Skip to content

Latest commit

 

History

History
380 lines (305 loc) · 13.3 KB

uftrace-graph.md

File metadata and controls

380 lines (305 loc) · 13.3 KB

% UFTRACE-GRAPH(1) Uftrace User Manuals % Namhyung Kim namhyung@gmail.com % Sep, 2018

NAME

uftrace-graph - Show function call graph

SYNOPSIS

uftrace graph [options] [FUNCTION]

DESCRIPTION

This command shows a function call graph for the binary or the given function in a uftrace record datafile. If the function name is omitted, whole function call graph will be shown. If a user provides a function name, it will show backtrace and calling functions. Each function in the output is annotated with a hit count and the total time spent running that function.

GRAPH OPTIONS

-f FIELD, --output-fields=FIELD : Customize field in the output. Possible values are: total, self and addr. Multiple fields can be set by using comma. Special field of 'none' can be used (solely) to hide all fields. Default is 'total'. See FIELDS.

--task : Print task graph instead of normal function graph. Each node in the output shows a process or thread(printed in green color).

--srcline : Show source location of each function if available.

--format=TYPE : Show format style output. Currently, normal and html styles are supported.

COMMON OPTIONS

-F FUNC, --filter=FUNC : Set filter to trace selected functions and their children functions. This option can be used more than once. See uftrace-replay(1) for an explanation of filters.

-N FUNC, --notrace=FUNC : Set filter not to trace selected functions and their children functions. This option can be used more than once. See uftrace-replay(1) for an explanation of filters.

-C FUNC, --caller-filter=FUNC : Set filter to trace callers of selected functions only. This option can be used more than once. See uftrace-replay(1) for an explanation of filters.

-T TRG, --trigger=TRG : Set trigger on selected functions. This option can be used more than once. See uftrace-replay(1) for an explanation of triggers.

-D DEPTH, --depth DEPTH : Set trace limit in nesting level.

-t TIME, --time-filter=TIME : Do not show functions which run under the time threshold. If some functions explicitly have the 'trace' trigger applied, those are always traced regardless of execution time.

-Z SIZE, --size-filter=SIZE : Do not show functions smaller than SIZE bytes.

-L LOCATION, --loc-filter=LOCATION : Set filter to trace selected source locations. This option can be used more than once.

--no-libcall : Do not show library calls.

--no-event : Do not show any events. Implies --no-sched.

--no-sched : Do not show schedule events.

--no-sched-preempt : Do not show preempt schedule events but show regular(sleeping) schedule events.

--match=TYPE : Use pattern match using TYPE. Possible types are regex and glob. Default is regex.

--with-syms=DIR : Read symbol data from the .sym files in DIR directory instead of the binary. This can be useful to deal with stripped binaries. The file name of the main binary should be the same when saved and used.

COMMON ANALYSIS OPTIONS

-H FUNC, --hide=FUNC : Set filter not to trace selected functions. It doesn't affect their subtrees, but hides only the given functions. This option can be used more than once. See uftrace-replay(1) for an explanation of filters.

--kernel-full : Show all kernel functions called outside of user functions.

--kernel-only : Show kernel functions only without user functions.

--event-full : Show all (user) events outside of user functions.

--tid=TID[,TID,...] : Only print functions called by the given tasks. To see the list of tasks in the data file, you can use uftrace report --task or uftrace info. This option can also be used more than once.

--demangle=TYPE : Use demangled C++ symbol names for filters, triggers, arguments and/or return values. Possible values are "full", "simple" and "no". Default is "simple" which ignores function arguments and template parameters.

-r RANGE, --time-range=RANGE : Only show functions executed within the time RANGE. The RANGE can be <start><stop> (separated by "") and one of <start> and <stop> can be omitted. The <start> and <stop> are timestamp or elapsed time if they have <time_unit> postfix, for example '100us'. The timestamp or elapsed time can be shown with -f time or -f elapsed option respectively in uftrace replay(1).

