bcc-hardirqs(8)

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

NAME

6       hardirqs  -  Measure  hard  IRQ (hard interrupt) event time. Uses Linux
7       eBPF/bcc.
8

SYNOPSIS

10       hardirqs [-h] [-T] [-N] [-C] [-d] [interval] [outputs]
11

DESCRIPTION

13       This summarizes the time spent servicing hard IRQs  (hard  interrupts),
14       and  can  show this time as either totals or histogram distributions. A
15       system-wide summary of this time is shown by the  %irq  column  of  mp‐
16       stat(1),  and  event  counts  (but not times) are shown by /proc/inter‐
17       rupts.
18
19       This tool uses the irq:irq_handler_entry and irq:irq_handler_exit  ker‐
20       nel  tracepoints, which is a stable tracing mechanism. BPF programs can
21       attach to tracepoints from Linux 4.7 only. An  older  version  of  this
22       tool  is  available  in  tools/old,  and uses kprobes instead of trace‐
23       points.
24
25       Since this uses BPF, only the root user can use this tool.
26

REQUIREMENTS

28       CONFIG_BPF and bcc.
29

OPTIONS

31       -h     Print usage message.
32
33       -T     Include timestamps on output.
34
35       -N     Output in nanoseconds.
36
37       -C     Count events only.
38
39       -d     Show IRQ time distribution as histograms.
40

EXAMPLES

42       Sum hard IRQ event time until Ctrl-C:
43              # hardirqs
44
45       Show hard IRQ event time as histograms:
46              # hardirqs -d
47
48       Print 1 second summaries, 10 times:
49              # hardirqs 1 10
50
51       1 second summaries, printed in nanoseconds, with timestamps:
52              # hardirqs -NT 1
53

FIELDS

55       HARDIRQ
56              The irq action name for this hard IRQ.
57
58       TOTAL_usecs
59              Total time spent in this hard IRQ in microseconds.
60
61       TOTAL_nsecs
62              Total time spent in this hard IRQ in nanoseconds.
63
64       usecs  Range of microseconds for this bucket.
65
66       nsecs  Range of nanoseconds for this bucket.
67
68       count  Number of hard IRQs in this time range.
69
70       distribution
71              ASCII representation of the distribution (the count column).
72

OVERHEAD

74       This traces kernel functions and maintains in-kernel counts, which  are
75       asynchronously  copied  to  user-space. While the rate of interrupts be
76       very high (>1M/sec), this is a relatively efficient way to trace  these
77       events,  and  so  the overhead is expected to be small for normal work‐
78       loads, but could become noticeable for heavy workloads.  Measure  in  a
79       test environment before use.
80

SOURCE

82       This is from bcc.
83
84              https://github.com/iovisor/bcc
85
86       Also  look  in  the bcc distribution for a companion _examples.txt file
87       containing example usage, output, and commentary for this tool.
88

OS

90       Linux
91

STABILITY

93       Unstable - in development.
94

AUTHOR

96       Brendan Gregg, Hengqi Chen
97