1 * What's new in version 2.6
3 - SystemTap now supports SDT probes with operands that refer to symbols.
5 - While in listing mode (-l/-L), probes printed are now more consistent
8 - Statement probes now support enumerated linenos to probe discontiguous
9 linenos using the form:
11 process.statement("foo@file.c:3,5-7,9")
13 - Statement counting is now suppressed in the generated c code for probes that
14 are non-recursive and loop-free. Statement counting can be turned back on in
17 * What's new in version 2.5, 2014-04-30
19 - Systemtap now supports backtracing through its own, invoking module.
21 - Java probes now support backtracing using the print_java_backtrace()
22 and sprint_java_backtrace() functions.
24 - Statement probes (e.g. process.statement) are now faster to resolve,
25 more precise, and work better with inlined functions.
27 - New switches have been added to help inspect the contents of installed
30 stap --dump-functions --> list all library functions and their args
31 stap --dump-probe-aliases --> list all library probe aliases
33 - The heuristic algorithms used to search for function-prologue
34 endings were improved, to cover more optimization (or
35 lack-of-optimization, or incorrect-debuginfo) cases. These
36 heuristics are necessary to find $context parameters for some
37 function-call/entry probes. We recommend programs be built with
38 CFLAGS+=-grecord-gcc-switches to feed information to the heuristics.
40 - The stap --use-server option now more correctly supports address:port
41 type parametrization, for manual use in the absence of avahi.
43 - A new probe alias "oneshot" allows a single quick script fragment to run,
46 - The argv tapset now merges translate-time and run-time positional
47 arguments, so all of these work:
49 stap -e 'probe oneshot {println(argv[1]," ",argv[2])}' hello world
51 stap -e 'probe oneshot {println(argv[1]," ",argv[2])}' \
52 -G argv_1=hello -G argv_2=world
54 staprun hello.ko argv_1=hello argv_2=world
56 - SystemTap now falls back on the symbol table for probing
57 functions in processes if the debuginfo is not available.
59 - SystemTap now supports a %( guru_mode == 0 /* or 1 */ %)
60 conditional for making dual-use scripts.
62 - SystemTap now supports UEFI/SecureBoot systems, via
63 machine-owner-keys maintained by a trusted stap-server on the
64 network. (Key enrollment requires a one-time reboot and BIOS
67 - SystemTap now reports more accurate and succinct errors on type
70 - Embedded-C functions may use STAP_RETURN(value) instead of the
71 more wordy STAP_RETVALUE assignment followed by a "goto out".
72 The macro supports numeric or string values as appropriate.
74 - Some struct-sockaddr fields are now individually decoded for
75 socket-related syscalls:
76 probe syscall.connect { println (uaddr_af, ":", uaddr_ip) }
78 - The documentation for the SystemTap initscript service and the
79 SystemTap compile-server service have been completely converted from
80 README files to man pages (see systemtap(8) and stap-server(8)).
82 - SystemTap is now capable of inserting modules early during the boot
83 process on dracut-based systems. See the 'onboot' command in
84 systemtap(8) for more information.
86 - DWARF probes can now use the '.callee[s]' variants, which allow more
87 precise function probing. For example, the probe point
89 process("myproc").function("foo").callee("bar")
91 will fire upon entering bar() from foo(). A '.callees' probe will
92 instead place probes on all callees of foo().
93 Note that this also means that probe point wildcards should be used
94 with more care. For example, use signal.*.return rather than
95 signal.*.*, which would also match '.callees'. See stapprobes(3stap)
96 for more info. This feature requires at least GCC 4.7.
98 - A few new functions in the task_time tapsets, as well as a new tapset
99 function task_ancestry(), which prints out the parentage of a process.
101 - The kprocess.exec probe has been updated to use syscall.execve, which
102 allows access to the new process' arguments (through the new 'argstr'
103 or 'args' variables) as well as giving better support across kernel
104 versions. Note also that the 'filename' variable now holds the
105 filename (quoted), or the address (unquoted) if it couldn't be
108 - The [s]println() function can now be called without any arguments to
109 simply print a newline.
111 - Suggestions are now provided when markers could not be resolved. For
112 example, process("stap").mark("benchmart") will suggest 'benchmark'.
114 - SystemTap colors can now be turned off by simply setting
115 SYSTEMTAP_COLORS to be empty, rather than having to make it invalid.
117 - There is a new context tapset function, pnlabel(), which returns the
118 name of the label which fired.
120 - The following tapset variables and functions are deprecated in
122 - The 'clone_flags', 'stack_start', 'stack_size',
123 'parent_tid_uaddr', and 'child_tid_uaddr' variables in the
124 'syscall.fork' and 'nd_syscall.fork' probe aliases.
125 - The '_sendflags_str()' and '_recvflags_str()' functions have been
126 deprecated in favor of the new '_msg_flags_str()' function.
127 - The 'flags' and 'flags_str' variables in the 'syscall.accept' and
128 'nd_syscall.accept' probe alias.
129 - The 'first', 'second', and 'uptr_uaddr' variables in the
130 'syscall.compat_sys_shmctl', and 'nd_syscall.compat_sys_shmctl'
131 probe aliases have been deprecated in favor of the new 'shmid',
132 'cmd', and 'buf_uaddr' variables.
134 * What's new in version 2.4, 2013-11-06
136 - Better suggestions are given in many of the semantic errors in which
137 alternatives are provided. Additionally, suggestions are now provided
138 when plt and trace probes could not be resolved. For example,
139 kernel.trace("sched_siwtch") will suggest 'sched_switch'.
141 - SystemTap is now smarter about error reporting. Errors from the same
142 source are considered duplicates and suppressed. A message is
143 displayed on exit if any errors/warnings were suppressed.
145 - Statistics aggregate typed objects are now implemented locklessly,
146 if the translator finds that they are only ever read (using the
147 foreach / @count / etc. constructs) in a probe-begin/end/error.
149 - SystemTap now supports probing inside virtual machines using the
150 libvirt and unix schemes, e.g.
152 stap -ve 'probe timer.s(1) { printf("hello!\n") }' \
153 --remote=libvirt://MyVirtualMachine
155 Virtual machines managed by libvirt can be prepared using stapvirt.
156 See stapvirt(1) and the --remote option in stap(1) for more details.
158 - Systemtap now checks for and uses (when available) the .gnu_debugdata
159 section which contains a subset of debuginfo, useful for backtraces
162 - SystemTap map variables are now allocated with vmalloc() instead of
163 with kmalloc(), which should cause memory to be less fragmented.
165 - Although SystemTap itself requires elfutils 0.148+, staprun only
166 requires elfutils 0.142+, which could be useful with the
167 '--disable-translator' configure switch.
169 - Under FIPS mode (/proc/sys/crypto/fips_enabled=1), staprun will
170 refuse to load systemtap modules (since these are not normally
171 signed with the kernel's build-time keys). This protection may
172 be suppressed with the $STAP_FIPS_OVERRIDE environment variable.
174 - The stap-server client & server code now enable all SSL/TLS
175 ciphers rather than just the "export" subset.
177 - For systems with in-kernel utrace, 'process.end' and 'thread.end'
178 probes will hit before the target's parent process is notified of
179 the target's death. This matches the behavior of newer kernels
180 without in-kernel utrace.
182 * What's new in version 2.3, 2013-07-25
184 - More context-accessing functions throw systemtap exceptions upon a
185 failure, whereas in previous versions they might return non-error
186 sentinel values like "" or "<unknown>". Use try { } / catch { }
187 around these, or new wrapper functions such as user_string_{n_}quoted()
188 that internally absorb exceptions.
190 - java("org.my.MyApp") probes are now restricted to pre-existing jvm pid's with
191 a listing in jps -l output to avoid recursive calls
193 - The tapset [nd_]syscall.semop parameter tsops_uaddr is renamed sops_uaddr for
194 consistency with [nd_]syscall.semtimedop.
196 - The udp.stp tapset adds some ip-address/port variables.
198 - A new guru-mode-only tapset function raise() is available to send signals
201 - Support for the standard Posix ERE named character classes has been
202 added to the regexp engine, e.g. [:digit:], [:alpha:], ...
204 - A substantial internal overhaul of the regexp engine has resulted in
205 correct behaviour on further obscure edge cases. The regexp engine
206 now implements the ERE standard and correctly passes the testsuite
207 for the glibc regexp engine (minus portions corresponding to
208 unimplemented features -- i.e. subexpression capture and reuse).
210 - Alternative functions are now suggested when function probes could not be
211 resolved. For example, kernel.function("vfs_reads") will suggest vfs_read.
212 Other probes for which suggestions are made are module.function,
213 process.function, and process.library.function.
215 - Has life been a bit bland lately? Want to spice things up? Why not write a
216 few faulty probes and feast your eyes upon the myriad of colours adorning
217 your terminal as SystemTap softly whispers in your ear... 'parse error'.
218 Search for '--color' in 'man stap' for more info.
220 - The following tapset functions are deprecated in release 2.3:
221 'stap_NFS_CLIENT', '__getfh_inode', '_success_check',
222 '_sock_prot_num', '_sock_fam_num', '_sock_state_num',
223 '_sock_type_num', and '_sock_flags_num'.
225 * What's new in version 2.2.1, 2013-05-16
226 * What's new in version 2.2, 2013-05-14
228 - Experimental support has been added for probing Java methods using
229 Byteman 2.0 as a backend. Java method probes can target method entries,
230 returns, or specific statements in the method as specified by line number.
232 probe java("org.my.MyApp").class("^java.lang.Object").method("foo(int)")
235 See java/README for information on how to set up Java/Byteman
236 functionality. Set env STAPBM_VERBOSE=yes for more tracing.
238 - The stap -l output and pn() tapset function's return value may be slightly
239 different for complicated web of wildcarded/aliased probes.
241 - The dyninst backend has improved in several aspects:
243 - Setting custom values for global variables is now supported, both
244 with -G when compiling a script, and from the stapdyn command line
245 when loading a precompiled module.
247 - A high-performance shared-memory-based transport is used for
250 - A systemd service file and tmpfile have been added to allow
251 systemtap-server to be managed natively by systemd.
253 - Due to the removal of register_timer_hook in recent kernels, the
254 behaviour of timer.profile has been changed slightly. This probe is
255 now an alias which uses the old mechanism where possible, but falls
256 back to perf.sw.cpu_clock or another mechanism when the kernel timer
257 hook is not available.
259 To require the kernel timer hook mechanism in your script, use
260 timer.profile.tick instead of timer.profile.
262 - The following tapset variables are deprecated in release 2.2:
263 - The 'origin' variables in the 'generic.fop.llseek',
264 'generic.fop.llseek.return', and 'nfs.fop.llseek' probes. The
265 'origin' variable has been replaced by the 'whence' variable.
