2 18us : lapic_next_event <-clockevents_program_event 2179 <idle>-0 3dN.2 19us : _raw_spin_unlock_irqrestore <-hrtimer_try_to_cancel 2180 <idle>-0 3dN.2 19us : sub_preempt_count <-_raw_spin_unlock_irqrestore 2181 <idle>-0 3dN.1 19us : hrtimer_forward <-tick_nohz_idle_exit 2182 <idle>-0 3dN.1 20us : ktime_add_safe <-hrtimer_forward 2183 <idle>-0 3dN.1 20us : ktime_add_safe <-hrtimer_forward 2184 <idle>-0 3dN.1 20us : hrtimer_start_range_ns <-hrtimer_start_expires.constprop.11 2185 <idle>-0 3dN.1 20us : __hrtimer_start_range_ns <-hrtimer_start_range_ns 2186 <idle>-0 3dN.1 21us : lock_hrtimer_base.isra.18 <-__hrtimer_start_range_ns 2187 <idle>-0 3dN.1 21us : _raw_spin_lock_irqsave <-lock_hrtimer_base.isra.18 2188 <idle>-0 3dN.1 21us : add_preempt_count <-_raw_spin_lock_irqsave 2189 <idle>-0 3dN.2 22us : ktime_add_safe <-__hrtimer_start_range_ns 2190 <idle>-0 3dN.2 22us : enqueue_hrtimer <-__hrtimer_start_range_ns 2191 <idle>-0 3dN.2 22us : tick_program_event <-__hrtimer_start_range_ns 2192 <idle>-0 3dN.2 23us : clockevents_program_event <-tick_program_event 2193 <idle>-0 3dN.2 23us : ktime_get <-clockevents_program_event 2194 <idle>-0 3dN.2 23us : lapic_next_event <-clockevents_program_event 2195 <idle>-0 3dN.2 24us : _raw_spin_unlock_irqrestore <-__hrtimer_start_range_ns 2196 <idle>-0 3dN.2 24us : sub_preempt_count <-_raw_spin_unlock_irqrestore 2197 <idle>-0 3dN.1 24us : account_idle_ticks <-tick_nohz_idle_exit 2198 <idle>-0 3dN.1 24us : account_idle_time <-account_idle_ticks 2199 <idle>-0 3.N.1 25us : sub_preempt_count <-cpu_idle 2200 <idle>-0 3.N.. 25us : schedule <-cpu_idle 2201 <idle>-0 3.N.. 25us : __schedule <-preempt_schedule 2202 <idle>-0 3.N.. 26us : add_preempt_count <-__schedule 2203 <idle>-0 3.N.1 26us : rcu_note_context_switch <-__schedule 2204 <idle>-0 3.N.1 26us : rcu_sched_qs <-rcu_note_context_switch 2205 <idle>-0 3dN.1 27us : rcu_preempt_qs <-rcu_note_context_switch 2206 <idle>-0 3.N.1 27us : _raw_spin_lock_irq <-__schedule 2207 <idle>-0 3dN.1 27us : add_preempt_count <-_raw_spin_lock_irq 2208 <idle>-0 3dN.2 28us : put_prev_task_idle <-__schedule 2209 <idle>-0 3dN.2 28us : pick_next_task_stop <-pick_next_task 2210 <idle>-0 3dN.2 28us : pick_next_task_rt <-pick_next_task 2211 <idle>-0 3dN.2 29us : dequeue_pushable_task <-pick_next_task_rt 2212 <idle>-0 3d..3 29us : __schedule <-preempt_schedule 2213 <idle>-0 3d..3 30us : 0:120:R ==> [003] 2448: 94:R sleep 2214 2215This isn't that big of a trace, even with function tracing enabled, 2216so I included the entire trace. 2217 2218The interrupt went off while when the system was idle. Somewhere 2219before task_woken_rt() was called, the NEED_RESCHED flag was set, 2220this is indicated by the first occurrence of the 'N' flag. 2221 2222Latency tracing and events 2223-------------------------- 2224As function tracing can induce a much larger latency, but without 2225seeing what happens within the latency it is hard to know what 2226caused it. There is a middle ground, and that is with enabling 2227events. 2228:: 2229 2230 # echo 0 > options/function-trace 2231 # echo wakeup_rt > current_tracer 2232 # echo 1 > events/enable 2233 # echo 1 > tracing_on 2234 # echo 0 > tracing_max_latency 2235 # chrt -f 5 sleep 1 2236 # echo 0 > tracing_on 2237 # cat trace 2238 # tracer: wakeup_rt 2239 # 2240 # wakeup_rt latency trace v1.1.5 on 3.8.0-test+ 2241 # -------------------------------------------------------------------- 2242 # latency: 6 us, #12/12, CPU#2 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4) 2243 # ----------------- 2244 # | task: sleep-5882 (uid:0 nice:0 policy:1 rt_prio:5) 2245 # ----------------- 2246 # 2247 # _------=> CPU# 2248 # / _-----=> irqs-off 2249 # | / _----=> need-resched 2250 # || / _---=> hardirq/softirq 2251 # ||| / _--=> preempt-depth 2252 # |||| / delay 2253 # cmd pid ||||| time | caller 2254 # \ / ||||| \ | / 2255 <idle>-0 2d.h4 0us : 0:120:R + [002] 5882: 94:R sleep 2256 <idle>-0 2d.h4 0us : ttwu_do_activate.constprop.87 <-try_to_wake_up 2257 <idle>-0 2d.h4 1us : sched_wakeup: comm=sleep pid=5882 prio=94 success=1 target_cpu=002 2258 <idle>-0 2dNh2 1us : hrtimer_expire_exit: hrtimer=ffff88007796feb8 2259 <idle>-0 2.N.2 2us : power_end: cpu_id=2 2260 <idle>-0 2.N.2 3us : cpu_idle: state=4294967295 cpu_id=2 2261 <idle>-0 2dN.3 4us : hrtimer_cancel: hrtimer=ffff88007d50d5e0 2262 <idle>-0 2dN.3 4us : hrtimer_start: hrtimer=ffff88007d50d5e0 function=tick_sched_timer expires=34311211000000 softexpires=34311211000000 2263 <idle>-0 2.N.2 5us : rcu_utilization: Start context switch 2264 <idle>-0 2.N.2 5us : rcu_utilization: End context switch 2265 <idle>-0 2d..3 6us : __schedule <-schedule 2266 <idle>-0 2d..3 6us : 0:120:R ==> [002] 5882: 94:R sleep 2267 2268 2269Hardware Latency Detector 2270------------------------- 2271 2272The hardware latency detector is executed by enabling the "hwlat" tracer. 2273 2274NOTE, this tracer will affect the performance of the system as it will 2275periodically make a CPU constantly busy with interrupts disabled. 2276:: 2277 2278 # echo hwlat > current_tracer 2279 # sleep 100 2280 # cat trace 2281 # tracer: hwlat 2282 # 2283 # entries-in-buffer/entries-written: 13/13 #P:8 2284 # 2285 # _-----=> irqs-off 2286 # / _----=> need-resched 2287 # | / _---=> hardirq/softirq 2288 # || / _--=> preempt-depth 2289 # ||| / delay 2290 # TASK-PID CPU# |||| TIMESTAMP FUNCTION 2291 # | | | |||| | | 2292 <...>-1729 [001] d... 678.473449: #1 inner/outer(us): 11/12 ts:1581527483.343962693 count:6 2293 <...>-1729 [004] d... 689.556542: #2 inner/outer(us): 16/9 ts:1581527494.889008092 count:1 2294 <...>-1729 [005] d... 714.756290: #3 inner/outer(us): 16/16 ts:1581527519.678961629 count:5 2295 <...>-1729 [001] d... 718.788247: #4 inner/outer(us): 9/17 ts:1581527523.889012713 count:1 2296 <...>-1729 [002] d... 719.796341: #5 inner/outer(us): 13/9 ts:1581527524.912872606 count:1 2297 <...>-1729 [006] d... 844.787091: #6 inner/outer(us): 9/12 ts:1581527649.889048502 count:2 2298 <...>-1729 [003] d... 849.827033: #7 inner/outer(us): 18/9 ts:1581527654.889013793 count:1 2299 <...>-1729 [007] d... 853.859002: #8 inner/outer(us): 9/12 ts:1581527658.889065736 count:1 2300 <...>-1729 [001] d... 855.874978: #9 inner/outer(us): 9/11 ts:1581527660.861991877 count:1 2301 <...>-1729 [001] d... 863.938932: #10 inner/outer(us): 9/11 ts:1581527668.970010500 count:1 nmi-total:7 nmi-count:1 2302 <...>-1729 [007] d... 878.050780: #11 inner/outer(us): 9/12 ts:1581527683.385002600 count:1 nmi-total:5 nmi-count:1 2303 <...>-1729 [007] d... 886.114702: #12 inner/outer(us): 9/12 ts:1581527691.385001600 count:1 2304 2305 2306The above output is somewhat the same in the header. All events will have 2307interrupts disabled 'd'. Under the FUNCTION title there is: 2308 2309 #1 2310 This is the count of events recorded that were greater than the 2311 tracing_threshold (See below). 2312 2313 inner/outer(us): 11/11 2314 2315 This shows two numbers as "inner latency" and "outer latency". The test 2316 runs in a loop checking a timestamp twice. The latency detected within 2317 the two timestamps is the "inner latency" and the latency detected 2318 after the previous timestamp and the next timestamp in the loop is 2319 the "outer latency". 2320 2321 ts:1581527483.343962693 2322 2323 The absolute timestamp that the first latency was recorded in the window. 2324 2325 count:6 2326 2327 The number of times a latency was detected during the window. 