1perf-kvm(1) 2=========== 3 4NAME 5---- 6perf-kvm - Tool to trace/measure kvm guest os 7 8SYNOPSIS 9-------- 10[verse] 11'perf kvm' [--host] [--guest] [--guestmount=<path> 12 [--guestkallsyms=<path> --guestmodules=<path> | --guestvmlinux=<path>]] 13 {top|record|report|diff|buildid-list} [<options>] 14'perf kvm' [--host] [--guest] [--guestkallsyms=<path> --guestmodules=<path> 15 | --guestvmlinux=<path>] {top|record|report|diff|buildid-list|stat} [<options>] 16'perf kvm stat [record|report|live] [<options>] 17 18DESCRIPTION 19----------- 20There are a couple of variants of perf kvm: 21 22 'perf kvm [options] top <command>' to generates and displays 23 a performance counter profile of guest os in realtime 24 of an arbitrary workload. 25 26 'perf kvm record <command>' to record the performance counter profile 27 of an arbitrary workload and save it into a perf data file. We set the 28 default behavior of perf kvm as --guest, so if neither --host nor --guest 29 is input, the perf data file name is perf.data.guest. If --host is input, 30 the perf data file name is perf.data.kvm. If you want to record data into 31 perf.data.host, please input --host --no-guest. The behaviors are shown as 32 following: 33 Default('') -> perf.data.guest 34 --host -> perf.data.kvm 35 --guest -> perf.data.guest 36 --host --guest -> perf.data.kvm 37 --host --no-guest -> perf.data.host 38 39 'perf kvm report' to display the performance counter profile information 40 recorded via perf kvm record. 41 42 'perf kvm diff' to displays the performance difference amongst two perf.data 43 files captured via perf record. 44 45 'perf kvm buildid-list' to display the buildids found in a perf data file, 46 so that other tools can be used to fetch packages with matching symbol tables 47 for use by perf report. As buildid is read from /sys/kernel/notes in os, then 48 if you want to list the buildid for guest, please make sure your perf data file 49 was captured with --guestmount in perf kvm record. 50 51 'perf kvm stat <command>' to run a command and gather performance counter 52 statistics. 53 Especially, perf 'kvm stat record/report' generates a statistical analysis 54 of KVM events. Currently, vmexit, mmio (x86 only) and ioport (x86 only) 55 events are supported. 'perf kvm stat record <command>' records kvm events 56 and the events between start and end <command>. 57 And this command produces a file which contains tracing results of kvm 58 events. 59 60 'perf kvm stat report' reports statistical data which includes events 61 handled time, samples, and so on. 62 63 'perf kvm stat live' reports statistical data in a live mode (similar to 64 record + report but with statistical data updated live at a given display 65 rate). 66 67OPTIONS 68------- 69-i:: 70--input=<path>:: 71 Input file name, for the 'report', 'diff' and 'buildid-list' subcommands. 72-o:: 73--output=<path>:: 74 Output file name, for the 'record' subcommand. Doesn't work with 'report', 75 just redirect the output to a file when using 'report'. 76--host:: 77 Collect host side performance profile. 78--guest:: 79 Collect guest side performance profile. 80--guestmount=<path>:: 81 Guest os root file system mount directory. Users mounts guest os 82 root directories under <path> by a specific filesystem access method, 83 typically, sshfs. For example, start 2 guest os. The one's pid is 8888 84 and the other's is 9999. 85 #mkdir ~/guestmount; 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>:: 90 Guest os /proc/kallsyms file copy. 'perf' kvm' reads it to get guest 91 kernel symbols. Users copy it out from guest os. 92--guestmodules=<path>:: 93 Guest os /proc/modules file copy. 'perf' kvm' reads it to get guest 94 kernel module information. Users copy it out from guest os. 95--guestvmlinux=<path>:: 96 Guest os kernel vmlinux. 97-v:: 98--verbose:: 99 Be more verbose (show counter open errors, etc). 100 101STAT REPORT OPTIONS 102------------------- 103--vcpu=<value>:: 104 analyze events which occur on this vcpu. (default: all vcpus) 105 106--event=<value>:: 107 event to be analyzed. Possible values: vmexit, mmio (x86 only), 108 ioport (x86 only). (default: vmexit) 109-k:: 110--key=<value>:: 111 Sorting key. Possible values: sample (default, sort by samples 112 number), time (sort by average time). 113-p:: 114--pid=:: 115 Analyze events only for given process ID(s) (comma separated list). 116 117STAT LIVE OPTIONS 118----------------- 119-d:: 120--display:: 121 Time in seconds between display updates 122 123-m:: 124--mmap-pages=:: 125 Number of mmap data pages (must be a power of two) or size 126 specification with appended unit character - B/K/M/G. The 127 size is rounded up to have nearest pages power of two value. 128 129-a:: 130--all-cpus:: 131 System-wide collection from all CPUs. 132 133-p:: 134--pid=:: 135 Analyze events only for given process ID(s) (comma separated list). 136 137--vcpu=<value>:: 138 analyze events which occur on this vcpu. (default: all vcpus) 139 140 141--event=<value>:: 142 event to be analyzed. Possible values: vmexit, 143 mmio (x86 only), ioport (x86 only). 144 (default: vmexit) 145 146-k:: 147--key=<value>:: 148 Sorting key. Possible values: sample (default, sort by samples 149 number), time (sort by average time). 150 151--duration=<value>:: 152 Show events other than HLT (x86 only) or Wait state (s390 only) 153 that take longer than duration usecs. 154 155--proc-map-timeout:: 156 When processing pre-existing threads /proc/XXX/mmap, it may take 157 a long time, because the file may be huge. A time out is needed 158 in such cases. 159 This option sets the time out limit. The default value is 500 ms. 160 161SEE ALSO 162-------- 163linkperf:perf-top[1], linkperf:perf-record[1], linkperf:perf-report[1], 164linkperf:perf-diff[1], linkperf:perf-buildid-list[1], 165linkperf:perf-stat[1] 166