266 - The 'page_index' variable in the 'vfs.block_sync_page' and
267 'vfs.buffer_migrate_page' probe aliases.
268 - The 'write_from' and 'write_upto' variables in the
269 '_vfs.block_prepare_write' and '_vfs.block_prepare_write.return'
271 - The 'regs' variable in the 'syscall.sigaltstack',
272 'nd_syscall.sigaltstack', 'syscall.fork', and 'nd_syscall.fork'
274 - The 'first', 'second', 'third', and 'uptr_uaddr' variables in the
275 'syscall.compat_sys_shmat' and 'nd_syscall.compat_sys_shmat' probe
278 - The following tapset functions are deprecated in release 2.2:
279 'ppos_pos', '_dev_minor', and '_dev_major'
281 - The folowing tapset functions used to return error strings instead
282 of raising an error. The original behavior is deprecated in release
285 'ctime', 'probemod', 'modname'
287 * What's new in version 2.1, 2013-02-13
289 - EMACS and VIM editor modes for systemtap source files are included / updated.
291 - The translator now eliminates duplicate tapset files between its
292 preferred directory (as configured during the build with --prefix=/
293 or specified with the -I /path option), and files it may find under
294 $XDG_DATA_DIRS. This should eliminate a class of conflicts between
295 parallel system- and hand-built systemtap installations.
297 - The translator accepts a --suppress-time-limits option, which defeats
298 time-related constraints, to allows probe handlers to run for indefinite
299 periods. It requires the guru mode (-g) flag to work. Add the earlier
300 --suppress-handler-errors flag for a gung-ho "just-keep-going" attitude.
302 - Perf event probes may now be read on demand. The counter probe is
303 defined using the counter-name part:
304 probe perf.type(0).config(0).counter("NAME"). The counter is
305 read in a user space probe using @perf("NAME"), e.g.
306 process("PROCESS").statement("func@file") {stat <<< @perf("NAME")}
308 - Perf event probes may now be bound to a specific task using the
309 process-name part: probe perf.type(0).config(0).process("NAME") { }
310 If the probed process name is not specified, then it is inferred
311 from the -c CMD argument.
313 - Some error messages and warnings now refer to additional information
314 that is found in man pages. These are generally named
315 error::FOO or warning::BAR (in the 7stap man page section)
320 - The dyninst backend has improved in several aspects:
321 - The runtime now allows much more concurrency when probing multithreaded
322 processes, and will also follow probes across forks.
323 - Several new probe types are now supported, including timers, function
324 return, and process.begin/end and process.thread.begin/end.
325 - Semaphores for SDT probes are now set properly.
326 - Attaching to existing processes with -x PID now works.
328 - The foreach looping construct can now sort aggregate arrays by the user's
329 choice of aggregating function. Previously, @count was implied. e.g.:
330 foreach ([x,y] in array @sum +) { println(@sum(array[x,y])) }
332 - Proof of concept support for regular expression matching has been added:
333 if ("aqqqqqb" =~ "q*b") { ... }
334 if ("abc" !~ "q*b") { ... }
336 The eventual aim is to support roughly the same functionality as
337 the POSIX Extended Regular Expressions implemented by glibc.
338 Currently missing features include extraction of the matched string
339 and subexpressions, and named character classes ([:alpha:], [:digit:], &c).
341 Special thanks go to the re2c project, whose public domain code this
342 functionality has been based on. For more info on re2c, see:
343 http://sourceforge.net/projects/re2c/
345 - The folowing tapset variables are deprecated in release 2.1 and will
346 be removed in release 2.2:
347 - The 'send2queue' variable in the 'signal.send' probe.
348 - The 'oldset_addr' and 'regs' variables in the 'signal.handle' probe.
350 - The following tapset probes are deprecated in release 2.1 and will
351 be removed in release 2.2:
353 - signal.handle.return
355 * What's new in version 2.0, 2012-10-09
357 - Systemtap includes a new prototype backend, which uses Dyninst to instrument
358 a user's own processes at runtime. This backend does not use kernel modules,
359 and does not require root privileges, but is restricted with respect to the
360 kinds of probes and other constructs that a script may use.
362 Users from source should configure --with-dyninst and install a
363 fresh dyninst snapshot such as that in Fedora rawhide. It may be
364 necessary to disable conflicting selinux checks; systemtap will advise.
366 Select this new backend with the new stap option --runtime=dyninst
367 and a -c target process, along with normal options. (-x target
368 processes are not supported in this prototype version.) For example:
370 stap --runtime=dyninst -c 'stap -l begin' \
371 -e 'probe process.function("main") { println("hi from dyninst!") }'
373 - To aid diagnosis, when a kernel panic occurs systemtap now uses
374 the panic_notifier_list facility to dump a summary of its trace
375 buffers to the serial console.
377 - The systemtap preprocessor now has a simple macro facility as follows:
379 @define add(a,b) %( ((@a)+(@b)) %)
380 @define probegin(x) %(
386 @probegin( foo = @add(40, 2); print(foo) )
388 Macros defined in the user script and regular tapset .stp files are
389 local to the file. To get around this, the tapset library can define
390 globally visible 'library macros' inside .stpm files. (A .stpm file
391 must contain a series of @define directives and nothing else.)
393 The status of the feature is experimental; semantics of macroexpansion
394 may change (unlikely) or expand in the future.
396 - Systemtap probe aliases may be used with additional suffixes
397 attached. The suffixes are passed on to the underlying probe
398 point(s) as shown below:
400 probe foo = bar, baz { }
401 probe foo.subfoo.option("gronk") { }
402 // expands to: bar.subfoo.option("gronk"), baz.subfoo.option("gronk")
404 In practical terms, this allows us to specify additional options to
405 certain tapset probe aliases, by writing e.g.
406 probe syscall.open.return.maxactive(5) { ... }
408 - To support the possibility of separate kernel and dyninst backends,
409 the tapsets have been reorganized into separate folders according to
410 backend. Thus kernel-specific tapsets are located under linux/, the
411 dyninst-specific ones under dyninst/
413 - The backtrace/unwind tapsets have been expanded to allow random
414 access to individual elements of the backtrace. (A caching mechanism
415 ensures that the backtrace computation run at most once for each
416 time a probe fires, regardless of how many times or what order the
417 query functions are called in.) New tapset functions are:
418 stack/ustack - return n'th element of backtrace
419 callers/ucallers - return first n elements of backtrace
420 print_syms/print_usyms - print full information on a list of symbols
421 sprint_syms/sprint_usyms - as above, but return info as a string
423 The following existing functions have been superseded by print_syms()
424 et al.; new scripts are recommended to avoid using them:
430 - The probefunc() tapset function is now myproc-unprivileged, and can
431 now be used in unprivileged scripts for such things as profiling in
432 userspace programs. For instance, try running
433 systemtap.examples/general/para-callgraph.stp in unprivileged mode
434 with a stapusr-permitted probe. The previous implementation of
435 probefunc() is available with "stap --compatible=1.8".
437 - Preprocessor conditional to vary code based on script privilege level:
438 unprivileged -- %( systemtap_privilege == "stapusr" %? ... %)
439 privileged -- %( systemtap_privilege != "stapusr" %? ... %)
440 or, alternately %( systemtap_privilege == "stapsys"
441 || systemtap_privilege == "stapdev" %? ... %)
443 - To ease migration to the embedded-C locals syntax introduced in 1.8
444 (namely, STAP_ARG_* and STAP_RETVALUE), the old syntax can now be
445 re-enabled on a per-function basis using the /* unmangled */ pragma:
447 function add_foo:long(a:long, b:long) %{ /* unmangled */
448 THIS->__retvalue = THIS->a + STAP_ARG_b;
451 Note that both the old and the new syntax may be used in an
452 /* unmangled */ function. Functions not marked /* unmangled */
453 can only use the new syntax.
455 - Adjacent string literals are now glued together irrespective of
456 intervening whitespace or comments:
457 "foo " "bar" --> "foo bar"
458 "foo " /* comment */ "bar" --> "foo bar"
459 Previously, the first pair of literals would be glued correctly,
460 while the second would cause a syntax error.
462 * What's new in version 1.8, 2012-06-17
464 - staprun accepts a -T timeout option to allow less frequent wake-ups
465 to poll for low-throughput output from scripts.
467 - When invoked by systemtap, the kbuild $PATH environment is sanitized
468 (prefixed with /usr/bin:/bin:) in an attempt to exclude compilers
469 other than the one the kernel was presumed built with.
471 - Printf formats can now use "%#c" to escape non-printing characters.
473 - Pretty-printed bitfields use integers and chars use escaped formatting
476 - The systemtap compile-server and client now support IPv6 networks.
477 - IPv6 addresses may now be specified on the --use-server option and will
478 be displayed by --list-servers, if the avahi-daemon service is running and
480 - Automatic server selection will automatically choose IPv4 or IPv6 servers
481 according to the normal server selection criteria when avahi-daemon is
482 running. One is not preferred over the other.
483 - The compile-server will automatically listen on IPv6 addresses, if
485 - To enable IPv6 in avahi-daemon, ensure that /etc/avahi/avahi-daemon.conf
486 contains an active "use-ipv6=yes" line. After adding this line run
487 "service avahi-daemon restart" to activate IPv6 support.
488 - See man stap(1) for details on how to use IPv6 addresses with the
491 - Support for DWARF4 .debug_types sections (for executables and shared
492 libraries compiled with recent GCC's -gdwarf-4 / -fdebug-types-section).
493 PR12997. SystemTap now requires elfutils 0.148+, full .debug_types support
494 depends on elfutils 0.154+.
496 - Systemtap modules are somewhat smaller & faster to compile. Their
497 debuginfo is now suppressed by default; use -B CONFIG_DEBUG_INFO=y to
500 - @var now an alternative language syntax for accessing DWARF variables
501 in uprobe and kprobe handlers (process, kernel, module). @var("somevar")
502 can be used where $somevar can be used. The @var syntax also makes it
503 possible to access non-local, global compile unit (CU) variables by
504 specifying the CU source file as follows @var("somevar@some/src/file.c").
505 This will provide the target variable value of global "somevar" as defined
506 in the source file "some/src/file.c". The @var syntax combines with all
507 normal features of DWARF target variables like @defined(), @entry(),
508 [N] array indexing, field access through ->, taking the address with
509 the & prefix and shallow or deep pretty printing with a $ or $$ suffix.
511 - Stap now has resource limit options:
517 All resource limiting has been moved from the compile server to stap
518 itself. When running the server as "stap-server", default resource
519 limit values are specified in ~stap-server/.systemtap/rc.
521 - Bug CVE-2012-0875 (kernel panic when processing malformed DWARF unwind data)
524 - The systemtap compile-server now supports multiple concurrent connections.