2328 2329 nmi-total:7 nmi-count:1 2330 2331 On architectures that support it, if an NMI comes in during the 2332 test, the time spent in NMI is reported in "nmi-total" (in 2333 microseconds). 2334 2335 All architectures that have NMIs will show the "nmi-count" if an 2336 NMI comes in during the test. 2337 2338hwlat files: 2339 2340 tracing_threshold 2341 This gets automatically set to "10" to represent 10 2342 microseconds. This is the threshold of latency that 2343 needs to be detected before the trace will be recorded. 2344 2345 Note, when hwlat tracer is finished (another tracer is 2346 written into "current_tracer"), the original value for 2347 tracing_threshold is placed back into this file. 2348 2349 hwlat_detector/width 2350 The length of time the test runs with interrupts disabled. 2351 2352 hwlat_detector/window 2353 The length of time of the window which the test 2354 runs. That is, the test will run for "width" 2355 microseconds per "window" microseconds 2356 2357 tracing_cpumask 2358 When the test is started. A kernel thread is created that 2359 runs the test. This thread will alternate between CPUs 2360 listed in the tracing_cpumask between each period 2361 (one "window"). To limit the test to specific CPUs 2362 set the mask in this file to only the CPUs that the test 2363 should run on. 2364 2365function 2366-------- 2367 2368This tracer is the function tracer. Enabling the function tracer 2369can be done from the debug file system. Make sure the 2370ftrace_enabled is set; otherwise this tracer is a nop. 2371See the "ftrace_enabled" section below. 2372:: 2373 2374 # sysctl kernel.ftrace_enabled=1 2375 # echo function > current_tracer 2376 # echo 1 > tracing_on 2377 # usleep 1 2378 # echo 0 > tracing_on 2379 # cat trace 2380 # tracer: function 2381 # 2382 # entries-in-buffer/entries-written: 24799/24799 #P:4 2383 # 2384 # _-----=> irqs-off 2385 # / _----=> need-resched 2386 # | / _---=> hardirq/softirq 2387 # || / _--=> preempt-depth 2388 # ||| / delay 2389 # TASK-PID CPU# |||| TIMESTAMP FUNCTION 2390 # | | | |||| | | 2391 bash-1994 [002] .... 3082.063030: mutex_unlock <-rb_simple_write 2392 bash-1994 [002] .... 3082.063031: __mutex_unlock_slowpath <-mutex_unlock 2393 bash-1994 [002] .... 3082.063031: __fsnotify_parent <-fsnotify_modify 2394 bash-1994 [002] .... 3082.063032: fsnotify <-fsnotify_modify 2395 bash-1994 [002] .... 3082.063032: __srcu_read_lock <-fsnotify 2396 bash-1994 [002] .... 3082.063032: add_preempt_count <-__srcu_read_lock 2397 bash-1994 [002] ...1 3082.063032: sub_preempt_count <-__srcu_read_lock 2398 bash-1994 [002] .... 3082.063033: __srcu_read_unlock <-fsnotify 2399 [...] 2400 2401 2402Note: function tracer uses ring buffers to store the above 2403entries. The newest data may overwrite the oldest data. 2404Sometimes using echo to stop the trace is not sufficient because 2405the tracing could have overwritten the data that you wanted to 2406record. For this reason, it is sometimes better to disable 2407tracing directly from a program. This allows you to stop the 2408tracing at the point that you hit the part that you are 2409interested in. To disable the tracing directly from a C program, 2410something like following code snippet can be used:: 2411 2412 int trace_fd; 2413 [...] 2414 int main(int argc, char *argv[]) { 2415 [...] 2416 trace_fd = open(tracing_file("tracing_on"), O_WRONLY); 2417 [...] 2418 if (condition_hit()) { 2419 write(trace_fd, "0", 1); 2420 } 2421 [...] 2422 } 2423 2424 2425Single thread tracing 2426--------------------- 2427 2428By writing into set_ftrace_pid you can trace a 2429single thread. For example:: 2430 2431 # cat set_ftrace_pid 2432 no pid 2433 # echo 3111 > set_ftrace_pid 2434 # cat set_ftrace_pid 2435 3111 2436 # echo function > current_tracer 2437 # cat trace | head 2438 # tracer: function 2439 # 2440 # TASK-PID CPU# TIMESTAMP FUNCTION 2441 # | | | | | 2442 yum-updatesd-3111 [003] 1637.254676: finish_task_switch <-thread_return 2443 yum-updatesd-3111 [003] 1637.254681: hrtimer_cancel <-schedule_hrtimeout_range 2444 yum-updatesd-3111 [003] 1637.254682: hrtimer_try_to_cancel <-hrtimer_cancel 2445 yum-updatesd-3111 [003] 1637.254683: lock_hrtimer_base <-hrtimer_try_to_cancel 2446 yum-updatesd-3111 [003] 1637.254685: fget_light <-do_sys_poll 2447 yum-updatesd-3111 [003] 1637.254686: pipe_poll <-do_sys_poll 2448 # echo > set_ftrace_pid 2449 # cat trace |head 2450 # tracer: function 2451 # 2452 # TASK-PID CPU# TIMESTAMP FUNCTION 2453 # | | | | | 2454 ##### CPU 3 buffer started #### 2455 yum-updatesd-3111 [003] 1701.957688: free_poll_entry <-poll_freewait 2456 yum-updatesd-3111 [003] 1701.957689: remove_wait_queue <-free_poll_entry 2457 yum-updatesd-3111 [003] 1701.957691: fput <-free_poll_entry 2458 yum-updatesd-3111 [003] 1701.957692: audit_syscall_exit <-sysret_audit 2459 yum-updatesd-3111 [003] 1701.957693: path_put <-audit_syscall_exit 2460 2461If you want to trace a function when executing, you could use 2462something like this simple program. 2463:: 2464 2465 #include <stdio.h> 2466 #include <stdlib.h> 2467 #include <sys/types.h> 2468 #include <sys/stat.h> 2469 #include <fcntl.h> 2470 #include <unistd.h> 2471 #include <string.h> 2472 2473 #define _STR(x) #x 2474 #define STR(x) _STR(x) 2475 #define MAX_PATH 256 2476 2477 const char *find_tracefs(void) 2478 { 2479 static char tracefs[MAX_PATH+1]; 2480 static int tracefs_found; 2481 char type[100]; 2482 FILE *fp; 2483 2484 if (tracefs_found) 2485 return tracefs; 2486 2487 if ((fp = fopen("/proc/mounts","r")) == NULL) { 2488 perror("/proc/mounts"); 2489 return NULL; 2490 } 2491 2492 while (fscanf(fp, "%*s %" 2493 STR(MAX_PATH) 2494 "s %99s %*s %*d %*d\n", 2495 tracefs, type) == 2) { 2496 if (strcmp(type, "tracefs") == 0) 2497 break; 2498 } 2499 fclose(fp); 2500 2501 if (strcmp(type, "tracefs") != 0) { 2502 fprintf(stderr, "tracefs not mounted"); 2503 return NULL; 2504 } 2505 2506 strcat(tracefs, "/tracing/"); 2507 tracefs_found = 1; 2508 2509 return tracefs; 2510 } 2511 2512 const char *tracing_file(const char *file_name) 2513 { 2514 static char trace_file[MAX_PATH+1]; 2515 snprintf(trace_file, MAX_PATH, "%s/%s", find_tracefs(), file_name); 2516 return trace_file; 2517 } 2518 2519 int main (int argc, char **argv) 2520 { 2521 if (argc < 1) 2522 exit(-1); 2523 2524 if (fork() > 0) { 2525 int fd, ffd; 2526 char line[64]; 2527 int s; 2528 2529 ffd = open(tracing_file("current_tracer"), O_WRONLY); 2530 if (ffd < 0) 2531 exit(-1); 2532 write(ffd, "nop", 3); 2533 2534 fd = open(tracing_file("set_ftrace_pid"), O_WRONLY); 2535 s = sprintf(line, "%d\n", getpid()); 2536 write(fd, line, s); 2537 2538 write(ffd, "function", 8); 2539 2540 close(fd); 2541 close(ffd); 2542 2543 execvp(argv[1], argv+1); 2544 } 2545 2546 return 0; 2547 } 2548 2549Or this simple script! 2550:: 2551 2552 #!/bin/bash 2553 2554 tracefs=`sed -ne 's/^tracefs \(.*\) tracefs.*/\1/p' /proc/mounts` 2555 echo 0 > $tracefs/tracing_on 2556 echo $$ > $tracefs/set_ftrace_pid 2557 echo function > $tracefs/current_tracer 2558 echo 1 > $tracefs/tracing_on 2559 exec "$@" 2560 2561 2562function graph tracer 2563--------------------------- 2564 2565This tracer is similar to the function tracer except that it 2566probes a function on its entry and its exit. This is done by 2567using a dynamically allocated stack of return addresses in each 2568task_struct. On function entry the tracer overwrites the return 2569address of each function traced to set a custom probe. Thus the 2570original return address is stored on the stack of return address 2571in the task_struct. 