1OPCONTROL(1) General Commands Manual OPCONTROL(1)
2
6 opcontrol - control OProfile profiling
7
9 opcontrol [ options ]
10
12 opcontrol can be used to start profiling, end a profiling session, dump
13 profile data, and set up the profiling parameters.
14
17 --help Show help message.
18 --version
20 Show version.
21 --list-events
23 Shows the monitorable events.
24 --init Load the OProfile module if required and make the OProfile
26 driver interface available.
27 --setup
29 Followed by list options for profiling setup. Store setup in
30 ~root/.oprofile/daemonrc. Optional.
31 --status
33 Show configuration information.
34 --start-daemon
36 Start the oprofile daemon without starting profiling. Not avail‐
37 able in 2.2/2.4 kernels.
38 --start
40 Start data collection with either arguments provided by --setup
41 or with information saved in ~root/.oprofile/daemonrc.
42 --dump Force a flush of the collected profiling data to the daemon.
44 --stop Stop data collection. Not available in 2.2/2.4 kernels.
46 --shutdown
48 Stop data collection and kill the daemon.
49 --reset
51 Clear out data from current session, but leaves saved sessions.
52 --save=sessionname
54 Save data from current session to sessionname.
55 --deinit
57 Shut down daemon. Unload the oprofile module and oprofilefs.
58 --session-dir=dir_path
60 Use sample database out of directory dir_path instead of the
61 default location (/var/lib/oprofile).
62 --buffer-size=num
64 Set kernel buffer to num samples. When using a 2.6 kernel buffer
65 watershed need to be tweaked when changing this value.
66 --buffer-watershed=num
68 Set kernel buffer watershed to num samples (2.6 only). When
69 it'll remain only buffer-size - buffer-watershed free entry in
70 the kernel buffer data will be flushed to daemon, most usefull
71 value are in the range [0.25 - 0.5] * buffer-size.
72 --cpu-buffer-size=num
74 Set kernel per cpu buffer to num samples (2.6 only). If you pro‐
75 file at high rate it can help to increase this if the log file
76 show excessive count of sample lost cpu buffer overflow.
77 --event=[event|"default"]
79 Specify an event to measure for the hardware performance coun‐
80 ters, or "default" for the default event. The event is of the
81 form "CPU_CLK_UNHALTED:30000:0:1:1" where the numeric values are
82 count, unit mask, kernel-space counting, user-space counting,
83 respectively. Note that this over-rides all previous events
84 selected; if you want two or more counters used simultaneously,
85 you must specify them on the same opcontrol invocation.
86 --separate=[none,lib,kernel,thread,cpu,all]
88 Separate samples based on the given separator. 'lib' separates
89 dynamically linked library samples per application. 'kernel'
90 separates kernel and kernel module samples per application;
91 'kernel' implies 'library'. 'thread' gives separation for each
92 thread and task. 'cpu' separates for each CPU. 'all' implies
93 all of the above options and 'none' turns off separation.
94 --callgraph=#depth
96 Enable callgraph sample collection with a maximum depth. Use 0
97 to disable callgraph profiling. This option is available on x86
98 using a 2.6+ kernel with callgraph support enabled. It is also
99 available on PowerPC using a 2.6.17+ kernel.
100 --image=[name,name...|"all"]
102 Only profile the given absolute paths to binaries, or "all" to
103 profile everything (the default).
104 --vmlinux=file
106 vmlinux kernel image.
107 --no-vmlinux
109 Use this when you don't have a kernel vmlinux file, and you
110 don't want to profile the kernel.
111 --verbose
113 Be verbose in the daemon log. This has a high overhead.
114 --kernel-range=start,end
116 Set kernel range vma address in hexadecimal.
117
120 --xen=file
121 Xen image
122 --active-domains=<list>
124 List of domain ids participating in a multi-domain profiling
125 session. Each of the specified domains must run an instance of
126 oprofile. The sequence of opcontrol commands in each domain must
127 follow a given order which is specified in the oprofile user
128 manual. If more than one domain is specified in <list> they
129 should be separated using commas. This option can only be used
130 in domain 0 which is the only domain that can coordinate a
131 multi-domain profiling session. Including domain 0 in the list
132 of active domains is optional. (e.g. --active-domains=2,5,6 and
133 --active-domains=0,2,5,6 are equivalent). This option can only
134 be specified if --start-daemon is also specified and it is only
135 valid for the current run of the oprofile daemon; e.g. the list
136 of active domains is not persistent.
137 --passive-domains=<list>or--domains=<list>
139 List of domain ids to be profiled, separated by commas. As
140 opposed to the --active-domains option, the domains specified
141 with this option do not need to run oprofile. This makes profil‐
142 ing multiple domains easier. However, with the passive-domains
143 option, samples in user level processes and kernel modules can‐
144 not be mapped to specific symbols and are aggregated under a
145 generic class. Both --active-domains and --passive-domains
146 options can be specified in the same command, but the same
147 domain cannot be specified in both options. This option can only
148 be specified if either --start or --start-daemon is specified on
149 the same command and it is only valid for the current run of the
150 oprofile daemon; e.g. the list of passive domains is not persis‐
151 tent.
152 --passive-images=<list>or--domains-images=<list>
154 List of kernel images associated with the domains specified in
155 the --passive-domains option, also separated by commas. The
156 association between the images and domains is based on the order
157 they are specified in both options.
158
161 No special environment variables are recognised by opcontrol.
162
165 /root/.oprofile/daemonrc
166 Configuration file for opcontrol
167 /var/lib/oprofile/samples/
169 The location of the generated sample files.
170
173 This man page is current for oprofile-0.9.6.
174
177 /usr/share/doc/oprofile-0.9.6/, oprofile(1)
1784th Berkeley Distribution Tue 07 June 2011 OPCONTROL(1)