525 Specify the desired maximum number of concurrent connections with
526 the new stap-server/stap-serverd --max-threads option. Specify a
527 value of '0' to tell the server not to spawn any new threads (handle
528 all connections serially in the main thread). The default value is
529 the number of processor cores on the host.
531 - The following tapset functions are deprecated in release 1.8 and will be
532 removed in release 1.9:
535 - SystemTap now mangles local variables to avoid collisions with C
536 headers included by tapsets. This required a change in how
537 embedded-C functions access local parameters and the return value slot.
539 Instead of THIS->foo in an embedded-C function, please use the newly
540 defined macro STAP_ARG_foo (substitute the actual name of the
541 argument for 'foo'); instead of THIS->__retvalue, use the newly
542 defined STAP_RETVALUE. All of the tapsets and test cases have been
543 adapted to use this new notation.
545 If you need to run code which uses the old THIS-> notation, run stap
546 with the --compatible=1.7 option.
548 - There is updated support for user-space probing against kernels >=
549 3.5, which have no utrace but do have the newer inode-uprobes work
550 by Srikar Dronamraju and colleagues. For kernels < 3.5, the
551 following 3 sets of kernel patches would need to be backported to
552 your kernel to use this preliminary user-space probing support:
554 - inode-uprobes patches:
555 - 2b144498350860b6ee9dc57ff27a93ad488de5dc
556 - 7b2d81d48a2d8e37efb6ce7b4d5ef58822b30d89
557 - a5f4374a9610fd7286c2164d4e680436727eff71
558 - 04a3d984d32e47983770d314cdb4e4d8f38fccb7
559 - 96379f60075c75b261328aa7830ef8aa158247ac
560 - 3ff54efdfaace9e9b2b7c1959a865be6b91de96c
561 - 35aa621b5ab9d08767f7bc8d209b696df281d715
562 - 900771a483ef28915a48066d7895d8252315607a
563 - e3343e6a2819ff5d0dfc4bb5c9fb7f9a4d04da73
564 - exec tracepoint kernel patch:
565 - 4ff16c25e2cc48cbe6956e356c38a25ac063a64d
566 - task_work_add kernel patches:
567 - e73f8959af0439d114847eab5a8a5ce48f1217c4
568 - 4d1d61a6b203d957777d73fcebf19d90b038b5b2
569 - 413cd3d9abeaef590e5ce00564f7a443165db238
570 - dea649b8ac1861107c5d91e1a71121434fc64193
571 - f23ca335462e3c84f13270b9e65f83936068ec2c
573 * What's new in version 1.7, 2012-02-01
575 - Map inserting and deleting is now significantly faster due to
576 improved hashing and larger hash tables. The hashes are also
577 now randomized to provide better protection against deliberate
580 - Formatted printing is faster by compiling the formatting directives
581 to C code rather than interpreting at run time.
583 - Systemtap loads extra command line options from $SYSTEMTAP_DIR/rc
584 ($HOME/.systemtap/rc by default) before the normal argc/argv. This
585 may be useful to activate site options such as --use-server or
586 --download-debuginfo or --modinfo.
588 - The stap-server has seen many improvements, and is no longer considered
591 - The stap-server service (initscript) now supports four new options:
596 These allow the specification of macro definitions to be passed to stap
597 by the server, the location of the log file, network port number and
598 NSS certificate database location respectively. These options are also
599 supported within individual server configuration files. See stap-server
600 and initscript/README.stap-server for details. The stap-server is no
601 longer activated by default.
603 - process("PATH").[library("PATH")].function("NAME").exported probes are now
604 supported to filter function() to only exported instances.
606 - The translator supports a new --suppress-handler-errors option, which
607 causes most runtime errors to be turned into quiet skipped probes. This
608 also disables the MAXERRORS and MAXSKIPPED limits.
610 - Translator warnings have been standardized and controlled by the -w / -W
613 - The translator supports a new --modinfo NAME=VALUE option to emit additional
614 MODULE_INFO(n,v) macros into the generated code.
616 - There is no more fixed maximum number of VMA pages that will be tracked
617 at runtime. This reduces memory use for those scripts that don't need any,
618 or only limited target process VMA tracking and allows easier system
619 wide probes inspecting shared library variables and/or user backtraces.
620 stap will now silently ignore -DTASK_FINDER_VMA_ENTRY_ITEMS.
622 - The tapset functions remote_id() and remote_uri() identify the member of a
623 swarm of "stap --remote FOO --remote BAR baz.stp" concurrent executions.
625 - Systemtap now supports a new privilege level and group, "stapsys", which
626 is equivalent to the privilege afforded by membership in the group "stapdev",
627 except that guru mode (-g) functionality may not be used. To support this, a
628 new option, --privilege=[stapusr|stapsys|stapdev] has been added.
629 --privilege=stapusr is equivalent to specifying the existing --unprivileged
630 option. --privilege=stapdev is the default. See man stap(1) for details.
632 - Scripts that use kernel.trace("...") probes compile much faster.
634 - The systemtap module cache is cleaned less frequently, governed by the
635 number of seconds in the $SYSTEMTAP_DIR/cache/cache_clean_interval_s file.
637 - SDT can now define up to 12 arguments in a probe point.
639 - Parse errors no longer generate a cascade of false errors. Instead, a
640 parse error skips the rest of the current probe or function, and resumes
641 at the next one. This should generate fewer and better messages.
643 - Global array wrapping is now supported for both associative and statistics typed
644 arrays using the '%' character to signify a wrapped array. For example,
645 'global foo%[100]' would allow the array 'foo' to be wrapped if more than 100
646 elements are inserted.
648 - process("PATH").library("PATH").plt("NAME") probes are now supported.
649 Wildcards are supported in the plt-name part, to refer to any function in the
650 program linkage table which matches the glob pattern and the rest of the
653 - A new option, --dump-probe-types, will dump a list of supported probe types.
654 If --unprivileged is also specified, the list will be limited to probe types
655 which are available to unprivileged users.
657 - Systemtap can now automatically download the required debuginfo
658 using abrt. The --download-debuginfo[=OPTION] can be used to
659 control this feature. Possible values are: 'yes', 'no', 'ask',
660 and a positive number representing the timeout desired. The
661 default behavior is to not automatically download the debuginfo.
663 - The translator has better support for probing C++ applications by
664 better undertanding of compilation units, nested types, templates,
665 as used in probe point and @cast constructs.
667 - On 2.6.29+ kernels, systemtap can now probe kernel modules that
668 arrive and/or depart during the run-time of a session. This allows
669 probing of device driver initialization functions, which had formerly been
672 - New tapset functions for cpu_clock and local_clock access were added.
674 - There is some limited preliminary support for user-space probing
675 against kernels such as linux-next, which have no utrace but do have
676 the newer inode-uprobes work by Srikar Dronamraju and colleagues.
678 - The following probe types are deprecated in release 1.7 and will be
679 removed in release 1.8:
680 kernel.function(number).inline
681 module(string).function(number).inline
682 process.function(number).inline
683 process.library(string).function(number).inline
684 process(string).function(number).inline
685 process(string).library(string).function(number).inline
687 - The systemtap-grapher is deprecated in release 1.7 and will be removed in
690 - The task_backtrace() tapset function was deprecated in 1.6 and has been
693 - MAXBACKTRACE did work in earlier releases, but has now been documented
694 in the stap 1 manual page.
696 - New tapset function probe_type(). Returns a short string describing
697 the low level probe handler type for the current probe point.
699 - Both unwind and symbol data is now only collected and emitted for
700 scripts actually using backtracing or function/data symbols.
701 Tapset functions are marked with /* pragma:symbols */ or
702 /* pragma:unwind */ to indicate they need the specific data.
704 - Kernel backtraces can now be generated for non-pt_regs probe context
705 if the kernel support dump_trace(). This enables backtraces from
706 certain timer probes and tracepoints.
708 - ubacktrace() should now also work for some kernel probes on x86 which can
709 use the dwarf unwinder to recover the user registers to provide
710 more accurate user backtraces.
712 - For s390x the systemtap runtime now properly splits kernel and user
713 addresses (which are in separate address spaces on that architecture)
714 which enable user space introspection.
716 - ppc and s390x now supports user backtraces through the DWARF unwinder.
718 - ppc now handles function descriptors as symbol names correctly.
720 - arm support kernel backtraces through the DWARF unwinder.
722 - arm now have a uprobes port which enables user probes. This still
723 requires some kernel patches (user_regsets and tracehook support for
726 - Starting in release 1.7, these old variables will be deprecated:
727 - The 'pid' variable in the 'kprocess.release' probe has been
728 deprecated in favor of the new 'released_pid' variable.
729 - The 'args' variable in the
730 '_sunrpc.clnt.create_client.rpc_new_client_inline' probe has been
731 deprecated in favor of the new internal-only '__args' variable.
733 - Experimental support for recent kernels without utrace has been
734 added for the following probe types:
737 process("PATH").begin
739 process(PID).thread.begin
740 process("PATH").thread.begin
745 process(PID).thread.end
746 process("PATH").thread.end
749 process("PATH").syscall
751 process(PID).syscall.return
752 process("PATH").syscall.return
753 process.syscall.return
755 - staprun disables kprobe-optimizations in recent kernels, as problems
756 were found. (PR13193)
758 * What's new in version 1.6, 2011-07-25
760 - Security fixes for CVE-2011-2503: read instead of mmap to load modules,
761 CVE-2011-2502: Don't allow path-based auth for uprobes
763 - The systemtap compile-server no longer uses the -k option when calling the
764 translator (stap). As a result, the server will now take advantage of the
765 module cache when compiling the same script more than once. You may observe
766 an improvement in the performance of the server in this situation.
768 - The systemtap compile-server and client now each check the version of the
769 other, allowing both to adapt when communicating with a down-level
770 counterpart. As a result, all version of the client can communicate
771 with all versions of the server and vice-versa. Client will prefer newer
772 servers when selecting a server automatically.
774 - SystemTap has improved support for the ARM architecture. The
775 kread() and kwrite() operations for ARM were corrected allowing many
776 of the tapsets probes and function to work properly on the ARM
779 - Staprun can now rename the module to a unique name with the '-R' option before
780 inserting it. Systemtap itself will also call staprun with '-R' by default.
781 This allows the same module to be inserted more than once, without conflicting
784 - Systemtap error messages now provide feedback when staprun or any other
785 process fails to launch. This also specifically covers when the user
786 doesn't have the proper permissions to launch staprun.
788 - Systemtap will now map - to _ in module names. Previously,
789 stap -L 'module("i2c-core").function("*")' would be empty. It now returns
790 a list had stap -L 'module("i2c_core").function("*") been specified.