2572 2573Probing on both ends of a function leads to special features 2574such as: 2575 2576- measure of a function's time execution 2577- having a reliable call stack to draw function calls graph 2578 2579This tracer is useful in several situations: 2580 2581- you want to find the reason of a strange kernel behavior and 2582 need to see what happens in detail on any areas (or specific 2583 ones). 2584 2585- you are experiencing weird latencies but it's difficult to 2586 find its origin. 2587 2588- you want to find quickly which path is taken by a specific 2589 function 2590 2591- you just want to peek inside a working kernel and want to see 2592 what happens there. 2593 2594:: 2595 2596 # tracer: function_graph 2597 # 2598 # CPU DURATION FUNCTION CALLS 2599 # | | | | | | | 2600 2601 0) | sys_open() { 2602 0) | do_sys_open() { 2603 0) | getname() { 2604 0) | kmem_cache_alloc() { 2605 0) 1.382 us | __might_sleep(); 2606 0) 2.478 us | } 2607 0) | strncpy_from_user() { 2608 0) | might_fault() { 2609 0) 1.389 us | __might_sleep(); 2610 0) 2.553 us | } 2611 0) 3.807 us | } 2612 0) 7.876 us | } 2613 0) | alloc_fd() { 2614 0) 0.668 us | _spin_lock(); 2615 0) 0.570 us | expand_files(); 2616 0) 0.586 us | _spin_unlock(); 2617 2618 2619There are several columns that can be dynamically 2620enabled/disabled. You can use every combination of options you 2621want, depending on your needs. 2622 2623- The cpu number on which the function executed is default 2624 enabled. It is sometimes better to only trace one cpu (see 2625 tracing_cpumask file) or you might sometimes see unordered 2626 function calls while cpu tracing switch. 2627 2628 - hide: echo nofuncgraph-cpu > trace_options 2629 - show: echo funcgraph-cpu > trace_options 2630 2631- The duration (function's time of execution) is displayed on 2632 the closing bracket line of a function or on the same line 2633 than the current function in case of a leaf one. It is default 2634 enabled. 2635 2636 - hide: echo nofuncgraph-duration > trace_options 2637 - show: echo funcgraph-duration > trace_options 2638 2639- The overhead field precedes the duration field in case of 2640 reached duration thresholds. 2641 2642 - hide: echo nofuncgraph-overhead > trace_options 2643 - show: echo funcgraph-overhead > trace_options 2644 - depends on: funcgraph-duration 2645 2646 ie:: 2647 2648 3) # 1837.709 us | } /* __switch_to */ 2649 3) | finish_task_switch() { 2650 3) 0.313 us | _raw_spin_unlock_irq(); 2651 3) 3.177 us | } 2652 3) # 1889.063 us | } /* __schedule */ 2653 3) ! 140.417 us | } /* __schedule */ 2654 3) # 2034.948 us | } /* schedule */ 2655 3) * 33998.59 us | } /* schedule_preempt_disabled */ 2656 2657 [...] 2658 2659 1) 0.260 us | msecs_to_jiffies(); 2660 1) 0.313 us | __rcu_read_unlock(); 2661 1) + 61.770 us | } 2662 1) + 64.479 us | } 2663 1) 0.313 us | rcu_bh_qs(); 2664 1) 0.313 us | __local_bh_enable(); 2665 1) ! 217.240 us | } 2666 1) 0.365 us | idle_cpu(); 2667 1) | rcu_irq_exit() { 2668 1) 0.417 us | rcu_eqs_enter_common.isra.47(); 2669 1) 3.125 us | } 2670 1) ! 227.812 us | } 2671 1) ! 457.395 us | } 2672 1) @ 119760.2 us | } 2673 2674 [...] 2675 2676 2) | handle_IPI() { 2677 1) 6.979 us | } 2678 2) 0.417 us | scheduler_ipi(); 2679 1) 9.791 us | } 2680 1) + 12.917 us | } 2681 2) 3.490 us | } 2682 1) + 15.729 us | } 2683 1) + 18.542 us | } 2684 2) $ 3594274 us | } 2685 2686Flags:: 2687 2688 + means that the function exceeded 10 usecs. 2689 ! means that the function exceeded 100 usecs. 2690 # means that the function exceeded 1000 usecs. 2691 * means that the function exceeded 10 msecs. 2692 @ means that the function exceeded 100 msecs. 2693 $ means that the function exceeded 1 sec. 2694 2695 2696- The task/pid field displays the thread cmdline and pid which 2697 executed the function. It is default disabled. 2698 2699 - hide: echo nofuncgraph-proc > trace_options 2700 - show: echo funcgraph-proc > trace_options 2701 2702 ie:: 2703 2704 # tracer: function_graph 2705 # 2706 # CPU TASK/PID DURATION FUNCTION CALLS 2707 # | | | | | | | | | 2708 0) sh-4802 | | d_free() { 2709 0) sh-4802 | | call_rcu() { 2710 0) sh-4802 | | __call_rcu() { 2711 0) sh-4802 | 0.616 us | rcu_process_gp_end(); 2712 0) sh-4802 | 0.586 us | check_for_new_grace_period(); 2713 0) sh-4802 | 2.899 us | } 2714 0) sh-4802 | 4.040 us | } 2715 0) sh-4802 | 5.151 us | } 2716 0) sh-4802 | + 49.370 us | } 2717 2718 2719- The absolute time field is an absolute timestamp given by the 2720 system clock since it started. A snapshot of this time is 2721 given on each entry/exit of functions 2722 2723 - hide: echo nofuncgraph-abstime > trace_options 2724 - show: echo funcgraph-abstime > trace_options 2725 2726 ie:: 2727 2728 # 2729 # TIME CPU DURATION FUNCTION CALLS 2730 # | | | | | | | | 2731 360.774522 | 1) 0.541 us | } 2732 360.774522 | 1) 4.663 us | } 2733 360.774523 | 1) 0.541 us | __wake_up_bit(); 2734 360.774524 | 1) 6.796 us | } 2735 360.774524 | 1) 7.952 us | } 2736 360.774525 | 1) 9.063 us | } 2737 360.774525 | 1) 0.615 us | journal_mark_dirty(); 2738 360.774527 | 1) 0.578 us | __brelse(); 2739 360.774528 | 1) | reiserfs_prepare_for_journal() { 2740 360.774528 | 1) | unlock_buffer() { 2741 360.774529 | 1) | wake_up_bit() { 2742 360.774529 | 1) | bit_waitqueue() { 2743 360.774530 | 1) 0.594 us | __phys_addr(); 2744 2745 2746The function name is always displayed after the closing bracket 2747for a function if the start of that function is not in the 2748trace buffer. 2749 2750Display of the function name after the closing bracket may be 2751enabled for functions whose start is in the trace buffer, 2752allowing easier searching with grep for function durations. 2753It is default disabled. 2754 2755 - hide: echo nofuncgraph-tail > trace_options 2756 - show: echo funcgraph-tail > trace_options 2757 2758 Example with nofuncgraph-tail (default):: 2759 2760 0) | putname() { 2761 0) | kmem_cache_free() { 2762 0) 0.518 us | __phys_addr(); 2763 0) 1.757 us | } 2764 0) 2.861 us | } 2765 2766 Example with funcgraph-tail:: 2767 2768 0) | putname() { 2769 0) | kmem_cache_free() { 2770 0) 0.518 us | __phys_addr(); 2771 0) 1.757 us | } /* kmem_cache_free() */ 2772 0) 2.861 us | } /* putname() */ 2773 2774The return value of each traced function can be displayed after 2775an equal sign "=". When encountering system call failures, it 2776can be very helpful to quickly locate the function that first 2777returns an error code. 2778 2779 - hide: echo nofuncgraph-retval > trace_options 2780 - show: echo funcgraph-retval > trace_options 2781 2782 Example with funcgraph-retval:: 2783 2784 1) | cgroup_migrate() { 2785 1) 0.651 us | cgroup_migrate_add_task(); /* = 0xffff93fcfd346c00 */ 2786 1) | cgroup_migrate_execute() { 2787 1) | cpu_cgroup_can_attach() { 2788 1) | cgroup_taskset_first() { 2789 1) 0.732 us | cgroup_taskset_next(); /* = 0xffff93fc8fb20000 */ 2790 1) 1.232 us | } /* cgroup_taskset_first = 0xffff93fc8fb20000 */ 2791 1) 0.380 us | sched_rt_can_attach(); /* = 0x0 */ 2792 1) 2.335 us | } /* cpu_cgroup_can_attach = -22 */ 2793 1) 4.369 us | } /* cgroup_migrate_execute = -22 */ 2794 1) 7.143 us | } /* cgroup_migrate = -22 */ 2795 2796The above example shows that the function cpu_cgroup_can_attach 2797returned the error code -22 firstly, then we can read the code 2798of this function to get the root cause. 