The release of the dynamic tracing system SystemTap 5.0 has been announced, providing debugging tools for the Linux platform similar to the DTrace technology. SystemTap allows for in-depth monitoring of a running Linux system, gathering statistics about application performance, profiling, and controlling system calls. Management is conducted through a command-line interface and a specialized C-like scripting language. The system is compatible with Linux kernels from version 3.10 to 6.6-rc0.
Companies such as Red Hat, IBM, Intel, Hitachi, and Oracle are involved in the development of the project. The catalog of examples features more than 190 scripts for various use cases suitable for monitoring memory distribution, input/output, disk operations, network traffic (for example, analyzing NFS performance), task scheduler operation, interrupt handling, use of system buffers, setting locks, executing system calls, signal processing, and more.
Key Changes:
- Support for the kernel branch 2.6.32 used in RHEL6 has been discontinued. Testing of SystemTap is now conducted with kernels 3.10 (RHEL7), 4.18 (RHEL8 + CentOS Stream 8), 5.14.0 (RHEL9 + CentOS Stream 9), 6.4 (Fedora 38), 6.5 (Fedora 39), and 6.6-rc (Fedora rawhide).
- Support has been added for the debuginfod process for dynamic retrieval of debugging information and metadata about installed executable files.
- Support has been added for debugging data in DWARF5 format, which can be used in the function print_ubacktrace_fileline().
- In the frontend (utility stap), when debuginfod is available, servers there is the capability to use masks when specifying target processes, for example, stap -e ‘probe debuginfod.process("/usr/*/curl").begin { log("hi")} . The details of the process information in the list output using the ‘-L’ option in conjunction with multiple ‘-v’ have been improved.
- In the backend, the performance of uprobe registration has increased and the launch of the kernel module has been accelerated. A faster and more reliable messaging system between the kernel and user space has been proposed.
- Run-time macros have been added: STP_TIMING_NSECS for obtaining check timing information in nanoseconds and STP_FORCE_STDOUT_TTY for overriding STP_STDOUT_NOT_ATTY.
Source: opennet.ru