792 - Systemtap now fills in missing process names to probe points, to
793 avoid having to name them twice twice:
794 % stap -e 'probe process("a.out").function("*") {}' -c 'a.out ...'
795 Now the probed process name is inferred from the -c CMD argument.
796 % stap -e 'probe process.function("*") {}' -c 'a.out ...'
798 - stap -L 'process("PATH").syscall' will now list context variables
800 - Depends on elfutils 0.142+.
802 - Deprecated task_backtrace:string (task:long). This function will go
803 away after 1.6. Please run your scripts with stap --check-version.
805 * What's new in version 1.5, 2011-05-23
807 - Security fixes for CVE-2011-1781, CVE-2011-1769: correct DW_OP_{mod,div}
810 - The compile server and its related tools (stap-gen-ert, stap-authorize-cert,
811 stap-sign-module) have been re-implemented in C++. Previously, these
812 components were a mix of bash scripts and C code. These changes should be
813 transparent to the end user with the exception of NSS certificate database
814 password prompting (see below). The old implementation would prompt more
815 than once for the same password in some situations.
817 - eventcount.stp now allows for event counting in the format of
818 'stap eventcount.stp process.end syscall.* ...', and also reports
819 corresponding event tid's.
821 - Systemtap checks that the build-id of the module being probed matches the
822 build-id saved in the systemtap module. Invoking systemtap with
823 -DSTP_NO_BUILDID_CHECK will bypass this build-id runtime verification. See
824 man ld(1) for info on --build-id.
826 - stapio will now report if a child process has an abnormal exit along with
827 the associated status or signal.
829 - Compiler optimization may sometimes result in systemtap not being able to
830 access a user-space probe argument. Compiling the application with
831 -DSTAP_SDT_ARG_CONSTRAINT=nr will force the argument to be an immediate or
832 register value which should enable systemtap to access the argument.
834 - GNU Gettext has now been intergrated with systemtap. Our translation
835 page can be found at http://www.transifex.net/projects/p/systemtap/ .
836 "make update-po" will generate the necessary files to use translated
837 messages. Please refer to the po/README file for more info and
838 please consider contributing to this I18N effort!
840 - The new addr() function returns the probe's instruction pointer.
842 - process("...").library("...") probes are now supported. Wildcards
843 are supported in the library-name part, to refer to any shared
844 library that is required by process-name, which matches the glob
845 pattern and the rest of the probe point.
847 - The "--remote USER@HOST" functionality can now be specified multiple times
848 to fan out on multiple targets. If the targets have distinct kernel and
849 architecture configurations, stap will automatically build the script
850 appropriately for each one. This option is also no longer considered
853 - The NSS certificate database generated for use by the compile server is now
854 generated with no password. Previously, a random password was generated and
855 used to access the database. This change should be transparent to most users.
856 However, if you are prompted for a password when using systemtap, then
857 running $libexecdir/stap-gen-cert should correct the problem.
859 - The timestamp tapset includes jiffies() and HZ() for lightweight approximate
862 - A powerful new command line option --version has been added.
864 - process.mark now supports $$parms for reading probe parameters.
866 - A new command line option, --use-server-on-error[=yes|no] is available
867 for stap. It instructs stap to retry compilation of a script using a
868 compile server if it fails on the local host. The default setting
871 - The following deprecated tools have been removed:
873 stap-authorize-server-cert
874 stap-authorize-signing-cert
875 stap-find-or-start-server
877 Use the --use-server, --trust-server and --list-servers options of stap
880 * What's new in version 1.4, 2011-01-17
882 - Security fixes for CVE-2010-4170, CVE-2010-4171: staprun module
885 - A new /* myproc-unprivileged */ marker is now available for embedded C
886 code and and expressions. Like the /* unprivileged */ marker, it makes
887 the code or expression available for use in unprivileged mode (see
888 --unprivileged). However, it also automatically adds a call to
889 assert_is_myproc() to the code or expression, thus, making it available
890 to the unprivileged user only if the target of the current probe is within
891 the user's own process.
893 - The experimental "--remote USER@HOST" option will run pass 5 on a given
894 ssh host, after building locally (or with --use-server) for that target.
896 - Warning messages from the script may now be suppressed with the stap
897 and/or staprun -w option. By default, duplicate warning messages are
898 suppressed (up to a certain limit). With stap --vp 00002 and above,
899 the duplicate elimination is defeated.
901 - The print_ubacktrace and usym* functions attempt to print the full
902 path of the user-space binaries' paths, instead of just the basename.
903 The maximum saved path length is set by -DTASK_FINDER_VMA_ENTRY_PATHLEN,
904 default 64. Warning messages are produced if unwinding fails due to
905 a missing 'stap -d MODULE' option, providing preloaded unwind data.
907 - The new tz_ctime() tapset function prints times in the local time zone.
909 - More kernel tracepoints are accessible to the kernel.trace("...") mechanism,
910 if kernel source trees or debuginfo are available. These formerly "hidden"
911 tracepoints are those that are declared somewhere other than the usual
912 include/linux/trace/ headers, such as xfs and kvm.
914 - debuginfo-based process("...").function/.statement/.mark probes support
915 wildcards in the process-name part, to refer to any executable files that
916 match the glob pattern and the rest of the probe point.
918 - The -t option now displays information per probe-point rather than a summary
919 for each probe. It also now shows the derivation chain for each probe-point.
921 - A rewrite of the sys/sdt.h header file provides zero-cost startup (few or
922 no ELF relocations) for the debuginfo-less near-zero-cost runtime probes.
923 Binaries compiled with earlier sdt.h versions remain supported. The
924 stap -L (listing) option now lists parameters for sys/sdt.h markers.
926 - The implementation of the integrated compile-server client has been
928 o --use-server now accepts an argument representing a particular server and
929 may be specified more than once.
930 o --list-servers now accepts an expanded range of arguments.
931 o a new --trust-servers option has been added to stap to replace several
932 old certificate-management scripts.
933 o The following tools are now deprecated and will be removed in release 1.5:
935 stap-authorize-server-cert
936 stap-authorize-signing-cert
937 stap-find-or-start-server
939 See man stap(1) for complete details.
941 - The compile-server now returns the uprobes.ko to the client when it is
942 required by the script being compiled. The integrated compile-server client
943 now makes it available to be loaded by staprun. The old (deprecated)
944 stap-client does not do this.
946 - process probes with scripts as the target are recognized by stap and the
947 interpreter would be selected for probing.
949 - Starting in release 1.5, these old variables/functions will be deprecated
950 and will only be available when the '--compatible=1.4' flag is used:
952 - In the 'syscall.add_key' probe, the 'description_auddr' variable
953 has been deprecated in favor of the new 'description_uaddr'
955 - In the 'syscall.fgetxattr', 'syscall.fsetxattr',
956 'syscall.getxattr', 'syscall.lgetxattr', and
957 'syscall.lremovexattr' probes, the 'name2' variable has been
958 deprecated in favor of the new 'name_str' variable.
959 - In the 'nd_syscall.accept' probe the 'flag_str' variable
960 has been deprecated in favor of the new 'flags_str' variable.
961 - In the 'nd_syscall.dup' probe the 'old_fd' variable has been
962 deprecated in favor of the new 'oldfd' variable.
963 - In the 'nd_syscall.fgetxattr', 'nd_syscall.fremovexattr',
964 'nd_syscall.fsetxattr', 'nd_syscall.getxattr', and
965 'nd_syscall.lremovexattr' probes, the 'name2' variable has been
966 deprecated in favor of the new 'name_str' variable.
967 - The tapset alias 'nd_syscall.compat_pselect7a' was misnamed. It should
968 have been 'nd_syscall.compat_pselect7' (without the trailing 'a').
969 - The tapset function 'cpuid' is deprecated in favor of the better known
971 - In the i386 'syscall.sigaltstack' probe, the 'ussp' variable has
972 been deprecated in favor of the new 'uss_uaddr' variable.
973 - In the ia64 'syscall.sigaltstack' probe, the 'ss_uaddr' and
974 'oss_uaddr' variables have been deprecated in favor of the new
975 'uss_uaddr' and 'uoss_uaddr' variables.
976 - The powerpc tapset alias 'syscall.compat_sysctl' was deprecated
977 and renamed 'syscall.sysctl32'.
978 - In the x86_64 'syscall.sigaltstack' probe, the 'regs_uaddr'
979 variable has been deprecated in favor of the new 'regs' variable.
981 * What's new in version 1.3, 2010-07-21
983 - The uprobes kernel module now has about half the overhead when probing
984 NOPs, which is particularly relevant for sdt.h markers.
986 - New stap option -G VAR=VALUE allows overriding global variables
987 by passing the settings to staprun as module options.
989 - The tapset alias 'syscall.compat_pselect7a' was misnamed. It should
990 have been 'syscall.compat_pselect7' (without the trailing 'a').
991 Starting in release 1.4, the old name will be deprecated and
992 will only be available when the '--compatible=1.3' flag is used.
994 - A new procfs parameter .umask(UMASK) which provides modification of
995 file permissions using the proper umask value. Default file
996 permissions for a read probe are 0400, 0200 for a write probe, and
997 0600 for a file with a read and write probe.
999 - It is now possible in some situations to use print_ubacktrace() to
1000 get a user space stack trace from a kernel probe point. e.g. for
1001 user backtraces when there is a pagefault:
1002 $ stap -d /bin/sort --ldd -e 'probe vm.pagefault {
1003 if (pid() == target()) {
1004 printf("pagefault @0x%x\n", address); print_ubacktrace();
1007 pagefault @0x7fea0595fa70
1008 0x000000384f07f958 : __GI_strcmp+0x12b8/0x1440 [libc-2.12.so]
1009 0x000000384f02824e : __gconv_lookup_cache+0xee/0x5a0 [libc-2.12.so]
1010 0x000000384f021092 : __gconv_find_transform+0x92/0x2cf [libc-2.12.so]
1011 0x000000384f094896 : __wcsmbs_load_conv+0x106/0x2b0 [libc-2.12.so]
1012 0x000000384f08bd90 : mbrtowc+0x1b0/0x1c0 [libc-2.12.so]
1013 0x0000000000404199 : ismbblank+0x39/0x90 [sort]
1014 0x0000000000404a4f : inittables_mb+0xef/0x290 [sort]
1015 0x0000000000406934 : main+0x174/0x2510 [sort]
1016 0x000000384f01ec5d : __libc_start_main+0xfd/0x1d0 [libc-2.12.so]
1017 0x0000000000402509 : _start+0x29/0x2c [sort]
1020 - New tapset functions to get a string representation of a stack trace:
1021 sprint_[u]backtrace() and sprint_[u]stack().
1023 - New tapset function to get the module (shared library) name for a
1024 user space address umodname:string(long). The module name will now
1025 also be in the output of usymdata() and in backtrace addresses even
1026 when they were not given with -d at the command line.