2799 2800When the option funcgraph-retval-hex is not set, the return value can 2801be displayed in a smart way. Specifically, if it is an error code, 2802it will be printed in signed decimal format, otherwise it will 2803printed in hexadecimal format. 2804 2805 - smart: echo nofuncgraph-retval-hex > trace_options 2806 - hexadecimal: echo funcgraph-retval-hex > trace_options 2807 2808 Example with funcgraph-retval-hex:: 2809 2810 1) | cgroup_migrate() { 2811 1) 0.651 us | cgroup_migrate_add_task(); /* = 0xffff93fcfd346c00 */ 2812 1) | cgroup_migrate_execute() { 2813 1) | cpu_cgroup_can_attach() { 2814 1) | cgroup_taskset_first() { 2815 1) 0.732 us | cgroup_taskset_next(); /* = 0xffff93fc8fb20000 */ 2816 1) 1.232 us | } /* cgroup_taskset_first = 0xffff93fc8fb20000 */ 2817 1) 0.380 us | sched_rt_can_attach(); /* = 0x0 */ 2818 1) 2.335 us | } /* cpu_cgroup_can_attach = 0xffffffea */ 2819 1) 4.369 us | } /* cgroup_migrate_execute = 0xffffffea */ 2820 1) 7.143 us | } /* cgroup_migrate = 0xffffffea */ 2821 2822At present, there are some limitations when using the funcgraph-retval 2823option, and these limitations will be eliminated in the future: 2824 2825- Even if the function return type is void, a return value will still 2826 be printed, and you can just ignore it. 2827 2828- Even if return values are stored in multiple registers, only the 2829 value contained in the first register will be recorded and printed. 2830 To illustrate, in the x86 architecture, eax and edx are used to store 2831 a 64-bit return value, with the lower 32 bits saved in eax and the 2832 upper 32 bits saved in edx. However, only the value stored in eax 2833 will be recorded and printed. 2834 2835- In certain procedure call standards, such as arm64's AAPCS64, when a 2836 type is smaller than a GPR, it is the responsibility of the consumer 2837 to perform the narrowing, and the upper bits may contain UNKNOWN values. 2838 Therefore, it is advisable to check the code for such cases. For instance, 2839 when using a u8 in a 64-bit GPR, bits [63:8] may contain arbitrary values, 2840 especially when larger types are truncated, whether explicitly or implicitly. 2841 Here are some specific cases to illustrate this point: 2842 2843 **Case One**: 2844 2845 The function narrow_to_u8 is defined as follows:: 2846 2847 u8 narrow_to_u8(u64 val) 2848 { 2849 // implicitly truncated 2850 return val; 2851 } 2852 2853 It may be compiled to:: 2854 2855 narrow_to_u8: 2856 < ... ftrace instrumentation ... > 2857 RET 2858 2859 If you pass 0x123456789abcdef to this function and want to narrow it, 2860 it may be recorded as 0x123456789abcdef instead of 0xef. 2861 2862 **Case Two**: 2863 2864 The function error_if_not_4g_aligned is defined as follows:: 2865 2866 int error_if_not_4g_aligned(u64 val) 2867 { 2868 if (val & GENMASK(31, 0)) 2869 return -EINVAL; 2870 2871 return 0; 2872 } 2873 2874 It could be compiled to:: 2875 2876 error_if_not_4g_aligned: 2877 CBNZ w0, .Lnot_aligned 2878 RET // bits [31:0] are zero, bits 2879 // [63:32] are UNKNOWN 2880 .Lnot_aligned: 2881 MOV x0, #-EINVAL 2882 RET 2883 2884 When passing 0x2_0000_0000 to it, the return value may be recorded as 2885 0x2_0000_0000 instead of 0. 2886 2887You can put some comments on specific functions by using 2888trace_printk() For example, if you want to put a comment inside 2889the __might_sleep() function, you just have to include 2890<linux/ftrace.h> and call trace_printk() inside __might_sleep():: 2891 2892 trace_printk("I'm a comment!\n") 2893 2894will produce:: 2895 2896 1) | __might_sleep() { 2897 1) | /* I'm a comment! */ 2898 1) 1.449 us | } 2899 2900 2901You might find other useful features for this tracer in the 2902following "dynamic ftrace" section such as tracing only specific 2903functions or tasks. 2904 2905dynamic ftrace 2906-------------- 2907 2908If CONFIG_DYNAMIC_FTRACE is set, the system will run with 2909virtually no overhead when function tracing is disabled. The way 2910this works is the mcount function call (placed at the start of 2911every kernel function, produced by the -pg switch in gcc), 2912starts of pointing to a simple return. (Enabling FTRACE will 2913include the -pg switch in the compiling of the kernel.) 2914 2915At compile time every C file object is run through the 2916recordmcount program (located in the scripts directory). This 2917program will parse the ELF headers in the C object to find all 2918the locations in the .text section that call mcount. Starting 2919with gcc version 4.6, the -mfentry has been added for x86, which 2920calls "__fentry__" instead of "mcount". Which is called before 2921the creation of the stack frame. 2922 2923Note, not all sections are traced. They may be prevented by either 2924a notrace, or blocked another way and all inline functions are not 2925traced. Check the "available_filter_functions" file to see what functions 2926can be traced. 2927 2928A section called "__mcount_loc" is created that holds 2929references to all the mcount/fentry call sites in the .text section. 2930The recordmcount program re-links this section back into the 2931original object. The final linking stage of the kernel will add all these 2932references into a single table. 2933 2934On boot up, before SMP is initialized, the dynamic ftrace code 2935scans this table and updates all the locations into nops. It 2936also records the locations, which are added to the 2937available_filter_functions list. Modules are processed as they 2938are loaded and before they are executed. When a module is 2939unloaded, it also removes its functions from the ftrace function 2940list. This is automatic in the module unload code, and the 2941module author does not need to worry about it. 2942 2943When tracing is enabled, the process of modifying the function 2944tracepoints is dependent on architecture. The old method is to use 2945kstop_machine to prevent races with the CPUs executing code being 2946modified (which can cause the CPU to do undesirable things, especially 2947if the modified code crosses cache (or page) boundaries), and the nops are 2948patched back to calls. But this time, they do not call mcount 2949(which is just a function stub). They now call into the ftrace 2950infrastructure. 2951 2952The new method of modifying the function tracepoints is to place 2953a breakpoint at the location to be modified, sync all CPUs, modify 2954the rest of the instruction not covered by the breakpoint. Sync 2955all CPUs again, and then remove the breakpoint with the finished 2956version to the ftrace call site. 2957 2958Some archs do not even need to monkey around with the synchronization, 2959and can just slap the new code on top of the old without any 2960problems with other CPUs executing it at the same time. 2961 2962One special side-effect to the recording of the functions being 2963traced is that we can now selectively choose which functions we 2964wish to trace and which ones we want the mcount calls to remain 2965as nops. 2966 2967Two files are used, one for enabling and one for disabling the 2968tracing of specified functions. They are: 2969 2970 set_ftrace_filter 2971 2972and 2973 2974 set_ftrace_notrace 2975 2976A list of available functions that you can add to these files is 2977listed in: 2978 2979 available_filter_functions 2980 2981:: 2982 2983 # cat available_filter_functions 2984 put_prev_task_idle 2985 kmem_cache_create 2986 pick_next_task_rt 2987 cpus_read_lock 2988 pick_next_task_fair 2989 mutex_lock 2990 [...] 