We are happy to announce the availability of the latest development
version of DTrace for Linux (2.0.0-1.8).
This new version is based on BPF and other Linux kernel tracing features
and is implemented entirely as a userspace application.
The functionality is still limited in comparison to the kernel module
based version of DTrace for Linux (currently at version 1.2.1-1).Â
Development continues in an incremental fashion to make the full feature
set of DTrace available using existing kernel features.
WHERE TO FIND IT?
The new version of DTrace for Linux is available at:
   https://github.com/oracle/dtrace-utils/tree/2.0-branch-dev
The most recent release tag is 2.0.0-1.8.
FEATURES
 - BEGIN and END probe (with full support for the expected semantics).
 - ERROR probe support.
 - Function Boundary Tracing (FBT) probes.
 - System call entry and exit probes.
 - Timer-based profile-* and tick-* probes.
 - Statically Defined Tracing (SDT) probes for kernel tracepoints.
 - Arguments for SDT probes can be retrieved with arg0-arg9.
 - pid provider support for userspace function boundary tracing.
 - Aggregated data collection support.
 - Aggregation functions: avg, count, llquantize, lquantize, max, min,
   quantize, stddev, and sum.
 - Aggregation actions: normalize() and denormalize().
 - Pretty-printing for aggregations using the printa() action.
 - Support for local and global variables with storage sizes greater
   than 8 bytes. This adds struct and union support for local and global
   variables.
 - Various features to support stack tracing.
 - Direct compilation of D source code into BPF programs.
 - Efficient use of pre-compiled BPF functions for library functions.
 - A bpflog option to request the BPF verifier log for loaded programs.
 - BPF program linking of dynamically generated code and pre-compiled code
   to facilitate code sharing and code re-use.
 - Improved integrated disassembler for generated BPF code at the
   clause and program level (-S in combination with the new -xdisasm=#\
   option).
 - Improved trace data buffer handling based on memory mapped perf event
   ring-buffers.
 - Deferred creation of kernel level probes. Listing available probes
does
   not require them to be created at the kernel level.
- [NEW] Speculative tracing: speculation(), speculate(), commit(), and
discard().
 - [NEW] Nearly complete support for strings: string constants in D clauses,
assignment, built-in variables, comparison operators, trace(), strlen(),
strchr(), strrchr(), index(), rindex(), strstr(), strjoin(), substr(),
and lltostr().
- [NEW] Host/network byte order conversion: htonl(), htonll(), htons(),
ntohl(), ntohll(), and ntohs().
- [NEW] Symbol resolving for compressed kernel modules.
- [NEW] Support to run dtrace under valgrind.
DEPENDENCIES
DTrace for Linux depends on libctf (part of newer binutils) -or-
libdtrace-ctf. While libctf is preferred, building against
libdtrace-ctf is fully supported. It can be found at:
   https://github.com/oracle/libdtrace-ctf
DTrace for Linux makes use of BPF library functions that are compiled at
build time. It depends on BPF support in GCC and binutils to generate
the pre-compiled BPF function library.
DTrace for Linux makes use of 3 additional features that are not
commonly available in the Linux kernel:
 - CTF data generation at compile time: this provides important datatype
   information for kernel and kernel module symbols.
 - waitfd() system call: this provides a mechanism to monitor for state
   changes of processes that are being traced.
 - /proc/kallsyms module grouping support: this adds annotation with module
   names to all symbols that are part of a kernel module even if the module
   is configured to be compiled into the kernel proper. This allows for
   consistent ways to refer to probes by module and function (or probe) name.
These additional support features for tracing are available at:
   https://github.com/oracle/dtrace-linux-kernel/tree/v2/5.14.9
Please consider joining our development list: dtrace-devel at oss.oracle.com
   Enjoy!
|