1028 - Kernel backtraces are now much faster (replaced a linear search
1029 with a binary search).
1031 - A new integrated compile-server client is now available as part of stap.
1033 o 'stap --use-server ...' is equivalent to 'stap-client ...'
1034 o 'stap --list-servers' is equivalent to 'stap-find-servers'
1035 o 'stap --list-servers=online' is equivalent to 'stap-find-servers --all'
1036 o stap-client and its related tools will soon be deprecated.
1037 o the nss-devel and avahi-devel packages are required for building stap with
1038 the integrated client (checked during configuration).
1039 o nss and avahi are required to run the integrated client.
1041 - A new operator @entry is available for automatically saving an expression
1042 at entry time for use in a .return probe.
1043 probe foo.return { println(get_cycles() - @entry(get_cycles())) }
1045 - Probe $target variables and @cast() can now use a suffix to print complex
1046 data types as strings. Use a single '$' for a shallow view, or '$$' for a
1047 deeper view that includes nested types. For example, with fs_struct:
1048 $fs$ : "{.users=%i, .lock={...}, .umask=%i,
1049 .in_exec=%i, .root={...}, .pwd={...}}"
1050 $fs$$ : "{.users=%i, .lock={.raw_lock={.lock=%u}}, .umask=%i, .in_exec=%i,
1051 .root={.mnt=%p, .dentry=%p}, .pwd={.mnt=%p, .dentry=%p}}"
1053 - The <sys/sdt.h> user-space markers no longer default to an implicit
1054 MARKER_NAME_ENABLED() semaphore check for each marker. To check for
1055 enabled markers use a .d declaration file, then:
1056 if (MARKER_NAME_ENABLED()) MARKER_NAME()
1058 - Hyphenated <sys/sdt.h> marker names such as process(...).mark("foo-bar")
1059 are now accepted in scripts. They are mapped to the double-underscore
1062 - More robust <sys/sdt.h> user-space markers support is included. For
1063 some platforms (x86*, ppc*), this can let systemtap probe the markers
1064 without debuginfo. This implementation also supports preserving
1065 the "provider" name associated with a marker:
1066 probe process("foo").provider("bar").mark("baz") to match
1067 STAP_PROBE<n>(bar, baz <...>)
1068 (Compile with -DSTAP_SDT_V1 to revert to the previous implementation.
1069 Systemtap supports pre-existing or new binaries using them.)
1071 - Embedded-C may be used within expressions as values, when in guru mode:
1072 num = %{ LINUX_VERSION_CODE %} // int64_t
1073 name = %{ /* string */ THIS_MODULE->name %} // const char*
1074 printf ("%s %x\n", name, num)
1075 The usual /* pure */, /* unprivileged */, and /* guru */ markers may be used
1076 as with embedded-C functions.
1078 - By default the systemtap-runtime RPM builds now include a shared
1079 library, staplog.so, that allows crash to extract systemtap data from
1082 - Iterating with "foreach" can now explicitly save the value for the loop.
1083 foreach(v = [i,j] in array)
1084 printf("array[%d,%s] = %d\n", i, j, v /* array[i,j] */)
1086 - The new "--ldd" option automatically adds any additional shared
1087 libraries needed by probed or -d-listed userspace binaries to the -d
1088 list, to enable symbolic backtracing through them. Similarly, the
1089 new "--all-modules" option automatically adds any currently loaded
1090 kernel modules (listed in /proc/modules) to the -d list.
1092 - A new family of set_kernel_* functions make it easier for gurus to write
1093 new values at arbitrary memory addresses.
1095 - Probe wildcards can now use '**' to cross the '.' separator.
1096 $ stap -l 'sys**open'
1100 - Backward compatibility flags (--compatible=VERSION, and matching
1101 script preprocessing predicate %( systemtap_v CMP "version" %)
1102 and a deprecation policy are being introduced, in case future
1103 tapset/language changes break valid scripts.
1105 * What's new in version 1.2, 2010-03-22
1107 - Prototype support for "perf events", where the kernel supports the
1108 2.6.33 in-kernel API. Probe points may refer to low-level
1109 perf_event_attr type/config numbers, or to a number of aliases
1110 defined in the new perf.stp tapset:
1111 probe perf.sw.cpu_clock, perf.type(0).config(4) { }
1113 - Type-casting can now use multiple headers to resolve codependencies.
1114 @cast(task, "task_struct",
1115 "kernel<linux/sched.h><linux/fs_struct.h>")->fs->umask
1117 - Tapset-related man pages have been renamed. 'man -k 3stap' should show
1118 the installed list, which due to prefixing should no longer collide over
1119 ordinary system functions.
1121 - User space marker arguments no longer use volatile if the version of gcc,
1122 which must be at least 4.5.0, supports richer DWARF debuginfo. Use cflags
1123 -DSTAP_SDT_VOLATILE=volatile or -DSTAP_SDT_VOLATILE= when building
1124 the sys/sdt.h application to override this one way or another.
1126 - A new construct for error handling is available. It is similar to c++
1127 exception catching, using try and catch as new keywords. Within a handler
1128 or function, the following is valid and may be nested:
1129 try { /* arbitrary statements */ }
1130 catch (er) { /* e.g. println("caught error ", er) */ }
1132 - A new command line flag '-W' forces systemtap to abort translation of
1133 a script if any warnings are produced. It is similar to gcc's -Werror.
1134 (If '-w' is also supplied to suppress warnings, it wins.)
1136 - A new predicate @defined is available for testing whether a
1137 particular $variable/expression is resolvable at translate time:
1138 probe foo { if (@defined($bar)) log ("$bar is available here") }
1140 - Adjacent string literals are glued together, making this
1142 probe process("/usr" @1 "/bin").function("*") { ... }
1144 - In order to limit potential impact from future security problems,
1145 the stap-server process does not permit its being launched as root.
1147 - On recent kernels, for some architectures/configurations, hardware
1148 breakpoint probes are supported. The probe point syntax is:
1150 probe kernel.data(ADDRESS).write
1151 probe kernel.data(ADDRESS).length(LEN).write
1152 probe kernel.data("SYMBOL_NAME").write
1154 * What's new in version 1.1, 2010-01-15
1156 - New tracepoint based tapset for memory subsystem.
1158 - The loading of signed modules by staprun is no longer allowed for
1159 ordinary, unprivileged users. This means that only root, members of
1160 the group 'stapdev' and members of the group 'stapusr' can load
1161 systemtap modules using staprun, stap or stap-client. The minimum
1162 privilege required to run arbitrary --unprivileged scripts is now
1163 'stapusr' membership.
1165 - The stap-server initscript is available. This initscript allows you
1166 to start systemtap compile servers as a system service and to manage
1167 these servers as a group or individually. The stap-server initscript
1168 is installed by the systemtap-server rpm. The build directory for
1169 the uprobes module (/usr/share/systemtap/runtime/uprobes) is made
1170 writable by the 'stap-server' group. All of the files generated when
1171 building the uprobes module, including the digital signature, are
1172 also writable by members of stap-server.
1174 See initscript/README.stap-server for details.
1176 - Some of the compile server client, server and certificate management
1177 tools have been moved from $bindir to $libexecdir/systemtap.
1178 You should use the new stap-server script or the stap-server initscript
1179 for server management where possible. The stap-server script provides the same
1180 functionality as the stap-server initscript except that the servers are
1181 run by the invoking user by default as opposed to servers started by the
1182 stap-server initscript which are run by the user stap-server
1183 by default. See stap-server(8) for more information.
1185 You may continue to use these tools by adding $libexecdir/systemtap to
1186 your path. You would need to do this, for example, if you are not root,
1187 you want to start a compile server and you are not running systemtap from a
1188 private installation. In this case you still need to use stap-start-server.
1190 - Any diagnostic output line that starts with "ERROR", as in
1191 error("foo"), will promote a "Pass 5: run failed", and the return
1194 - Systemtap now warns about global variables being referenced from other
1195 script files. This aims to protect against unintended local-vs-global
1196 namespace collisions such as:
1198 % cat some_tapset.stp
1199 probe baz.one = bar { foo = $foo; bar = $bar }
1200 % cat end_user_script.stp
1201 global foo # intended to be private variable
1202 probe timer.s(1) { foo ++ }
1203 probe baz.* { println(foo, pp()) }
1204 % stap end_user_script.stp
1205 WARNING: cross-file global variable reference to foo from some_tapset.stp
1207 - Preprocessor conditional for kernel configuration testing:
1208 %( CONFIG_foo == "y" %? ... %)
1210 - ftrace(msg:string) tapset function to send strings to the system-wide
1211 ftrace ring-buffer (if any).
1213 - Better support for richer DWARF debuginfo output from GCC 4.5
1214 (variable tracking assignments). Kernel modules are now always resolved
1215 against all their dependencies to find any info referring to missing
1216 symbols. DW_AT_const_value is now supported when no DW_AT_location
1219 * What's new in verson 1.0, 2009-09-22
1221 - process().mark() probes now use an enabling semaphore to reduce the
1222 computation overhead of dormant probes.
1224 - The function spec for dwarf probes now supports C++ scopes, so you can
1225 limit the probes to specific namespaces or classes. Multiple scopes
1226 can be specified, and they will be matched progressively outward.
1227 probe process("foo").function("std::vector<*>::*") { }
1228 probe process("foo").function("::global_function") { }
1230 - It is now possible to cross-compile systemtap scripts for foreign
1231 architectures, using the new '-a ARCH' and '-B OPT=VALUE' flags.
1232 For example, put arm-linux-gcc etc. into your $PATH, and point
1233 systemtap at the target kernel build tree with:
1234 stap -a arm -B CROSS_COMPILE=arm-linux- -r /build/tree [...]
1235 The -B option is passed to kbuild make. -r identifies the already
1236 configured/built kernel tree and -a its architecture (kbuild ARCH=...).
1237 Systemtap will infer -p4.
1239 - Cross compilation using the systemtap client and server
1240 - stap-start-server now accepts the -r, -R, -I, -B and -a options in
1241 order to start a cross compiling server. The server will correctly
1242 advertise itself with respect to the kernel release and architecture
1243 that it compiles for.
1244 - When specified on stap-client, the -r and -a options will be
1245 considered when searching for a suitable server.
1247 - When using the systemtap client and server udp port 5353 must be open
1248 in your firewall in order for the client to find servers using
1249 avahi-browse. Also the systemtap server will choose a random port in
1250 the range 1024-63999 for accepting ssl connections.