EXAMPLES

This command show data like below:

$ uftrace record loop

$ uftrace replay
# DURATION    TID     FUNCTION
            [24447] | main() {
            [24447] |   foo() {
   8.134 us [24447] |     loop();
   7.296 us [24447] |     loop();
   7.234 us [24447] |     loop();
  24.324 us [24447] |   } /* foo */
            [24447] |   foo() {
   7.234 us [24447] |     loop();
   7.231 us [24447] |     loop();
   7.231 us [24447] |     loop();
  22.302 us [24447] |   } /* foo */
            [24447] |   bar() {
  10.100 ms [24447] |     usleep();
  10.138 ms [24447] |   } /* bar */
  10.293 ms [24447] | } /* main */

Running the graph command shows function call graph like below:

$ uftrace graph
# Function Call Graph for 'loop' (session: 073f1e84aa8b09d3)
========== FUNCTION CALL GRAPH ==========
  10.293 ms : (1) loop
  10.293 ms : (1) main
  46.626 us :  +-(2) foo
  44.360 us :  | (6) loop
            :  |
  10.138 ms :  +-(1) bar
  10.100 ms :    (1) usleep

The graph root is not a function, but the executing process itself. The left side shows total time running the function on the right side. The number in parentheses before the function name is the invocation count. As you can see, main was called once and ran for 10 msec. It called foo twice and then foo called loop 6 times in total. The printed time is the total execution time for all function invocations.

It can also be seen that main called bar once and that bar then called usleep once. To minimize nesting, the output shows calls at the same level if only a single call path exists. Since the nodes usleep and main are not directly connected, usleep is not called from main directly.

Running the graph command on the main function shows called functions and backtrace like below:

$ uftrace graph main
# Function Call Graph for 'main' (session: 073f1e84aa8b09d3)
=============== BACKTRACE ===============
 backtrace #0: hit 1, time  10.293 ms
   [0] main (0x4004f0)

========== FUNCTION CALL GRAPH ==========
# TOTAL TIME   FUNCTION
   10.293 ms : (1) main
   46.626 us :  +-(2) foo
   44.360 us :  | (6) loop
             :  |
   10.138 ms :  +-(1) bar
   10.100 ms :    (1) usleep

Note that the 'main' is the top-level function so it has no backtrace above itself. Running the graph command on a leaf function looks like below.

$ uftrace graph loop
# Function Call Graph for 'loop' (session: 073f1e84aa8b09d3)
=============== BACKTRACE ===============
 backtrace #0: hit 6, time  44.360 us
   [0] main (0x4004b0)
   [1] foo (0x400622)
   [2] loop (0x400f5f6)

========== FUNCTION CALL GRAPH ==========
# TOTAL TIME   FUNCTION
   44.360 us : (6) loop

The backtrace shows that loop is called from foo and that foo is called from main. Since loop is a leaf function, it didn't call any other function. In this case, loop was called only from a single path so backtrace #0 is hit 6 times.

While graph command shows function-level call graph, --task option makes the output in task-level graph which shows how processes and threads are created. The term here task includes process and thread.

For example, the task graph of GCC compiler can be shown as follows:

$ uftrace record --force /usr/bin/gcc hello.c

$ uftrace graph --task
========== TASK GRAPH ==========
# TOTAL TIME   SELF TIME     TID     TASK NAME
  159.854 ms    4.440 ms  [ 82723] : gcc
                                   :  |
   90.951 ms   90.951 ms  [ 82734] :  +----cc1
                                   :  |
   17.150 ms   17.150 ms  [ 82735] :  +----as
                                   :  |
   45.183 ms    6.076 ms  [ 82736] :  +----collect2
                                   :        |
   38.880 ms   38.880 ms  [ 82737] :        +----ld

The above output shows gcc created cc1, as, and collect2 processes then collect2 created ld process.

TOTAL TIME is the lifetime of the task from its creation to termination, and SELF TIME is also lifetime, but it excludes internal idle time. TID is the thread id of the task.

The following shows task graph of uftrace recording itself. It shows uftrace created t-abc process, and also created many threads whose names are all WriterThread.

