With recent kernel tracepoints for mmap/process tracking, as used
by the perf tool, we may be able to use a non-utrace backend for
the task_finder. That in turn could enable utrace-free user-space
<sdt.h> probing, using the kprobe or similar non-utrace backend.
As of commit c0f710f991 (which merged the dsmith/task_finder2 branch into the master branch), there is a non-utrace backend for the task_finder for kernels without utrace.
A kernel must support specific tracepoints (CONFIG_TRACEPOINTS) for this to work:
and one of the following:
- the not quite upstream 'sched_process_exec' tracepoint; or
- the ftrace_set_filter() function (which depends on CONFIG_FTRACE)