1252 - Support for unprivileged users:
1253 ***********************************************************************
1254 * WARNING!!!!!!!!!! *
1255 * This feature is EXPERIMENTAL at this time and should be used with *
1256 * care. This feature allows systemtap kernel modules to be loaded by *
1257 * unprivileged users. The user interface and restrictions will change *
1258 * as this feature evolves. *
1259 ***********************************************************************
1260 - Systemtap modules generated from scripts which use a restricted
1261 subset of the features available may be loaded by staprun for
1262 unprivileged users. Previously, staprun would load modules only for
1263 root or for members of the groups stapdev and stapusr.
1264 - Using the --unprivileged option on stap enables translation-time
1265 checking for use by unprivileged users (see restrictions below).
1266 - All modules deemed suitable for use by unprivileged users will be
1267 signed by the systemtap server when --unprivileged is specified on
1268 stap-client. See module signing in release 0.9.8 and stap-server in
1270 - Modules signed by trusted signers (servers) and verified by staprun
1271 will be loaded by staprun regardless of the user's privilege level.
1272 - The system administrator asserts the trustworthiness of a signer
1273 (server) by running stap-authorize-signing-cert <cert-file> as root,
1274 where the <cert-file> can be found in
1275 ~<user>/.systemtap/ssl/server/stap.cert for servers started by
1276 ordinary users and in $sysconfdir/systemtap/ssl/server/stap.cert for
1277 servers started by root.
1278 - Restrictions are intentionally strict at this time and may be
1279 relaxed in the future:
1280 - probe points are restricted to:
1281 begin, begin(n), end, end(n), error, error(n), never,
1282 timer.{jiffies,s,sec,ms,msec,us,usec,ns,nsec}(n)*, timer.hz(n),
1283 process.* (for processes owned by the user).
1284 - use of embedded C code is not allowed.
1285 - use of tapset functions is restricted.
1286 - some tapset functions may not be used at all. A message will be
1287 generated at module compilation time.
1288 - some actions by allowed tapset functions may only be performed
1289 in the context of the user's own process. A runtime fault will
1290 occur in these situations, for example, direct memory access.
1291 - The is_myproc() tapset function has been provided so that
1292 tapset writers for unprivileged users can check that the
1293 context is of the users own process before attempting these
1295 - accessing the kernel memory space is not allowed.
1296 - The following command line options may not be used by stap-client
1298 - The following environment variables are ignored by stap-client:
1299 SYSTEMTAP_RUNTIME, SYSTEMTAP_TAPSET, SYSTEMTAP_DEBUGINFO_PATH
1300 - nss and nss-tools are required to use this feature.
1302 - Support output file switching by SIGUSR2. Users can command running
1303 stapio to switch output file by sending SIGUSR2.
1305 - Memory consumption for scripts involving many uprobes has been
1306 dramatically reduced.
1308 - The preprocessor now supports || and && in the conditions.
1309 e.g. %( arch == "x86_64" || arch == "ia64" %: ... %)
1311 - The systemtap notion of "architecture" now matches the kernel's, rather
1312 than that of "uname -m". This means that 32-bit i386 family are all
1313 known as "i386" rather than "i386" or "i686"; "ppc64" as "powerpc";
1314 "s390x" as "s390", and so on. This is consistent between the new
1315 "-a ARCH" flag and the script-level %( arch ... %) conditional.
1317 - It is now possible to define multiple probe aliases with the same name.
1318 A probe will expand to all matching aliases.
1321 probe foo { } # expands twice, once to bar and once to baz
1323 - A new experimental transport mechanism, using ftrace's ring_buffer,
1324 has been added. This may become the default transport mechanism in
1325 future versions of systemtap. To test this new transport mechanism,
1326 define 'STP_USE_RING_BUFFER'.
1328 - Support for recognizing DW_OP_{stack,implicit}_value DWARF expressions
1329 as emitted by GCC 4.5.
1331 * What's new in version 0.9.9, 2009-08-04
1333 - Systemwide kernel .function.return (kretprobe) maxactive defaults may
1334 be overridden with the -DKRETACTIVE=nnn parameter.
1336 - Translation pass 2 is significantly faster by avoiding unnecessary
1337 searching through a kernel build/module directory tree.
1339 - When compiled against elfutils 0.142 systemtap now handles the new
1340 DW_OP_call_frame_CFA generated by by GCC.
1342 - uprobes and ustack() are more robust when used on applications that
1343 depend on prelinked/separate debuginfo shared libraries.
1345 - User space PROBE marks are not always found with or without separate
1346 debuginfo. The .probes section itself is now always put in the main
1347 elf file and marked as allocated. When building pic code the section
1348 is marked writable. The selinux memory check problems seen with
1349 programs using STAP_PROBES is fixed.
1351 - statement() probes can now override "address not at start of statement"
1352 errors in guru mode. They also provide alternative addresses to use
1355 - The stapgraph application can generate graphs of data and events
1356 emitted by systemtap scripts in real time. Run "stapgraph
1357 testsuite/systemtap.examples/general/grapher.stp" for an example of
1358 graphing the system load average and keyboard events.
1360 - Dwarf probes now show parameters and local variables in the verbose
1363 - Symbol aliases are now resolved to their canonical dwarf names. For
1364 example, probing "malloc" in libc resolves to "__libc_malloc".
1366 - The syntax for dereferencing $target variables and @cast() gained new
1368 - Array indexes can now be arbitrary numeric expressions.
1369 - Array subscripts are now supported on pointer types.
1370 - An '&' operator before a @cast or $target returns the address of the
1371 final component, especially useful for nested structures.
1373 - For reading all probe variables, kernel.mark now supports $$vars and
1374 $$parms, and process.syscall now supports $$vars.
1376 - The SNMP tapset provides probes and functions for many network
1377 statistics. See stapprobes.snmp(3stap) for more details.
1379 - The dentry tapset provides functions to map kernel VFS directory entries
1380 to file or full path names: d_path(), d_name() and reverse_path_walk().
1382 - SystemTap now has userspace markers in its own binaries, and the stap
1383 tapset provides the available probepoints and local variables.
1385 - Miscellaneous new tapset functions:
1386 - pgrp() returns the process group ID of the current process
1387 - str_replace() performs string replacement
1389 * What's new in version 0.9.8, 2009-06-11
1391 - Miscellaneous new tapset functions:
1392 - sid() returns the session ID of the current process
1393 - stringat() indexes a single character from a string.
1395 - Using %M in print formats for hex dumps can now print entire buffers,
1396 instead of just small numbers.
1398 - Dwarfless syscalls: The nd_syscalls tapset is now available to probe
1399 system calls without requiring kernel debugging information. All of
1400 the same probepoints in the normal syscalls tapset are available with
1401 an "nd_" prefix, e.g. syscall.open becomes nd_syscall.open. Most
1402 syscall arguments are also available by name in nd_syscalls.
1404 - Module signing: If the appropriate nss libraries are available on your
1405 system, stap-server will sign each compiled module using a self-generated
1406 certificate. This is the first step toward extending authority to
1407 load certain modules to unprivileged users. For now, if the system
1408 administrator adds a certificate to a database of trusted signers
1409 (stap-authorize-signing-cert), modules signed using that certificate
1410 will be verified by staprun against tampering. Otherwise, you should
1411 notice no difference in the operation of stap or staprun.
1413 * What's new in version 0.9.7, 2009-04-23
1415 - @cast can now determine its type information using an explicit header
1416 specification. For example:
1417 @cast(tv, "timeval", "<sys/time.h>")->tv_sec
1418 @cast(task, "task_struct", "kernel<linux/sched.h>")->tgid
1420 - The overlapping process.* tapsets are now separated. Those probe points
1421 documented in stapprobes(3stap) remain the same. Those that were formerly
1422 in stapprobes.process(3stap) have been renamed to kprocess, to reflect
1423 their kernel perspective on processes.
1425 - The --skip-badvars option now also suppresses run-time error
1426 messages that would otherwise result from erroneous memory accesses.
1427 Such accesses can originate from $context expressions fueled by
1428 erroneous debug data, or by kernel_{long,string,...}() tapset calls.
1430 - New probes kprobe.function(FUNCTION) and kprobe.function(FUNCTION).return
1431 for dwarfless probing. These postpone function address resolution to
1432 run-time and use the kprobe symbol-resolution mechanism.
1433 Probing of absolute statements can be done using the
1434 kprobe.statement(ADDRESS).absolute construct.
1436 - EXPERIMENTAL support for user process unwinding. A new collection of
1437 tapset functions have been added to handle user space backtraces from
1438 probe points that support them (currently process and timer probes -
1439 for timer probes test whether or not in user space first with the
1440 already existing user_mode() function). The new tapset functions are:
1441 uaddr - User space address of current running task.
1442 usymname - Return the symbol of an address in the current task.
1443 usymdata - Return the symbol and module offset of an address.
1444 print_ustack - Print out stack for the current task from string.
1445 print_ubacktrace - Print stack back trace for current task.
1446 ubacktrace - Hex backtrace of current task stack.
1447 Please read http://sourceware.org/ml/systemtap/2009-q2/msg00364.html
1448 on the current restrictions and possible changes in the future and
1449 give feedback if you want to influence future developments.
1451 * What's new in version 0.9.5, 2009-03-27
1453 - New probes process().insn and process().insn.block that allows
1454 inspection of the process after each instruction or block of
1455 instructions executed. So to count the total number of instructions
1456 a process executes during a run do something like:
1457 $ stap -e 'global steps; probe process("/bin/ls").insn {steps++}
1458 probe end {printf("Total instructions: %d\n", steps);}' \
1460 This feature can slow down execution of a process somewhat.
1462 - Systemtap probes and function man pages extracted from the tapsets
1463 are now available under 3stap. To show the page for probe vm.pagefault
1464 or the stap function pexecname do:
1465 $ man 3stap vm.pagefault
1466 $ man 3stap pexecname
1468 - Kernel tracepoints are now supported for probing predefined kernel
1469 events without any debuginfo. Tracepoints incur less overhead than
1470 kprobes, and context parameters are available with full type
1471 information. Any kernel 2.6.28 and later should have defined
1472 tracepoints. Try the following to see what's available:
1473 $ stap -L 'kernel.trace("*")'
1475 - Typecasting with @cast now supports modules search paths, which is
1476 useful in case there are multiple places where the type definition
1477 may be found. For example:
1478 @cast(sdev, "scsi_device", "kernel:scsi_mod")->sdev_state
1480 - On-file flight recorder is supported. It allows stap to record huge
1481 trace log on the disk and to run in background.
1482 Passing -F option with -o option runs stap in background mode. In this
1483 mode, staprun is detached from console, and stap itself shows staprun's
1485 Specifying the max size and the max number of log files are also available
1486 by passing -S option. This option has one or two arguments seperated by
1487 a comma. The first argument is the max size of a log file in MB. If the
1488 size of a log file exceeds it, stap switches to the next log file
1489 automatically. The second is how many files are kept on the disk. If the
1490 number of log files exceeds it, the oldest log file is removed
1491 automatically. The second argument can be omitted.