2991 2992If I am only interested in sys_nanosleep and hrtimer_interrupt:: 2993 2994 # echo sys_nanosleep hrtimer_interrupt > set_ftrace_filter 2995 # echo function > current_tracer 2996 # echo 1 > tracing_on 2997 # usleep 1 2998 # echo 0 > tracing_on 2999 # cat trace 3000 # tracer: function 3001 # 3002 # entries-in-buffer/entries-written: 5/5 #P:4 3003 # 3004 # _-----=> irqs-off 3005 # / _----=> need-resched 3006 # | / _---=> hardirq/softirq 3007 # || / _--=> preempt-depth 3008 # ||| / delay 3009 # TASK-PID CPU# |||| TIMESTAMP FUNCTION 3010 # | | | |||| | | 3011 usleep-2665 [001] .... 4186.475355: sys_nanosleep <-system_call_fastpath 3012 <idle>-0 [001] d.h1 4186.475409: hrtimer_interrupt <-smp_apic_timer_interrupt 3013 usleep-2665 [001] d.h1 4186.475426: hrtimer_interrupt <-smp_apic_timer_interrupt 3014 <idle>-0 [003] d.h1 4186.475426: hrtimer_interrupt <-smp_apic_timer_interrupt 3015 <idle>-0 [002] d.h1 4186.475427: hrtimer_interrupt <-smp_apic_timer_interrupt 3016 3017To see which functions are being traced, you can cat the file: 3018:: 3019 3020 # cat set_ftrace_filter 3021 hrtimer_interrupt 3022 sys_nanosleep 3023 3024 3025Perhaps this is not enough. The filters also allow glob(7) matching. 3026 3027 ``<match>*`` 3028 will match functions that begin with <match> 3029 ``*<match>`` 3030 will match functions that end with <match> 3031 ``*<match>*`` 3032 will match functions that have <match> in it 3033 ``<match1>*<match2>`` 3034 will match functions that begin with <match1> and end with <match2> 3035 3036.. note:: 3037 It is better to use quotes to enclose the wild cards, 3038 otherwise the shell may expand the parameters into names 3039 of files in the local directory. 3040 3041:: 3042 3043 # echo 'hrtimer_*' > set_ftrace_filter 3044 3045Produces:: 3046 3047 # tracer: function 3048 # 3049 # entries-in-buffer/entries-written: 897/897 #P:4 3050 # 3051 # _-----=> irqs-off 3052 # / _----=> need-resched 3053 # | / _---=> hardirq/softirq 3054 # || / _--=> preempt-depth 3055 # ||| / delay 3056 # TASK-PID CPU# |||| TIMESTAMP FUNCTION 3057 # | | | |||| | | 3058 <idle>-0 [003] dN.1 4228.547803: hrtimer_cancel <-tick_nohz_idle_exit 3059 <idle>-0 [003] dN.1 4228.547804: hrtimer_try_to_cancel <-hrtimer_cancel 3060 <idle>-0 [003] dN.2 4228.547805: hrtimer_force_reprogram <-__remove_hrtimer 3061 <idle>-0 [003] dN.1 4228.547805: hrtimer_forward <-tick_nohz_idle_exit 3062 <idle>-0 [003] dN.1 4228.547805: hrtimer_start_range_ns <-hrtimer_start_expires.constprop.11 3063 <idle>-0 [003] d..1 4228.547858: hrtimer_get_next_event <-get_next_timer_interrupt 3064 <idle>-0 [003] d..1 4228.547859: hrtimer_start <-__tick_nohz_idle_enter 3065 <idle>-0 [003] d..2 4228.547860: hrtimer_force_reprogram <-__rem 3066 3067Notice that we lost the sys_nanosleep. 3068:: 3069 3070 # cat set_ftrace_filter 3071 hrtimer_run_queues 3072 hrtimer_run_pending 3073 hrtimer_init 3074 hrtimer_cancel 3075 hrtimer_try_to_cancel 3076 hrtimer_forward 3077 hrtimer_start 3078 hrtimer_reprogram 3079 hrtimer_force_reprogram 3080 hrtimer_get_next_event 3081 hrtimer_interrupt 3082 hrtimer_nanosleep 3083 hrtimer_wakeup 3084 hrtimer_get_remaining 3085 hrtimer_get_res 3086 hrtimer_init_sleeper 3087 3088 3089This is because the '>' and '>>' act just like they do in bash. 3090To rewrite the filters, use '>' 3091To append to the filters, use '>>' 3092 3093To clear out a filter so that all functions will be recorded 3094again:: 3095 3096 # echo > set_ftrace_filter 3097 # cat set_ftrace_filter 3098 # 3099 3100Again, now we want to append. 3101 3102:: 3103 3104 # echo sys_nanosleep > set_ftrace_filter 3105 # cat set_ftrace_filter 3106 sys_nanosleep 3107 # echo 'hrtimer_*' >> set_ftrace_filter 3108 # cat set_ftrace_filter 3109 hrtimer_run_queues 3110 hrtimer_run_pending 3111 hrtimer_init 3112 hrtimer_cancel 3113 hrtimer_try_to_cancel 3114 hrtimer_forward 3115 hrtimer_start 3116 hrtimer_reprogram 3117 hrtimer_force_reprogram 3118 hrtimer_get_next_event 3119 hrtimer_interrupt 3120 sys_nanosleep 3121 hrtimer_nanosleep 3122 hrtimer_wakeup 3123 hrtimer_get_remaining 3124 hrtimer_get_res 3125 hrtimer_init_sleeper 3126 3127 3128The set_ftrace_notrace prevents those functions from being 3129traced. 3130:: 3131 3132 # echo '*preempt*' '*lock*' > set_ftrace_notrace 3133 3134Produces:: 3135 3136 # tracer: function 3137 # 3138 # entries-in-buffer/entries-written: 39608/39608 #P:4 3139 # 3140 # _-----=> irqs-off 3141 # / _----=> need-resched 3142 # | / _---=> hardirq/softirq 3143 # || / _--=> preempt-depth 3144 # ||| / delay 3145 # TASK-PID CPU# |||| TIMESTAMP FUNCTION 3146 # | | | |||| | | 3147 bash-1994 [000] .... 4342.324896: file_ra_state_init <-do_dentry_open 3148 bash-1994 [000] .... 4342.324897: open_check_o_direct <-do_last 3149 bash-1994 [000] .... 4342.324897: ima_file_check <-do_last 3150 bash-1994 [000] .... 4342.324898: process_measurement <-ima_file_check 3151 bash-1994 [000] .... 4342.324898: ima_get_action <-process_measurement 3152 bash-1994 [000] .... 4342.324898: ima_match_policy <-ima_get_action 3153 bash-1994 [000] .... 4342.324899: do_truncate <-do_last 3154 bash-1994 [000] .... 4342.324899: setattr_should_drop_suidgid <-do_truncate 3155 bash-1994 [000] .... 4342.324899: notify_change <-do_truncate 3156 bash-1994 [000] .... 4342.324900: current_fs_time <-notify_change 3157 bash-1994 [000] .... 4342.324900: current_kernel_time <-current_fs_time 3158 bash-1994 [000] .... 4342.324900: timespec_trunc <-current_fs_time 3159 3160We can see that there's no more lock or preempt tracing. 3161 3162Selecting function filters via index 3163------------------------------------ 3164 3165Because processing of strings is expensive (the address of the function 3166needs to be looked up before comparing to the string being passed in), 3167an index can be used as well to enable functions. This is useful in the 3168case of setting thousands of specific functions at a time. By passing 3169in a list of numbers, no string processing will occur. Instead, the function 3170at the specific location in the internal array (which corresponds to the 3171functions in the "available_filter_functions" file), is selected. 3172 3173:: 3174 3175 # echo 1 > set_ftrace_filter 3176 3177Will select the first function listed in "available_filter_functions" 3178 3179:: 3180 3181 # head -1 available_filter_functions 3182 trace_initcall_finish_cb 3183 3184 # cat set_ftrace_filter 3185 trace_initcall_finish_cb 3186 3187 # head -50 available_filter_functions | tail -1 3188 x86_pmu_commit_txn 3189 3190 # echo 1 50 > set_ftrace_filter 3191 # cat set_ftrace_filter 3192 trace_initcall_finish_cb 3193 x86_pmu_commit_txn 3194 3195Dynamic ftrace with the function graph tracer 3196--------------------------------------------- 3197 3198Although what has been explained above concerns both the 3199function tracer and the function-graph-tracer, there are some 3200special features only available in the function-graph tracer. 