4 _by [Mathieu Desnoyers](mailto:mathieu.desnoyers@efficios.com)_
7 LTTng kernel modules are Linux kernel modules which make
8 [LTTng](http://lttng.org/) kernel tracing possible. They include
9 essential control modules and many probes which instrument numerous
10 interesting parts of Linux. LTTng-modules builds against a vanilla or
11 distribution kernel, with no need for additional patches.
13 Other notable features:
15 - Produces [CTF](http://www.efficios.com/ctf)
16 (Common Trace Format) natively,
17 - Tracepoints, function tracer, CPU Performance Monitoring Unit (PMU)
18 counters, kprobes, and kretprobes support,
19 - Have the ability to attach _context_ information to events in the
20 trace (e.g., any PMU counter, PID, PPID, TID, command name, etc).
21 All the extra information fields to be collected with events are
22 optional, specified on a per-tracing-session basis (except for
23 timestamp and event ID, which are mandatory).
29 To build and install LTTng-modules, you will need to have your kernel
30 headers available (or access to your full kernel source tree), and do:
33 sudo make modules_install
36 The above commands will build LTTng-modules against your
37 current kernel. If you need to build LTTng-modules against a custom
40 make KERNELDIR=/path/to/custom/kernel
41 sudo make KERNELDIR=/path/to/custom/kernel modules_install
42 sudo depmod -a kernel_version
45 ### Kernel built-in support
47 It is also possible to build these modules as part of a kernel image. Simply
48 run the [`scripts/built-in.sh`](scripts/built-in.sh) script with the path to
49 your kernel source directory as an argument. It will symlink the
50 lttng-modules directory in the kernel sources and add an include in the kernel
53 Then configure your kernel as usual and enable the `CONFIG_LTTNG` option.
56 ### Required kernel config options
58 Make sure your target kernel has the following config options enabled:
60 - `CONFIG_MODULES`: loadable module support (not strictly required
61 when built into the kernel),
62 - `CONFIG_KALLSYMS`: see files in [`wrapper`](wrapper); this is
63 necessary until the few required missing symbols are exported to GPL
64 modules from mainline,
65 - `CONFIG_HIGH_RES_TIMERS`: needed for LTTng 2.x clock source,
66 - `CONFIG_TRACEPOINTS`: kernel tracepoint instrumentation
67 (enabled as a side-effect of any of the perf/ftrace/blktrace
68 instrumentation features).
69 - `CONFIG_KPROBES` (5.7+): use kallsyms for kernel 5.7 and newer.
72 ### Supported (optional) kernel config options
74 The following kernel configuration options will affect the features
77 - `CONFIG_HAVE_SYSCALL_TRACEPOINTS`: system call tracing:
79 lttng enable-event -k --syscall
80 lttng enable-event -k -a
82 - `CONFIG_PERF_EVENTS`: performance counters:
84 lttng add-context -t perf:*
86 - `CONFIG_EVENT_TRACING`: needed to allow block layer tracing
87 - `CONFIG_KPROBES`: dynamic probes:
89 lttng enable-event -k --probe ...
91 - `CONFIG_KRETPROBES`: dynamic function entry/return probes:
93 lttng enable-event -k --function ...
95 - `CONFIG_KALLSYMS_ALL`: state dump of mapping between block device
98 Customization/Extension
99 -----------------------
101 The lttng-modules source includes definitions for the actual callback
102 functions that will be attached to the kernel tracepoints by lttng.
103 The lttng-modules project implements its own macros generating these
104 callbacks: the LTTNG_TRACEPOINT_EVENT macro family found in
105 instrumentation/events/lttng-module/. In order to show up in a
106 lttng-modules trace, a kernel tracepoint must be defined within the
107 kernel tree, and also defined within lttng-modules with the
108 LTTNG_TRACEPOINT_EVENT macro family. Customizations or extensions must
109 be done by modifying instances of these macros within the lttng-modules
115 Use [LTTng-tools](https://lttng.org/download) to control the tracer.
116 The session daemon of LTTng-tools should automatically load the LTTng
117 kernel modules when needed. Use [Babeltrace](https://lttng.org/babeltrace)
118 to print traces as a human-readable text log.
124 Linux kernels >= 3.0 are supported.
130 ### About perf PMU counters support
132 Each PMU counter has its zero value set when it is attached to a context with
133 add-context. Therefore, it is normal that the same counters attached to both the
134 stream context and event context show different values for a given event; what
135 matters is that they increment at the same rate.