1493 For example, this will record output on log files each of them is smaller
1494 than 1024MB and keep last 3 logs, in background.
1495 % stap -F -o /tmp/staplog -S 1024,3 script.stp
1497 - In guru mode (-g), the kernel probing blacklist is disabled, leaving
1498 only a subset - the kernel's own internal kprobe blacklist - to attempt
1499 to filter out areas unsafe to probe. The differences may be enough to
1500 probe more interrupt handlers.
1502 - Variables unavailable in current context may be skipped by setting a
1503 session level flag with command line option --skip-badvars now available.
1504 This replaces any dwarf $variable expressions that could not be resolved
1505 with literal numeric zeros, along with a warning message.
1507 - Both kernel markers and kernel tracepoint support argument listing
1508 through stap -L 'kernel.mark("*")' or stap -L 'kernel.trace("*")'
1510 - Users can use -DINTERRUPTIBLE=0 to prevent interrupt reentrancy in
1511 their script, at the cost of a bit more overhead to toggle the
1514 - Added reentrancy debugging. If stap is run with the arguments
1515 "-t -DDEBUG_REENTRANCY", additional warnings will be printed for
1516 every reentrancy event, including the probe points of the
1517 resident and interloper probes.
1519 - Default to --disable-pie for configure.
1520 Use --enable-pie to turn it back on.
1522 - Improved sdt.h compatibility and test suite for static dtrace
1523 compatible user space markers.
1525 - Some architectures now use syscall wrappers (HAVE_SYSCALL_WRAPPERS).
1526 The syscall tapset has been enhanced to take care of the syscall
1527 wrappers in this release.
1529 - Security fix for CVE-2009-0784: stapusr module-path checking race.
1531 * What's new in version 0.9, 2009-02-19
1533 - Typecasting is now supported using the @cast operator. A script can
1534 define a pointer type for a "long" value, and then access type members
1535 using the same syntax as with $target variables. For example, this will
1536 retrieve the parent pid from a kernel task_struct:
1537 @cast(pointer, "task_struct", "kernel")->parent->pid
1539 - process().mark() probes are now possible to trace static user space
1540 markers put in programs with the STAP_PROBE macro using the new
1541 sys/sdt.h include file. This also provides dtrace compatible markers
1542 through DTRACE_PROBE and an associated python 'dtrace' script that
1543 can be used in builds based on dtrace that need dtrace -h or -G
1546 - For those that really want to run stap from the build tree there is
1547 now the 'run-stap' script in the top-level build directory that sets
1548 up the SYSTEMTAP_TAPSET, SYSTEMTAP_RUNTIME, SYSTEMTAP_STAPRUN, and
1549 SYSTEMTAP_STAPIO environment variables (installing systemtap, in a
1550 local prefix, is still recommended for common use).
1552 - Systemtap now comes with a new Beginners Guide that walks the user
1553 through their first steps setting up stap, understanding how it all
1554 works, introduces some useful scripts and describes some common
1555 pitfalls. It isn't created by default since it needs a Publican
1556 setup, but full build instructions can be found in the wiki:
1557 http://sourceware.org/systemtap/wiki/PublicanQuikHowto
1558 An online version can be found at:
1559 http://sourceware.org/systemtap/SystemTap_Beginners_Guide.pdf
1561 - Standard tapsets included with Systemtap were modified to include
1562 extractable documentation information based on the kernel-doc
1563 infrastructure. When configured --enabled-docs a HTML and PDF
1564 version of the Tapset Reference Manual is produced explaining probes
1565 defined in each tapset.
1567 - The systemtap client and compile server are now available.
1568 These allow you to compile a systemtap module on a host other than
1569 the one which it will be run, providing the client and server
1570 are compatible. Other than using a server for passes 1 through
1571 4, the client behaves like the 'stap' front end itself. This
1572 means, among other things, that the client will automatically
1573 load the resulting module on the local host unless -p[1234]
1574 was specified. See stap-server(8) for more details.
1575 The client/server now use SSL for network connection security and
1578 The systemtap client and server are prototypes only. Interfaces, options
1579 and usage may change at any time.
1581 - function("func").label("label") probes are now supported to allow matching
1582 the label of a function.
1584 - Systemtap initscript is available. This initscript allows you to run
1585 systemtap scripts as system services (in flight recorder mode) and
1586 control those scripts individually.
1587 See README.systemtap for details.
1589 - The stap "-r DIR" option may be used to identify a hand-made kernel
1590 build directory. The tool determines the appropriate release string
1591 automatically from the directory.
1593 - Serious problems associated with user-space probing in shared libraries
1594 were corrected, making it now possible to experiment with probe shared
1595 libraries. Assuming dwarf debugging information is installed, use this
1596 twist on the normal syntax:
1598 probe process("/lib64/libc-2.8.so").function("....") { ... }
1600 This would probe all threads that call into that library. Running
1601 "stap -c CMD" or "stap -x PID" naturally restricts this to the target
1602 command+descendants only. $$vars etc. may be used.
1604 - For scripts that sometimes terminate with excessive "skipped" probes,
1605 rerunning the script with "-t" (timing) will print more details about
1606 the skippage reasons.
1608 - Symbol tables and unwind (backtracing) data support were formerly
1609 compiled in for all probed modules as identified by the script
1610 (kernel; module("name"); process("file")) plus those listed by the
1611 stap "-d BINARY" option. Now, this data is included only if the systemtap
1612 script uses tapset functions like probefunc() or backtrace() that require
1613 such information. This shrinks the probe modules considerably for the rest.
1615 - Per-pass verbosity control is available with the new "--vp {N}+" option.
1616 "stap --vp 040" adds 4 units of -v verbosity only to pass 2. This is useful
1617 for diagnosing errors from one pass without excessive verbosity from others.
1619 - Most probe handlers now run with interrupts enabled, for improved
1620 system responsiveness and less probing overhead. This may result
1621 in more skipped probes, for example if a reentrant probe handler
1622 is attempted from within an interrupt handler. It may also make the
1623 systemtap overload detection facility more likely to be triggered, as
1624 interrupt handlers' run time would be included in the self-assessed
1625 overhead of running probe handlers.
1627 * What's new in version 0.8, 2008-11-13
1629 - Cache limiting is now available. If the compiled module cache size is
1630 over a limit specified in the $SYSTEMTAP_DIR/cache/cache_mb_limit file,
1631 some old cache entries will be unlinked. See man stap(1) for more.
1633 - Error and warning messages are now followed by source context displaying
1634 the erroneous line/s and a handy '^' in the following line pointing to the
1637 - A bug reporting tool "stap-report" is now available which will quickly
1638 retrieve much of the information requested here:
1639 http://sourceware.org/systemtap/wiki/HowToReportBugs
1641 - The translator can resolve members of anonymous structs / unions:
1642 given struct { int foo; struct { int bar; }; } *p;
1643 this now works: $p->bar
1645 - The stap "-F" flag activates "flight recorder" mode, which consists of
1646 translating the given script as usual, but implicitly launching it into
1647 the background with staprun's existing "-L" (launch) option. A user
1648 can later reattach to the module with "staprun -A MODULENAME".
1650 - Additional context variables are available on user-space syscall probes.
1651 - $argN ($arg1, $arg2, ... $arg6) in process(PATH_OR_PID).syscall
1652 gives you the argument of the system call.
1653 - $return in process(PATH_OR_PID).syscall.return gives you the return
1654 value of the system call.
1656 - Target process mode (stap -c CMD or -x PID) now implicitly restricts all
1657 "process.*" probes to the given child process. (It does not affect
1658 kernel.* or other probe types.) The CMD string is normally run directly,
1659 rather than via a /bin/sh -c subshell, since then utrace/uprobe probes
1660 receive a fairly "clean" event stream. If metacharacters like
1661 redirection operators were present in CMD, then "sh -c CMD" is still
1662 used, and utrace/uprobe probes will receive events from the shell.
1664 % stap -e 'probe process.syscall, process.end {
1665 printf("%s %d %s\n", execname(), pid(), pp())}'\
1667 ls 2323 process.syscall
1668 ls 2323 process.syscall
1671 - Probe listing mode is improved: "-L" lists available script-level variables
1673 % stap -L 'syscall.*open*'
1674 syscall.mq_open name:string name_uaddr:long filename:string mode:long u_attr_uaddr:long oflag:long argstr:string
1675 syscall.open name:string filename:string flags:long mode:long argstr:string
1676 syscall.openat name:string filename:string flags:long mode:long argstr:string
1678 - All user-space-related probes support $PATH-resolved executable
1681 probe process("ls").syscall {}
1682 probe process("./a.out").syscall {}
1684 work now, instead of just
1686 probe process("/bin/ls").syscall {}
1687 probe process("/my/directory/a.out").syscall {}
1689 - Prototype symbolic user-space probing support:
1691 # stap -e 'probe process("ls").function("*").call {
1692 log (probefunc()." ".$$parms)
1697 - debugging information for the named program
1698 - a version of utrace in the kernel that is compatible with the "uprobes"
1699 kernel module prototype. This includes RHEL5 and older Fedora, but not
1700 yet current lkml-track utrace; a "pass 4a"-time build failure means
1701 your system cannot use this yet.
1703 - Global variables which are written to but never read are now
1704 automatically displayed when the session does a shutdown. For example:
1706 global running_tasks
1707 probe timer.profile {running_tasks[pid(),tid()] = execname()}
1708 probe timer.ms(8000) {exit()}
1710 - A formatted string representation of the variables, parameters, or local
1711 variables at a probe point is now supported via the special $$vars,
1712 $$parms, and $$locals context variables, which expand to a string
1713 containing a list "var1=0xdead var2=0xbeef var3=?". (Here, var3 exists
1714 but is for some reason unavailable.) In return probes only, $$return
1715 expands to an empty string for a void function, or "return=0xf00".
1718 * What's new in version 0.7, 2008-07-15
1720 - .statement("func@file:*") and .statement("func@file:M-N") probes are now
1721 supported to allow matching a range of lines in a function. This allows
1722 tracing the execution of a function.
1724 - Scripts relying on probe point wildcards like "syscall.*" that expand
1725 to distinct kprobes are processed significantly faster than before.
1727 - The vector of script command line arguments is available in a
1728 tapset-provided global array argv[]. It is indexed 1 ... argc,
1729 another global. This can substitute for of preprocessor
1730 directives @NNN that fail at parse time if there are not
1733 printf("argv: %s %s %s", argv[1], argv[2], argv[3])
1735 - .statement("func@file+line") probes are now supported to allow a
1736 match relative to the entry of the function incremented by line
1737 number. This allows using the same systemtap script if the rest
1738 of the file.c source only changes slightly.
