A new development of the kernel Thomas Gleixner and Ingo Molnar, promises to illustrate events in the Linux system is better than traditional tools.
The technique called “trace” tool presented offers, according to Thomas Gleixner a way the events in the Linux system and better able to pursue study. Although there is already using tools like ” strace “a way to trace system events, but according to this Gleixner not satisfactory.
“Trace” is now to fill this gap and give users an easy way to examining the system. The tool serves as a basis for measuring speed which actually presented Perf architecture that allows easier access to kernel internals, and already provided a variety of profiling tools entails. Some of the features are already implemented in the kernel, but others have to be installed first. Gleixner therefore recommends the installation of the tip kernels , the functionality of “trace” to be able to test fully. The minimum requirement on the kernel side represents the support of syscall events (CONFIG_FTRACE_SYSCALLS) dar.
The philosophy of the new tools that are first collected information about a system and evaluated at a later date. It does not matter what information you have. “Trace” is clearly both individual processes and groups of processes, system-wide CPU calls and calls. It records the system calls, tasks, events (fork, clone, Exit), page faults, and mmap and scheduler events. Maintains it, the tool, in contrast to strace broken in page faults, including the complete mapping of the stack and heap. Existing sessions or traces may also examined at a later date and are filtered according to your needs.
How Gleixner wrote in his announcement, as the “trace” at the beginning of development and will be expanded gradually in the future. Among other things planned that it has considered network or block IO events and more trace points, especially at higher levels.