3201 3202If you want to trace only one function and all of its children, 3203you just have to echo its name into set_graph_function:: 3204 3205 echo __do_fault > set_graph_function 3206 3207will produce the following "expanded" trace of the __do_fault() 3208function:: 3209 3210 0) | __do_fault() { 3211 0) | filemap_fault() { 3212 0) | find_lock_page() { 3213 0) 0.804 us | find_get_page(); 3214 0) | __might_sleep() { 3215 0) 1.329 us | } 3216 0) 3.904 us | } 3217 0) 4.979 us | } 3218 0) 0.653 us | _spin_lock(); 3219 0) 0.578 us | page_add_file_rmap(); 3220 0) 0.525 us | native_set_pte_at(); 3221 0) 0.585 us | _spin_unlock(); 3222 0) | unlock_page() { 3223 0) 0.541 us | page_waitqueue(); 3224 0) 0.639 us | __wake_up_bit(); 3225 0) 2.786 us | } 3226 0) + 14.237 us | } 3227 0) | __do_fault() { 3228 0) | filemap_fault() { 3229 0) | find_lock_page() { 3230 0) 0.698 us | find_get_page(); 3231 0) | __might_sleep() { 3232 0) 1.412 us | } 3233 0) 3.950 us | } 3234 0) 5.098 us | } 3235 0) 0.631 us | _spin_lock(); 3236 0) 0.571 us | page_add_file_rmap(); 3237 0) 0.526 us | native_set_pte_at(); 3238 0) 0.586 us | _spin_unlock(); 3239 0) | unlock_page() { 3240 0) 0.533 us | page_waitqueue(); 3241 0) 0.638 us | __wake_up_bit(); 3242 0) 2.793 us | } 3243 0) + 14.012 us | } 3244 3245You can also expand several functions at once:: 3246 3247 echo sys_open > set_graph_function 3248 echo sys_close >> set_graph_function 3249 3250Now if you want to go back to trace all functions you can clear 3251this special filter via:: 3252 3253 echo > set_graph_function 3254 3255 3256ftrace_enabled 3257-------------- 3258 3259Note, the proc sysctl ftrace_enable is a big on/off switch for the 3260function tracer. By default it is enabled (when function tracing is 3261enabled in the kernel). If it is disabled, all function tracing is 3262disabled. This includes not only the function tracers for ftrace, but 3263also for any other uses (perf, kprobes, stack tracing, profiling, etc). It 3264cannot be disabled if there is a callback with FTRACE_OPS_FL_PERMANENT set 3265registered. 3266 3267Please disable this with care. 3268 3269This can be disable (and enabled) with:: 3270 3271 sysctl kernel.ftrace_enabled=0 3272 sysctl kernel.ftrace_enabled=1 3273 3274 or 3275 3276 echo 0 > /proc/sys/kernel/ftrace_enabled 3277 echo 1 > /proc/sys/kernel/ftrace_enabled 3278 3279 3280Filter commands 3281--------------- 3282 3283A few commands are supported by the set_ftrace_filter interface. 3284Trace commands have the following format:: 3285 3286 <function>:<command>:<parameter> 3287 3288The following commands are supported: 3289 3290- mod: 3291 This command enables function filtering per module. The 3292 parameter defines the module. For example, if only the write* 3293 functions in the ext3 module are desired, run: 3294 3295 echo 'write*:mod:ext3' > set_ftrace_filter 3296 3297 This command interacts with the filter in the same way as 3298 filtering based on function names. Thus, adding more functions 3299 in a different module is accomplished by appending (>>) to the 3300 filter file. Remove specific module functions by prepending 3301 '!':: 3302 3303 echo '!writeback*:mod:ext3' >> set_ftrace_filter 3304 3305 Mod command supports module globbing. Disable tracing for all 3306 functions except a specific module:: 3307 3308 echo '!*:mod:!ext3' >> set_ftrace_filter 3309 3310 Disable tracing for all modules, but still trace kernel:: 3311 3312 echo '!*:mod:*' >> set_ftrace_filter 3313 3314 Enable filter only for kernel:: 3315 3316 echo '*write*:mod:!*' >> set_ftrace_filter 3317 3318 Enable filter for module globbing:: 3319 3320 echo '*write*:mod:*snd*' >> set_ftrace_filter 3321 3322- traceon/traceoff: 3323 These commands turn tracing on and off when the specified 3324 functions are hit. The parameter determines how many times the 3325 tracing system is turned on and off. If unspecified, there is 3326 no limit. For example, to disable tracing when a schedule bug 3327 is hit the first 5 times, run:: 3328 3329 echo '__schedule_bug:traceoff:5' > set_ftrace_filter 3330 3331 To always disable tracing when __schedule_bug is hit:: 3332 3333 echo '__schedule_bug:traceoff' > set_ftrace_filter 3334 3335 These commands are cumulative whether or not they are appended 3336 to set_ftrace_filter. To remove a command, prepend it by '!' 3337 and drop the parameter:: 3338 3339 echo '!__schedule_bug:traceoff:0' > set_ftrace_filter 3340 3341 The above removes the traceoff command for __schedule_bug 3342 that have a counter. To remove commands without counters:: 3343 3344 echo '!__schedule_bug:traceoff' > set_ftrace_filter 3345 3346- snapshot: 3347 Will cause a snapshot to be triggered when the function is hit. 3348 :: 3349 3350 echo 'native_flush_tlb_others:snapshot' > set_ftrace_filter 3351 3352 To only snapshot once: 3353 :: 3354 3355 echo 'native_flush_tlb_others:snapshot:1' > set_ftrace_filter 3356 3357 To remove the above commands:: 3358 3359 echo '!native_flush_tlb_others:snapshot' > set_ftrace_filter 3360 echo '!native_flush_tlb_others:snapshot:0' > set_ftrace_filter 3361 3362- enable_event/disable_event: 3363 These commands can enable or disable a trace event. Note, because 3364 function tracing callbacks are very sensitive, when these commands 3365 are registered, the trace point is activated, but disabled in 3366 a "soft" mode. That is, the tracepoint will be called, but 3367 just will not be traced. The event tracepoint stays in this mode 3368 as long as there's a command that triggers it. 3369 :: 3370 3371 echo 'try_to_wake_up:enable_event:sched:sched_switch:2' > \ 3372 set_ftrace_filter 3373 3374 The format is:: 3375 3376 <function>:enable_event:<system>:<event>[:count] 3377 <function>:disable_event:<system>:<event>[:count] 3378 3379 To remove the events commands:: 3380 3381 echo '!try_to_wake_up:enable_event:sched:sched_switch:0' > \ 3382 set_ftrace_filter 3383 echo '!schedule:disable_event:sched:sched_switch' > \ 3384 set_ftrace_filter 3385 3386- dump: 3387 When the function is hit, it will dump the contents of the ftrace 3388 ring buffer to the console. This is useful if you need to debug 3389 something, and want to dump the trace when a certain function 3390 is hit. Perhaps it's a function that is called before a triple 3391 fault happens and does not allow you to get a regular dump. 3392 3393- cpudump: 3394 When the function is hit, it will dump the contents of the ftrace 3395 ring buffer for the current CPU to the console. Unlike the "dump" 3396 command, it only prints out the contents of the ring buffer for the 3397 CPU that executed the function that triggered the dump. 3398 3399- stacktrace: 3400 When the function is hit, a stack trace is recorded. 3401 3402trace_pipe 3403---------- 3404 3405The trace_pipe outputs the same content as the trace file, but 3406the effect on the tracing is different. Every read from 3407trace_pipe is consumed. This means that subsequent reads will be 3408different. The trace is live. 3409:: 3410 3411 # echo function > current_tracer 3412 # cat trace_pipe > /tmp/trace.out & 3413 [1] 4153 3414 # echo 1 > tracing_on 3415 # usleep 1 3416 # echo 0 > tracing_on 3417 # cat trace 3418 # tracer: function 3419 # 3420 # entries-in-buffer/entries-written: 0/0 #P:4 3421 # 3422 # _-----=> irqs-off 3423 # / _----=> need-resched 3424 # | / _---=> hardirq/softirq 3425 # || / _--=> preempt-depth 3426 # ||| / delay 3427 # TASK-PID CPU# |||| TIMESTAMP FUNCTION 3428 # | | | |||| | | 3429 3430 # 3431 # cat /tmp/trace.out 3432 bash-1994 [000] .... 5281.568961: mutex_unlock <-rb_simple_write 3433 bash-1994 [000] .... 5281.568963: __mutex_unlock_slowpath <-mutex_unlock 3434 bash-1994 [000] .... 5281.568963: __fsnotify_parent <-fsnotify_modify 3435 bash-1994 [000] .... 5281.568964: fsnotify <-fsnotify_modify 3436 bash-1994 [000] .... 5281.568964: __srcu_read_lock <-fsnotify 3437 bash-1994 [000] .... 5281.568964: add_preempt_count <-__srcu_read_lock 3438 bash-1994 [000] ...1 5281.568965: sub_preempt_count <-__srcu_read_lock 3439 bash-1994 [000] .... 