1740 - A probe listing mode is available.
1749 - More user-space probe types are added:
1751 probe process(PID).begin { }
1752 probe process("PATH").begin { }
1753 probe process(PID).thread.begin { }
1754 probe process("PATH").thread.begin { }
1755 probe process(PID).end { }
1756 probe process("PATH").end { }
1757 probe process(PID).thread.end { }
1758 probe process("PATH").thread.end { }
1759 probe process(PID).syscall { }
1760 probe process("PATH").syscall { }
1761 probe process(PID).syscall.return { }
1762 probe process("PATH").syscall.return { }
1764 - Globals now accept ; terminators
1767 global little[10], big[5];
1769 * What's new in version 0.6, 2007-12-15
1771 - A copy of the systemtap tutorial and language reference guide
1774 - There is a new format specifier, %m, for the printf family of
1775 functions. It functions like %s, except that it does not stop when
1776 a nul ('\0') byte is encountered. The number of bytes output is
1777 determined by the precision specifier. The default precision is 1.
1780 printf ("%m", "My String") // prints one character: M
1781 printf ("%.5", myString) // prints 5 bytes beginning at the start
1784 - The %b format specifier for the printf family of functions has been enhanced
1787 1) When the width and precision are both unspecified, the default is %8.8b.
1788 2) When only one of the width or precision is specified, the other defaults
1789 to the same value. For example, %4b == %.4b == %4.4b
1790 3) Nul ('\0') bytes are used for field width padding. For example,
1792 printf ("%b", 0x1111deadbeef2222) // prints all eight bytes
1793 printf ("%4.2b", 0xdeadbeef) // prints \0\0\xbe\xef
1795 - Dynamic width and precision are now supported for all printf family format
1796 specifiers. For example:
1800 printf ("%*.*b", four, two, 0xdeadbbeef) // prints \0\0\xbe\xef
1801 printf ("%*d", four, two) // prints <space><space><space>2
1803 - Preprocessor conditional expressions can now include wildcard style
1804 matches on kernel versions.
1805 %( kernel_vr != "*xen" %? foo %: bar %)
1807 - Prototype support for user-space probing is showing some progress.
1808 No symbolic notations are supported yet (so no probing by function names,
1809 file names, process names, and no access to $context variables), but at
1810 least it's something:
1812 probe process(PID).statement(ADDRESS).absolute { }
1814 This will set a uprobe on the given process-id and given virtual address.
1815 The proble handler runs in kernel-space as usual, and can generally use
1816 existing tapset functions.
1818 - Crash utility can retrieve systemtap's relay buffer from a kernel dump
1819 image by using staplog which is a crash extension module. To use this
1820 feature, type commands as below from crash(8)'s command line:
1822 crash> extend staplog.so
1823 crash> help systemtaplog
1825 Then, you can see more precise help message.
1827 - You can share a relay buffer amoung several scripts and merge outputs from
1828 several scripts by using "-DRELAY_HOST" and "-DRELAY_GUEST" options.
1832 % stap -ve 'probe begin{}' -o merged.out -DRELAY_HOST &
1833 # wait until starting the host.
1834 % stap -ve 'probe begin{print("hello ");exit()}' -DRELAY_GUEST
1835 % stap -ve 'probe begin{print("world\n");exit()}' -DRELAY_GUEST
1837 Then, you'll see "hello world" in merged.out.
1839 - You can add a conditional statement for each probe point or aliase, which
1840 is evaluated when the probe point is hit. If the condition is false, the
1841 whole probe body(including aliases) is skipped. For example:
1844 probe syscall.* if (switch) { ... }
1845 probe procfs.write {switch = strtol($value,10)} /* enable/disable ctrl */
1847 - Systemtap will warn you if your script contains unused variables or
1848 functions. This is helpful in case of misspelled variables. If it
1849 doth protest too much, turn it off with "stap -w ...".
1851 - You can add error-handling probes to a script, which are run if a
1852 script was stopped due to errors. In such a case, "end" probes are
1853 not run, but "error" ones are.
1855 probe error { println ("oops, errors encountered; here's a report anyway")
1856 foreach (coin in mint) { println (coin) } }
1858 - In a related twist, one may list probe points in order of preference,
1859 and mark any of them as "sufficient" beyond just "optional". Probe
1860 point sequence expansion stops if a sufficient-marked probe point has a hit.
1861 This is useful for probes on functions that may be in a module (CONFIG_FOO=m)
1862 or may have been compiled into the kernel (CONFIG_FOO=y), but we don't know
1865 probe module("sd").function("sd_init_command") ? ,
1866 kernel.function("sd_init_command") ? { ... }
1868 which might match neither, now one can write this:
1870 probe module("sd").function("sd_init_command") ! , /* <-- note excl. mark */
1871 kernel.function("sd_init_command") { ... }
1873 - New security model. To install a systemtap kernel module, a user
1874 must be one of the following: the root user; a member of the
1875 'stapdev' group; or a member of the 'stapusr' group. Members of the
1876 stapusr group can only use modules located in the
1877 /lib/modules/VERSION/systemtap directory (where VERSION is the
1878 output of "uname -r").
1880 - .statement("...@file:line") probes now apply heuristics to allow an
1881 approximate match for the line number. This works similarly to gdb,
1882 where a breakpoint placed on an empty source line is automatically
1883 moved to the next statement. A silly bug that made many $target
1884 variables inaccessible to .statement() probes was also fixed.
1886 - LKET has been retired. Please let us know on <systemtap@sourceware.org>
1887 if you have been a user of the tapset/tools, so we can help you find
1890 - New families of printing functions println() and printd() have been added.
1891 println() is like print() but adds a newline at the end;
1892 printd() is like a sequence of print()s, with a specified field delimiter.
1894 * What's new since version 0.5.14?, 2007-07-03
1896 - The way in which command line arguments for scripts are substituted has
1897 changed. Previously, $1 etc. would interpret the corresponding command
1898 line argument as an numeric literal, and @1 as a string literal. Now,
1899 the command line arguments are pasted uninterpreted wherever $1 etc.
1900 appears at the beginning of a token. @1 is similar, but is quoted as
1901 a string. This change does not modify old scripts, but has the effect
1902 of permitting substitution of arbitrary token sequences.
1904 # This worked before, and still does:
1905 % stap -e 'probe timer.s($1) {}' 5
1906 # Now this also works:
1907 % stap -e 'probe syscall.$1 {log(@1)}' open
1908 # This won't crash, just signal a recursion error:
1910 # As before, $1... is recognized only at the beginning of a token
1911 % stap -e 'probe begin {foo$1=5}'
1913 * What's new since version 0.5.13?, 2007-03-26
1915 - The way in which systemtap resolves function/inline probes has changed:
1916 .function(...) - now refers to all functions, inlined or not
1917 .inline(...) - is deprecated, use instead:
1918 .function(...).inline - filters function() to only inlined instances
1919 .function(...).call - filters function() to only non-inlined instances
1920 .function(...).return - as before, but now pairs best with .function().call
1921 .statement() is unchanged.
1923 * What's new since version 0.5.12?, 2007-01-01
1925 - When running in -p4 (compile-only) mode, the compiled .ko file name
1926 is printed on standard output.
1928 - An array element with a null value such as zero or an empty string
1929 is now preserved, and will show up in a "foreach" loop or "in" test.
1930 To delete such an element, the scripts needs to use an explicit
1931 "delete array[idx]" statement rather than something like "array[idx]=0".
1933 - The new "-P" option controls whether prologue searching heuristics
1934 will be activated for function probes. This was needed to get correct
1935 debugging information (dwarf location list) data for $target variables.
1936 Modern compilers (gcc 4.1+) tend not to need this heuristic, so it is
1937 no longer default. A new configure flag (--enable-prologues) restores
1938 it as a default setting, and is appropriate for older compilers (gcc 3.*).
1940 - Each systemtap module prints a one-line message to the kernel informational
1941 log when it starts. This line identifies the translator version, base
1942 address of the probe module, a broken-down memory consumption estimate, and
1943 the total number of probes. This is meant as a debugging / auditing aid.
1945 - Begin/end probes are run with interrupts enabled (but with
1946 preemption disabled). This will allow begin/end probes to be
1947 longer, to support generating longer reports.
1949 - The numeric forms of kernel.statement() and kernel.function() probe points
1950 are now interpreted as relocatable values - treated as relative to the
1951 _stext symbol in that kernel binary. Since some modern kernel images
1952 are relocated to a different virtual address at startup, such addresses
1953 may shift up or down when actually inserted into a running kernel.
1955 kernel.statement(0xdeadbeef): validated, interpreted relative to _stext,
1956 may map to 0xceadbeef at run time.
1958 In order to specify unrelocated addresses, use the new ".absolute"
1959 probe point suffix for such numeric addresses. These are only
1960 allowed in guru mode, and provide access to no $target variables.
1961 They don't use debugging information at all, actually.
1963 kernel.statement(0xfeedface).absolute: raw, unvalidated, guru mode only
1965 * What's new since version 0.5.10?, 2006-10-19
1967 - Offline processing of debugging information, enabling general
1968 cross-compilation of probe scripts to remote hosts, without
1969 requiring identical module/memory layout. This slows down
1970 compilation/translation somewhat.
1972 - Kernel symbol table data is loaded by staprun at startup time
1973 rather than compiled into the module.
1975 - Support the "limit" keyword for foreach iterations:
1976 foreach ([x,y] in ary limit 5) { ... }
1977 This implicitly exits after the fifth iteration. It also enables
1978 more efficient key/value sorting.
1980 - Support the "maxactive" keyword for return probes:
1981 probe kernel.function("sdfsdf").maxactive(848) { ... }
1982 This allows up to 848 concurrently outstanding entries to
1983 the sdfsdf function before one returns. The default maxactive
1984 number is smaller, and can result in missed return probes.
1986 - Support accessing of saved function arguments from within
1987 return probes. These values are saved by a synthesized
1988 function-entry probe.
1990 - Add substantial version/architecture checking in compiled probes to
1991 assert correct installation of debugging information and correct
1992 execution on a compatible kernel.
1994 - Add probe-time checking for sufficient free stack space when probe
1995 handlers are invoked, as a safety improvement.
1997 - Add an optional numeric parameter for begin/end probe specifications,
1998 to order their execution.
1999 probe begin(10) { } /* comes after */ probe begin(-10) {}
2001 - Add an optional array size declaration, which is handy for very small
2003 global little[5], big[20000]
2005 - Include some example scripts along with the documentation.
2007 - Change the start-time allocation of probe memory to avoid causing OOM
2008 situations, and to abort cleanly if free kernel memory is short.
2010 - Automatically use the kernel DWARF unwinder, if present, for stack
2013 - Many minor bug fixes, performance, tapset, and error message