$ uftrace record -P. ./uftrace record -d uftrace.data.abc t-abc

$ uftrace graph --task
========== TASK GRAPH ==========
# TOTAL TIME   SELF TIME     TID     TASK NAME
  404.929 ms  321.692 ms  [  4230] : uftrace
                                   :  |
  278.662 us  278.662 us  [  4241] :  +----t-abc
                                   :  |
   33.754 ms    4.061 ms  [  4242] :  +-WriterThread
   27.415 ms  120.992 us  [  4244] :  +-WriterThread
   27.212 ms    8.119 ms  [  4245] :  +-WriterThread
   26.754 ms    6.616 ms  [  4248] :  +-WriterThread
   26.859 ms    8.154 ms  [  4247] :  +-WriterThread
   26.509 ms    1.645 ms  [  4243] :  +-WriterThread
   25.320 ms   57.350 us  [  4246] :  +-WriterThread
   24.757 ms    4.391 ms  [  4249] :  +-WriterThread
   26.040 ms    3.707 ms  [  4250] :  +-WriterThread
   24.004 ms    3.999 ms  [  4251] :  +-WriterThread

Please note that the indentation depth of thread is different from process.

Running the graph command with --srcline option shows source location in call graph like below:

$ uftrace record --srcline t-abc
$ uftrace graph --srcline
# Function Call Graph for 't-abc' (session: 60195bac953d8736)
========== FUNCTION CALL GRAPH ==========
# TOTAL TIME   FUNCTION [SOURCE]
  8.909 us : (1) t-abc
  1.260 us :  +-(1) __monstartup
           :  |
  0.179 us :  +-(1) __cxa_atexit
           :  |
  7.470 us :  +-(1) main [tests/s-abc.c:26]
  5.522 us :    (1) a [tests/s-abc.c:11]
  4.912 us :    (1) b [tests/s-abc.c:16]
  4.176 us :    (1) c [tests/s-abc.c:21]
  0.794 us :    (1) getpid

FIELDS

The uftrace allows for user to customize the graph output with some of fields. Here the field means info on the left side of the colon (:) character. By default it uses time only, but you can use other fields in any order like:

$ uftrace record tests/t-abc
$ uftrace graph -f total,self,addr
# Function Call Graph for 't-sort' (session: b007f4b7cf792878)
========== FUNCTION CALL GRAPH ==========
# TOTAL TIME  SELF TIME      ADDRESS     FUNCTION
   10.145 ms              561f652cd610 : (1) t-sort
   10.145 ms   39.890 us  561f652cd610 : (1) main
   16.773 us    0.734 us  561f652cd7ce :  +-(2) foo
   16.039 us   16.039 us  561f652cd7a0 :  | (6) loop
                                       :  |
   10.088 ms   14.740 us  561f652cd802 :  +-(1) bar
   10.073 ms   10.073 ms  561f652cd608 :    (1) usleep

Each field has following meaning:

  • total: function execution time in total
  • self : function execution time excluding its children's
  • addr : address of the function

The default value is 'total'. If given field name starts with "+", then it'll be appended to the default fields. So "-f +addr" is as same as "-f total,addr". And it also accepts a special field name of 'none' which disables the field display and shows function output only.

$ uftrace graph -f none
# Function Call Graph for 't-sort' (session: b007f4b7cf792878)
========== FUNCTION CALL GRAPH ==========
(1) t-sort
(1) main
 +-(2) foo
 | (6) loop
 |
 +-(1) bar
   (1) usleep

This output can be useful when comparing two different call graph outputs using diff tool.

It also supports field customization for task graph. The default field is set to total,self,tid, but the field option can also be used as follows:

$ uftrace graph --task -f tid,self
========== TASK GRAPH ==========
#    TID     SELF TIME   TASK NAME
  [ 82723]    4.440 ms : gcc
                       :  |
  [ 82734]   90.951 ms :  +----cc1
                       :  |
  [ 82735]   17.150 ms :  +----as
                       :  |
  [ 82736]    6.076 ms :  +----collect2
                       :        |
  [ 82737]   38.880 ms :        +----ld

Each field has following meaning:

  • total: total task lifetime from its creation to termination
  • self : task execution time excluding its idle time
  • tid : task id (obtained by gettid(2))

It also accepts a special field none, which hides all the fields on the left.

$ uftrace graph --task -f none
========== TASK GRAPH ==========
gcc
 |
 +----cc1
 |
 +----as
 |
 +----collect2
       |
       +----ld

SEE ALSO

uftrace(1), uftrace-record(1), uftrace-replay(1), uftrace-tui(1)