1PERF-LOCK(1) perf Manual PERF-LOCK(1)
2
3
4
6 perf-lock - Analyze lock events
7
9 perf lock {record|report|script|info|contention}
10
12 You can analyze various lock behaviours and statistics with this perf
13 lock command.
14
15 'perf lock record <command>' records lock events
16 between start and end <command>. And this command
17 produces the file "perf.data" which contains tracing
18 results of lock events.
19
20 'perf lock report' reports statistical data.
21
22 'perf lock script' shows raw lock events.
23
24 'perf lock info' shows metadata like threads or addresses
25 of lock instances.
26
27 'perf lock contention' shows contention statistics.
28
30 -i, --input=<file>
31 Input file name. (default: perf.data unless stdin is a fifo)
32
33 --output=<file>
34 Output file name for perf lock contention and report.
35
36 -v, --verbose
37 Be more verbose (show symbol address, etc).
38
39 -q, --quiet
40 Do not show any warnings or messages. (Suppress -v)
41
42 -D, --dump-raw-trace
43 Dump raw trace in ASCII.
44
45 -f, --force
46 Don’t complain, do it.
47
48 --vmlinux=<file>
49 vmlinux pathname
50
51 --kallsyms=<file>
52 kallsyms pathname
53
55 -k, --key=<value>
56 Sorting key. Possible values: acquired (default), contended,
57 avg_wait, wait_total, wait_max, wait_min.
58
59 -F, --field=<value>
60 Output fields. By default it shows all the fields but users can
61 customize that using this. Possible values: acquired, contended,
62 avg_wait, wait_total, wait_max, wait_min.
63
64 -c, --combine-locks
65 Merge lock instances in the same class (based on name).
66
67 -t, --threads
68 The -t option is to show per-thread lock stat like below:
69
70 $ perf lock report -t -F acquired,contended,avg_wait
71
72 Name acquired contended avg wait (ns)
73
74 perf 240569 9 5784
75 swapper 106610 19 543
76 :15789 17370 2 14538
77 ContainerMgr 8981 6 874
78 sleep 5275 1 11281
79 ContainerThread 4416 4 944
80 RootPressureThr 3215 5 1215
81 rcu_preempt 2954 0 0
82 ContainerMgr 2560 0 0
83 unnamed 1873 0 0
84 EventManager_De 1845 1 636
85 futex-default-S 1609 0 0
86
87 -E, --entries=<value>
88 Display this many entries.
89
91 -t, --threads
92 dump thread list in perf.data
93
94 -m, --map
95 dump map of lock instances (address:name table)
96
97 CONTENTION OPTIONS
98
99
100 .ft C
101 -k::
102 --key=<value>::
103 Sorting key. Possible values: contended, wait_total (default),
104 wait_max, wait_min, avg_wait.
105
106 -F::
107 --field=<value>::
108 Output fields. By default it shows all but the wait_min fields
109 and users can customize that using this. Possible values:
110 contended, wait_total, wait_max, wait_min, avg_wait.
111
112 -t::
113 --threads::
114 Show per-thread lock contention stat
115
116 -b::
117 --use-bpf::
118 Use BPF program to collect lock contention stats instead of
119 using the input data.
120
121 -a::
122 --all-cpus::
123 System-wide collection from all CPUs.
124
125 -C::
126 --cpu=<value>::
127 Collect samples only on the list of CPUs provided. Multiple CPUs can be
128 provided as a comma-separated list with no space: 0,1. Ranges of CPUs
129 are specified with -: 0-2. Default is to monitor all CPUs.
130
131 -p::
132 --pid=<value>::
133 Record events on existing process ID (comma separated list).
134
135 --tid=<value>::
136 Record events on existing thread ID (comma separated list).
137
138 -M::
139 --map-nr-entries=<value>::
140 Maximum number of BPF map entries (default: 16384).
141 This will be aligned to a power of 2.
142
143 --max-stack=<value>::
144 Maximum stack depth when collecting lock contention (default: 8).
145
146 --stack-skip=<value>::
147 Number of stack depth to skip when finding a lock caller (default: 3).
148
149 -E::
150 --entries=<value>::
151 Display this many entries.
152
153 -l::
154 --lock-addr::
155 Show lock contention stat by address
156
157 -o::
158 --lock-owner::
159 Show lock contention stat by owners. Implies --threads and
160 requires --use-bpf.
161
162 -Y::
163 --type-filter=<value>::
164 Show lock contention only for given lock types (comma separated list).
165 Available values are:
166 semaphore, spinlock, rwlock, rwlock:R, rwlock:W, rwsem, rwsem:R, rwsem:W,
167 rtmutex, rwlock-rt, rwlock-rt:R, rwlock-rt:W, pcpu-sem, pcpu-sem:R, pcpu-sem:W,
168 mutex
169
170 Note that RW-variant of locks have :R and :W suffix. Names without the
171 suffix are shortcuts for the both variants. Ex) rwsem = rwsem:R + rwsem:W.
172
173 -L::
174 --lock-filter=<value>::
175 Show lock contention only for given lock addresses or names (comma separated list).
176
177 -S::
178 --callstack-filter=<value>::
179 Show lock contention only if the callstack contains the given string.
180 Note that it matches the substring so 'rq' would match both 'raw_spin_rq_lock'
181 and 'irq_enter_rcu'.
182
183 -x::
184 --field-separator=<SEP>::
185 Show results using a CSV-style output to make it easy to import directly
186 into spreadsheets. Columns are separated by the string specified in SEP.
187
188
189 SEE ALSO
190 .ft
191
192
193 perf(1)
194
195
196
197perf 11/28/2023 PERF-LOCK(1)