5281.568965: __srcu_read_unlock <-fsnotify 3440 bash-1994 [000] .... 5281.568967: sys_dup2 <-system_call_fastpath 3441 3442 3443Note, reading the trace_pipe file will block until more input is 3444added. This is contrary to the trace file. If any process opened 3445the trace file for reading, it will actually disable tracing and 3446prevent new entries from being added. The trace_pipe file does 3447not have this limitation. 3448 3449trace entries 3450------------- 3451 3452Having too much or not enough data can be troublesome in 3453diagnosing an issue in the kernel. The file buffer_size_kb is 3454used to modify the size of the internal trace buffers. The 3455number listed is the number of entries that can be recorded per 3456CPU. To know the full size, multiply the number of possible CPUs 3457with the number of entries. 3458:: 3459 3460 # cat buffer_size_kb 3461 1408 (units kilobytes) 3462 3463Or simply read buffer_total_size_kb 3464:: 3465 3466 # cat buffer_total_size_kb 3467 5632 3468 3469To modify the buffer, simple echo in a number (in 1024 byte segments). 3470:: 3471 3472 # echo 10000 > buffer_size_kb 3473 # cat buffer_size_kb 3474 10000 (units kilobytes) 3475 3476It will try to allocate as much as possible. If you allocate too 3477much, it can cause Out-Of-Memory to trigger. 3478:: 3479 3480 # echo 1000000000000 > buffer_size_kb 3481 -bash: echo: write error: Cannot allocate memory 3482 # cat buffer_size_kb 3483 85 3484 3485The per_cpu buffers can be changed individually as well: 3486:: 3487 3488 # echo 10000 > per_cpu/cpu0/buffer_size_kb 3489 # echo 100 > per_cpu/cpu1/buffer_size_kb 3490 3491When the per_cpu buffers are not the same, the buffer_size_kb 3492at the top level will just show an X 3493:: 3494 3495 # cat buffer_size_kb 3496 X 3497 3498This is where the buffer_total_size_kb is useful: 3499:: 3500 3501 # cat buffer_total_size_kb 3502 12916 3503 3504Writing to the top level buffer_size_kb will reset all the buffers 3505to be the same again. 3506 3507Snapshot 3508-------- 3509CONFIG_TRACER_SNAPSHOT makes a generic snapshot feature 3510available to all non latency tracers. (Latency tracers which 3511record max latency, such as "irqsoff" or "wakeup", can't use 3512this feature, since those are already using the snapshot 3513mechanism internally.) 3514 3515Snapshot preserves a current trace buffer at a particular point 3516in time without stopping tracing. Ftrace swaps the current 3517buffer with a spare buffer, and tracing continues in the new 3518current (=previous spare) buffer. 3519 3520The following tracefs files in "tracing" are related to this 3521feature: 3522 3523 snapshot: 3524 3525 This is used to take a snapshot and to read the output 3526 of the snapshot. Echo 1 into this file to allocate a 3527 spare buffer and to take a snapshot (swap), then read 3528 the snapshot from this file in the same format as 3529 "trace" (described above in the section "The File 3530 System"). Both reads snapshot and tracing are executable 3531 in parallel. When the spare buffer is allocated, echoing 3532 0 frees it, and echoing else (positive) values clear the 3533 snapshot contents. 3534 More details are shown in the table below. 3535 3536 +--------------+------------+------------+------------+ 3537 |status\\input | 0 | 1 | else | 3538 +==============+============+============+============+ 3539 |not allocated |(do nothing)| alloc+swap |(do nothing)| 3540 +--------------+------------+------------+------------+ 3541 |allocated | free | swap | clear | 3542 +--------------+------------+------------+------------+ 3543 3544Here is an example of using the snapshot feature. 3545:: 3546 3547 # echo 1 > events/sched/enable 3548 # echo 1 > snapshot 3549 # cat snapshot 3550 # tracer: nop 3551 # 3552 # entries-in-buffer/entries-written: 71/71 #P:8 3553 # 3554 # _-----=> irqs-off 3555 # / _----=> need-resched 3556 # | / _---=> hardirq/softirq 3557 # || / _--=> preempt-depth 3558 # ||| / delay 3559 # TASK-PID CPU# |||| TIMESTAMP FUNCTION 3560 # | | | |||| | | 3561 <idle>-0 [005] d... 2440.603828: sched_switch: prev_comm=swapper/5 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=snapshot-test-2 next_pid=2242 next_prio=120 3562 sleep-2242 [005] d... 2440.603846: sched_switch: prev_comm=snapshot-test-2 prev_pid=2242 prev_prio=120 prev_state=R ==> next_comm=kworker/5:1 next_pid=60 next_prio=120 3563 [...] 3564 <idle>-0 [002] d... 2440.707230: sched_switch: prev_comm=swapper/2 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=snapshot-test-2 next_pid=2229 next_prio=120 3565 3566 # cat trace 3567 # tracer: nop 3568 # 3569 # entries-in-buffer/entries-written: 77/77 #P:8 3570 # 3571 # _-----=> irqs-off 3572 # / _----=> need-resched 3573 # | / _---=> hardirq/softirq 3574 # || / _--=> preempt-depth 3575 # ||| / delay 3576 # TASK-PID CPU# |||| TIMESTAMP FUNCTION 3577 # | | | |||| | | 3578 <idle>-0 [007] d... 2440.707395: sched_switch: prev_comm=swapper/7 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=snapshot-test-2 next_pid=2243 next_prio=120 3579 snapshot-test-2-2229 [002] d... 2440.707438: sched_switch: prev_comm=snapshot-test-2 prev_pid=2229 prev_prio=120 prev_state=S ==> next_comm=swapper/2 next_pid=0 next_prio=120 3580 [...] 3581 3582 3583If you try to use this snapshot feature when current tracer is 3584one of the latency tracers, you will get the following results. 3585:: 3586 3587 # echo wakeup > current_tracer 3588 # echo 1 > snapshot 3589 bash: echo: write error: Device or resource busy 3590 # cat snapshot 3591 cat: snapshot: Device or resource busy 3592 3593 3594Instances 3595--------- 3596In the tracefs tracing directory, there is a directory called "instances". 3597This directory can have new directories created inside of it using 3598mkdir, and removing directories with rmdir. The directory created 3599with mkdir in this directory will already contain files and other 3600directories after it is created. 3601:: 3602 3603 # mkdir instances/foo 3604 # ls instances/foo 3605 buffer_size_kb buffer_total_size_kb events free_buffer per_cpu 3606 set_event snapshot trace trace_clock trace_marker trace_options 3607 trace_pipe tracing_on 3608 3609As you can see, the new directory looks similar to the tracing directory 3610itself. In fact, it is very similar, except that the buffer and 3611events are agnostic from the main directory, or from any other 3612instances that are created. 3613 3614The files in the new directory work just like the files with the 3615same name in the tracing directory except the buffer that is used 3616is a separate and new buffer. The files affect that buffer but do not 3617affect the main buffer with the exception of trace_options. Currently, 3618the trace_options affect all instances and the top level buffer 3619the same, but this may change in future releases. That is, options 3620may become specific to the instance they reside in. 3621 3622Notice that none of the function tracer files are there, nor is 3623current_tracer and available_tracers. This is because the buffers 3624can currently only have events enabled for them. 3625:: 3626 3627 # mkdir instances/foo 3628 # mkdir instances/bar 3629 # mkdir instances/zoot 3630 # echo 100000 > buffer_size_kb 3631 # echo 1000 > instances/foo/buffer_size_kb 3632 # echo 5000 > instances/bar/per_cpu/cpu1/buffer_size_kb 3633 # echo function > current_trace 3634 # echo 1 > instances/foo/events/sched/sched_wakeup/enable 3635 # echo 1 > instances/foo/events/sched/sched_wakeup_new/enable 3636 # echo 1 > instances/foo/events/sched/sched_switch/enable 3637 # echo 1 > instances/bar/events/irq/enable 3638 # echo 1 > instances/zoot/events/syscalls/enable 3639 # cat trace_pipe 3640 CPU:2 [LOST 11745 EVENTS] 3641 bash-2044 [002] .... 