Documenting kernel tracepoints
There are a number of tracepoints in the latest kernels, but not much documentation on the tracepoints. If a very recent version of systemtap is available on the system, a list of the probe points can be obtained with:
stap -p2 -e 'probe kernel.trace("*") {exit()}'|sort
However, this only provides the names. It doesn't provide information about what information the probe point provides or the arguments available at the probe point.
Currently, a number of kernel functions and structures are documented with embedded comments that are extracted with kernel-doc. Seems like it would be reasonable to extend this to support tracepoints. Any thoughts or comments about this approach?
-Will
Quoted 7 lines · William Cohen
There are a number of tracepoints in the latest kernels, but not much documentation on the tracepoints. If a very recent version of systemtap is available on the system, a list of the probe points can be obtained with:
stap -p2 -e 'probe
kernel.trace("*"){exit()}'|sort
Quoted 8 lines
However, this only provides the names. It doesn't provide information about what information the probe point provides or the arguments available at the probe point.
Currently, a number of kernel functions and structures are documented with embedded comments that are extracted with kernel-doc. Seems like it would be reasonable to extend this to support tracepoints. Any thoughts or comments about this approach?
FYI, in the latest tracing tree (targeted for 2.6.30) all tracepoints show up under /debug/tracing/events/.
Ingo
Quoted 18 lines · Ingo Molnar
Quoted 7 lines · William Cohen
There are a number of tracepoints in the latest kernels, but not much documentation on the tracepoints. If a very recent version of systemtap is available on the system, a list of the probe points can be obtained with:
stap -p2 -e 'probe
kernel.trace("*"){exit()}'|sortQuoted 8 lines
However, this only provides the names. It doesn't provide information about what information the probe point provides or the arguments available at the probe point.
Currently, a number of kernel functions and structures are documented with embedded comments that are extracted with kernel-doc. Seems like it would be reasonable to extend this to support tracepoints. Any thoughts or comments about this approach?
Quoted 4 lines
FYI, in the latest tracing tree (targeted for 2.6.30) all tracepoints show up under /debug/tracing/events/.
Ingo
The /debug/tracing/events provides event names and information about the available arguments. However, it doesn't provide the detail comparable to kerneldoc embedded comments summarizing the tracepoint's purpose or what the arguments are.
-Will