1PERF-KVM(1) perf Manual PERF-KVM(1)
2
6 perf-kvm - Tool to trace/measure kvm guest os
7
9 perf kvm [--host] [--guest] [--guestmount=<path>
10 [--guestkallsyms=<path> --guestmodules=<path> | --guestvmlinux=<path>]]
11 {top|record|report|diff|buildid-list} [<options>]
12 perf kvm [--host] [--guest] [--guestkallsyms=<path> --guestmodules=<path>
13 | --guestvmlinux=<path>] {top|record|report|diff|buildid-list|stat} [<options>]
14 'perf kvm stat [record|report|live] [<options>]
15
17 There are a couple of variants of perf kvm:
18 'perf kvm [options] top <command>' to generates and displays
20 a performance counter profile of guest os in realtime
21 of an arbitrary workload.
22 'perf kvm record <command>' to record the performance counter profile
24 of an arbitrary workload and save it into a perf data file. We set the
25 default behavior of perf kvm as --guest, so if neither --host nor --guest
26 is input, the perf data file name is perf.data.guest. If --host is input,
27 the perf data file name is perf.data.kvm. If you want to record data into
28 perf.data.host, please input --host --no-guest. The behaviors are shown as
29 following:
30 Default('') -> perf.data.guest
31 --host -> perf.data.kvm
32 --guest -> perf.data.guest
33 --host --guest -> perf.data.kvm
34 --host --no-guest -> perf.data.host
35 'perf kvm report' to display the performance counter profile information
37 recorded via perf kvm record.
38 'perf kvm diff' to displays the performance difference amongst two perf.data
40 files captured via perf record.
41 'perf kvm buildid-list' to display the buildids found in a perf data file,
43 so that other tools can be used to fetch packages with matching symbol tables
44 for use by perf report. As buildid is read from /sys/kernel/notes in os, then
45 if you want to list the buildid for guest, please make sure your perf data file
46 was captured with --guestmount in perf kvm record.
47 'perf kvm stat <command>' to run a command and gather performance counter
49 statistics.
50 Especially, perf 'kvm stat record/report' generates a statistical analysis
51 of KVM events. Currently, vmexit, mmio (x86 only) and ioport (x86 only)
52 events are supported. 'perf kvm stat record <command>' records kvm events
53 and the events between start and end <command>.
54 And this command produces a file which contains tracing results of kvm
55 events.
56 'perf kvm stat report' reports statistical data which includes events
58 handled sample, percent_sample, time, percent_time, max_t, min_t, mean_t.
59 'perf kvm stat live' reports statistical data in a live mode (similar to
61 record + report but with statistical data updated live at a given display
62 rate).
63
65 -i, --input=<path>
66 Input file name, for the report, diff and buildid-list subcommands.
67 -o, --output=<path>
69 Output file name, for the record subcommand. Doesn’t work with
70 report, just redirect the output to a file when using report.
71 --host
73 Collect host side performance profile.
74 --guest
76 Collect guest side performance profile.
77 --guestmount=<path>
79 Guest OS root file system mount directory. Users mount guest OS
80 root directories under <path> by a specific filesystem access
81 method, typically, sshfs. For example, start 2 guest OS, one’s pid
82 is 8888 and the other’s is 9999:
83 $ mkdir ~/guestmount
85 $ cd ~/guestmount
86 $ sshfs -o allow_other,direct_io -p 5551 localhost:/ 8888/
87 $ sshfs -o allow_other,direct_io -p 5552 localhost:/ 9999/
88 $ perf kvm --host --guest --guestmount=~/guestmount top
89 --guestkallsyms=<path>
91 Guest OS /proc/kallsyms file copy. perf reads it to get guest
92 kernel symbols. Users copy it out from guest OS.
93 --guestmodules=<path>
95 Guest OS /proc/modules file copy. perf reads it to get guest kernel
96 module information. Users copy it out from guest OS.
97 --guestvmlinux=<path>
99 Guest OS kernel vmlinux.
100 --guest-code
102 Indicate that guest code can be found in the hypervisor process,
103 which is a common case for KVM test programs.
104 --stdio
106 Use the stdio interface.
107 -v, --verbose
109 Be more verbose (show counter open errors, etc).
110
112 --vcpu=<value>
113 analyze events which occur on this vcpu. (default: all vcpus)
114 --event=<value>
116 event to be analyzed. Possible values: vmexit, mmio (x86 only),
117 ioport (x86 only). (default: vmexit)
118 -k, --key=<value>
120 Sorting key. Possible values: sample (default, sort by samples
121 number), percent_sample (sort by sample percentage), time (sort by
122 average time), precent_time (sort by time percentage), max_t (sort
123 by maximum time), min_t (sort by minimum time), mean_t (sort by
124 mean time).
125 -p, --pid=
127 Analyze events only for given process ID(s) (comma separated list).
128
130 -d, --display
131 Time in seconds between display updates
132 -m, --mmap-pages=
134 Number of mmap data pages (must be a power of two) or size
135 specification with appended unit character - B/K/M/G. The size is
136 rounded up to have nearest pages power of two value.
137 -a, --all-cpus
139 System-wide collection from all CPUs.
140 -p, --pid=
142 Analyze events only for given process ID(s) (comma separated list).
143 --vcpu=<value>
145 analyze events which occur on this vcpu. (default: all vcpus)
146 --event=<value>
148 event to be analyzed. Possible values: vmexit, mmio (x86 only),
149 ioport (x86 only). (default: vmexit)
150 -k, --key=<value>
152 Sorting key. Possible values: sample (default, sort by samples
153 number), time (sort by average time).
154 --duration=<value>
156 Show events other than HLT (x86 only) or Wait state (s390 only)
157 that take longer than duration usecs.
158 --proc-map-timeout
160 When processing pre-existing threads /proc/XXX/mmap, it may take a
161 long time, because the file may be huge. A time out is needed in
162 such cases. This option sets the time out limit. The default value
163 is 500 ms.
164
166 perf-top(1), perf-record(1), perf-report(1), perf-diff(1), perf-
167 buildid-list(1), perf-stat(1)
168perf 11/28/2023 PERF-KVM(1)