bcc-funcinterval(8)

1funcinterval(8)             System Manager's Manual            funcinterval(8)
2
3
4

NAME

6       funcinterval - Time interval between the same function, tracepoint as a
7       histogram.
8

SYNOPSIS

10       funcinterval [-h] [-p PID] [-i INTERVAL] [-d DURATION] [-T]  [-u]  [-m]
11       [-v] pattern
12

DESCRIPTION

14       This tool times interval between the same function as a histogram.
15
16       eBPF/bcc  is very suitable for platform performance tuning.  By funcla‐
17       tency, we can profile specific functions to know how latency this func‐
18       tion  costs.  However,  sometimes performance drop is not about the la‐
19       tency of function but the interval between function calls.   funcinter‐
20       val is born for this purpose.
21
22       This  tool uses in-kernel eBPF maps for storing timestamps and the his‐
23       togram, for efficiency.
24
25       WARNING: This uses dynamic tracing of (what can be many) functions,  an
26       activity that has had issues on some kernel versions (risk of panics or
27       freezes). Test, and know what you are doing, before use.
28
29       Since this uses BPF, only the root user can use this tool.
30

REQUIREMENTS

32       CONFIG_BPF and bcc.
33

OPTIONS

35       pattern Function name.  -h Print usage message.
36
37       -p PID Trace this process ID only.
38
39       -i INTERVAL
40              Print output every interval seconds.
41
42       -d DURATION
43              Total duration of trace, in seconds.
44
45       -T     Include timestamps on output.
46
47       -u     Output histogram in microseconds.
48
49       -m     Output histogram in milliseconds.
50
51       -v     Print the BPF program (for debugging purposes).
52

EXAMPLES

54       Time the interval of do_sys_open() kernel function as a histogram:
55              # funcinterval do_sys_open
56
57       Time the interval of xhci_ring_ep_doorbell(), in microseconds:
58              # funcinterval -u xhci_ring_ep_doorbell
59
60       Time the interval of do_nanosleep(), in milliseconds
61              # funcinterval -m do_nanosleep
62
63       Output every 5 seconds, with timestamps:
64              # funcinterval -mTi 5 vfs_read
65
66       Time process 181 only:
67              # funcinterval -p 181 vfs_read
68
69       Time the interval of mm_vmscan_direct_reclaim_begin tracepoint:
70              # funcinterval t:vmscan:mm_vmscan_direct_reclaim_begin
71
72       Time the interval of c:malloc used by top every 3 seconds:
73              # funcinterval -p `pidof -s top` -i 3 c:malloc
74
75       Time /usr/local/bin/python main function:
76              # funcinterval /usr/local/bin/python:main
77

FIELDS

79       necs   Nanosecond range
80
81       usecs  Microsecond range
82
83       msecs  Millisecond range
84
85       count  How many calls fell into this range
86
87       distribution
88              An ASCII bar chart to visualize the distribution (count column)
89

OVERHEAD

91       This traces kernel functions and maintains in-kernel timestamps  and  a
92       histogram,  which  are  asynchronously copied to user-space. While this
93       method is very efficient, the rate of kernel functions can also be very
94       high  (>1M/sec),  at which point the overhead is expected to be measur‐
95       able. Measure in a test environment  and  understand  overheads  before
96       use. You can also use funccount to measure the rate of kernel functions
97       over a short duration, to set some expectations before use.
98

SOURCE

100       This is from bcc.
101
102              https://github.com/iovisor/bcc
103
104       Also look in the bcc distribution for a  companion  _examples.txt  file
105       containing example usage, output, and commentary for this tool.
106

OS

108       Linux
109

STABILITY

111       Unstable - in development.
112

AUTHOR

114       Edward Wu
115