10594.481032: _raw_spin_lock_irqsave <-get_page_from_freelist 3642 bash-2044 [002] d... 10594.481032: add_preempt_count <-_raw_spin_lock_irqsave 3643 bash-2044 [002] d..1 10594.481032: __rmqueue <-get_page_from_freelist 3644 bash-2044 [002] d..1 10594.481033: _raw_spin_unlock <-get_page_from_freelist 3645 bash-2044 [002] d..1 10594.481033: sub_preempt_count <-_raw_spin_unlock 3646 bash-2044 [002] d... 10594.481033: get_pageblock_flags_group <-get_pageblock_migratetype 3647 bash-2044 [002] d... 10594.481034: __mod_zone_page_state <-get_page_from_freelist 3648 bash-2044 [002] d... 10594.481034: zone_statistics <-get_page_from_freelist 3649 bash-2044 [002] d... 10594.481034: __inc_zone_state <-zone_statistics 3650 bash-2044 [002] d... 10594.481034: __inc_zone_state <-zone_statistics 3651 bash-2044 [002] .... 10594.481035: arch_dup_task_struct <-copy_process 3652 [...] 3653 3654 # cat instances/foo/trace_pipe 3655 bash-1998 [000] d..4 136.676759: sched_wakeup: comm=kworker/0:1 pid=59 prio=120 success=1 target_cpu=000 3656 bash-1998 [000] dN.4 136.676760: sched_wakeup: comm=bash pid=1998 prio=120 success=1 target_cpu=000 3657 <idle>-0 [003] d.h3 136.676906: sched_wakeup: comm=rcu_preempt pid=9 prio=120 success=1 target_cpu=003 3658 <idle>-0 [003] d..3 136.676909: sched_switch: prev_comm=swapper/3 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=rcu_preempt next_pid=9 next_prio=120 3659 rcu_preempt-9 [003] d..3 136.676916: sched_switch: prev_comm=rcu_preempt prev_pid=9 prev_prio=120 prev_state=S ==> next_comm=swapper/3 next_pid=0 next_prio=120 3660 bash-1998 [000] d..4 136.677014: sched_wakeup: comm=kworker/0:1 pid=59 prio=120 success=1 target_cpu=000 3661 bash-1998 [000] dN.4 136.677016: sched_wakeup: comm=bash pid=1998 prio=120 success=1 target_cpu=000 3662 bash-1998 [000] d..3 136.677018: sched_switch: prev_comm=bash prev_pid=1998 prev_prio=120 prev_state=R+ ==> next_comm=kworker/0:1 next_pid=59 next_prio=120 3663 kworker/0:1-59 [000] d..4 136.677022: sched_wakeup: comm=sshd pid=1995 prio=120 success=1 target_cpu=001 3664 kworker/0:1-59 [000] d..3 136.677025: sched_switch: prev_comm=kworker/0:1 prev_pid=59 prev_prio=120 prev_state=S ==> next_comm=bash next_pid=1998 next_prio=120 3665 [...] 3666 3667 # cat instances/bar/trace_pipe 3668 migration/1-14 [001] d.h3 138.732674: softirq_raise: vec=3 [action=NET_RX] 3669 <idle>-0 [001] dNh3 138.732725: softirq_raise: vec=3 [action=NET_RX] 3670 bash-1998 [000] d.h1 138.733101: softirq_raise: vec=1 [action=TIMER] 3671 bash-1998 [000] d.h1 138.733102: softirq_raise: vec=9 [action=RCU] 3672 bash-1998 [000] ..s2 138.733105: softirq_entry: vec=1 [action=TIMER] 3673 bash-1998 [000] ..s2 138.733106: softirq_exit: vec=1 [action=TIMER] 3674 bash-1998 [000] ..s2 138.733106: softirq_entry: vec=9 [action=RCU] 3675 bash-1998 [000] ..s2 138.733109: softirq_exit: vec=9 [action=RCU] 3676 sshd-1995 [001] d.h1 138.733278: irq_handler_entry: irq=21 name=uhci_hcd:usb4 3677 sshd-1995 [001] d.h1 138.733280: irq_handler_exit: irq=21 ret=unhandled 3678 sshd-1995 [001] d.h1 138.733281: irq_handler_entry: irq=21 name=eth0 3679 sshd-1995 [001] d.h1 138.733283: irq_handler_exit: irq=21 ret=handled 3680 [...] 3681 3682 # cat instances/zoot/trace 3683 # tracer: nop 3684 # 3685 # entries-in-buffer/entries-written: 18996/18996 #P:4 3686 # 3687 # _-----=> irqs-off 3688 # / _----=> need-resched 3689 # | / _---=> hardirq/softirq 3690 # || / _--=> preempt-depth 3691 # ||| / delay 3692 # TASK-PID CPU# |||| TIMESTAMP FUNCTION 3693 # | | | |||| | | 3694 bash-1998 [000] d... 140.733501: sys_write -> 0x2 3695 bash-1998 [000] d... 140.733504: sys_dup2(oldfd: a, newfd: 1) 3696 bash-1998 [000] d... 140.733506: sys_dup2 -> 0x1 3697 bash-1998 [000] d... 140.733508: sys_fcntl(fd: a, cmd: 1, arg: 0) 3698 bash-1998 [000] d... 140.733509: sys_fcntl -> 0x1 3699 bash-1998 [000] d... 140.733510: sys_close(fd: a) 3700 bash-1998 [000] d... 140.733510: sys_close -> 0x0 3701 bash-1998 [000] d... 140.733514: sys_rt_sigprocmask(how: 0, nset: 0, oset: 6e2768, sigsetsize: 8) 3702 bash-1998 [000] d... 140.733515: sys_rt_sigprocmask -> 0x0 3703 bash-1998 [000] d... 140.733516: sys_rt_sigaction(sig: 2, act: 7fff718846f0, oact: 7fff71884650, sigsetsize: 8) 3704 bash-1998 [000] d... 140.733516: sys_rt_sigaction -> 0x0 3705 3706You can see that the trace of the top most trace buffer shows only 3707the function tracing. The foo instance displays wakeups and task 3708switches. 3709 3710To remove the instances, simply delete their directories: 3711:: 3712 3713 # rmdir instances/foo 3714 # rmdir instances/bar 3715 # rmdir instances/zoot 3716 3717Note, if a process has a trace file open in one of the instance 3718directories, the rmdir will fail with EBUSY. 3719 3720 3721Stack trace 3722----------- 3723Since the kernel has a fixed sized stack, it is important not to 3724waste it in functions. A kernel developer must be conscious of 3725what they allocate on the stack. If they add too much, the system 3726can be in danger of a stack overflow, and corruption will occur, 3727usually leading to a system panic. 3728 3729There are some tools that check this, usually with interrupts 3730periodically checking usage. But if you can perform a check 3731at every function call that will become very useful. As ftrace provides 3732a function tracer, it makes it convenient to check the stack size 3733at every function call. This is enabled via the stack tracer. 3734 3735CONFIG_STACK_TRACER enables the ftrace stack tracing functionality. 3736To enable it, write a '1' into /proc/sys/kernel/stack_tracer_enabled. 3737:: 3738 3739 # echo 1 > /proc/sys/kernel/stack_tracer_enabled 3740 3741You can also enable it from the kernel command line to trace 3742the stack size of the kernel during boot up, by adding "stacktrace" 3743to the kernel command line parameter. 3744 3745After running it for a few minutes, the output looks like: 3746:: 3747 3748 # cat stack_max_size 3749 2928 3750 3751 # cat stack_trace 3752 Depth Size Location (18 entries) 3753 ----- ---- -------- 3754 0) 2928 224 update_sd_lb_stats+0xbc/0x4ac 3755 1) 2704 160 find_busiest_group+0x31/0x1f1 3756 2) 2544 256 load_balance+0xd9/0x662 3757 3) 2288 80 idle_balance+0xbb/0x130 3758 4) 2208 128 __schedule+0x26e/0x5b9 3759 5) 2080 16 schedule+0x64/0x66 3760 6) 2064 128 schedule_timeout+0x34/0xe0 3761 7) 1936 112 wait_for_common+0x97/0xf1 3762 8) 1824 16 wait_for_completion+0x1d/0x1f 3763 9) 1808 128 flush_work+0xfe/0x119 3764 10) 1680 16 tty_flush_to_ldisc+0x1e/0x20 3765 11) 1664 48 input_available_p+0x1d/0x5c 3766 12) 1616 48 n_tty_poll+0x6d/0x134 3767 13) 1568 64 tty_poll+0x64/0x7f 3768 14) 1504 880 do_select+0x31e/0x511 3769 15) 624 400 core_sys_select+0x177/0x216 3770 16) 224 96 sys_select+0x91/0xb9 3771 17) 128 128 system_call_fastpath+0x16/0x1b 3772 3773Note, if -mfentry is being used by gcc, functions get traced before 3774they set up the stack frame. This means that leaf level functions 3775are not tested by the stack tracer when -mfentry is used. 3776 3777Currently, -mfentry is used by gcc 4.6.0 and above on x86 only. 3778 3779More 3780---- 3781More details can be found in the source code, in the `kernel/trace/*.c` files. 3782