1 * What's new in version 2.5
3 - Statement probes (e.g. process.statement) are now faster to resolve,
4 more precise, and work better with inlined functions.
6 - New switches have been added to help inspect the contents of installed
9 stap --dump-function --> lists all library functions and their args
10 stap --dump-probe-aliases --> list all library probe aliases
12 - The heuristic algorithms used to search for function-prologue
13 endings were improved, to cover more optimization (or
14 lack-of-optimization, or incorrect-debuginfo) cases. These
15 heuristics are necessary to find $context parameters for some
16 function-call/entry probes. We recommend programs be built with
17 CFLAGS+=-grecord-gcc-switches to feed information to the heuristics.
19 - The stap --use-server option now more correctly supports address:port
20 type parametrization, for manual use in the absence of avahi.
22 - A new probe alias "oneshot" allows a single quick script fragment to run,
25 - The argv tapset now merges translate-time and run-time positional
26 arguments, so all of these work:
28 stap -e 'probe oneshot {println(argv[1]," ",argv[2])}' hello world
30 stap -e 'probe oneshot {println(argv[1]," ",argv[2])}' \
31 -G argv_1=hello -G argv_2=world
33 staprun hello.ko argv_1=hello argv_2=world
35 - SystemTap now falls back on the symbol table for probing
36 functions in processes if the debuginfo is not available.
38 - SystemTap now supports a %( guru_mode == 0 /* or 1 */ %)
39 conditional for making dual-use scripts.
41 - SystemTap now supports UEFI/SecureBoot systems, via
42 machine-owner-keys maintained by a trusted stap-server on the
43 network. (Key enrollment requires a one-time reboot and BIOS
46 - SystemTap now reports more accurate and succinct errors on type
49 - Embedded-C functions may use STAP_RETURN(value) instead of the
50 more wordy STAP_RETVALUE assignment followed by a "goto out".
51 The macro supports numeric or string values as appropriate.
53 - Some struct-sockaddr fields are now individually decoded for
54 socket-related syscalls:
55 probe syscall.connect { println (uaddr_af, ":", uaddr_ip) }
57 - The documentation for the SystemTap initscript service and the
58 SystemTap compile-server service have been completely converted from
59 README files to man pages (see systemtap(8) and stap-server(8)).
61 - SystemTap is now capable of inserting modules early during the boot
62 process on dracut-based systems. See the 'onboot' command in
63 systemtap(8) for more information.
65 - DWARF probes can now use the '.callee[s]' variants, which allow more
66 precise function probing. For example, the probe point
68 process("myproc").function("foo").callee("bar")
70 will fire upon entering bar() from foo(). A '.callees' probe will
71 instead place probes on all callees of foo().
72 Note that this also means that probe point wildcards should be used
73 with more care. For example, use signal.*.return rather than
74 signal.*.*, which would also match '.callees'. See stapprobes(3stap)
75 for more info. This feature requires at least GCC 4.7.
77 - A few new functions in the task_time tapsets, as well as a new tapset
78 function task_ancestry(), which prints out the parentage of a process.
80 - The kprocess.exec probe has been updated to use syscall.execve, which
81 allows access to the new process' arguments (through the new 'argstr'
82 or 'args' variables) as well as giving better support across kernel
83 versions. Note also that the 'filename' variable now holds the
84 filename (quoted), or the address (unquoted) if it couldn't be
87 - The [s]println() function can now be called without any arguments to
88 simply print a newline.
90 - Suggestions are now provided when markers could not be resolved. For
91 example, process("stap").mark("benchmart") will suggest 'benchmark'.
93 - SystemTap colors can now be turned off by simply setting
94 SYSTEMTAP_COLORS to be empty, rather than having to make it invalid.
96 - There is a new context tapset function, pnlabel(), which returns the
97 name of the label which fired.
99 - The following tapset variables and functions are deprecated in
101 - The 'clone_flags', 'stack_start', 'stack_size',
102 'parent_tid_uaddr', and 'child_tid_uaddr' variables in the
103 'syscall.fork' and 'nd_syscall.fork' probe aliases.
104 - The '_sendflags_str()' and '_recvflags_str()' functions have been
105 deprecated in favor of the new '_msg_flags_str()' function.
106 - The 'flags' and 'flags_str' variables in the 'syscall.accept' and
107 'nd_syscall.accept' probe alias.
109 * What's new in version 2.4, 2013-11-06
111 - Better suggestions are given in many of the semantic errors in which
112 alternatives are provided. Additionally, suggestions are now provided
113 when plt and trace probes could not be resolved. For example,
114 kernel.trace("sched_siwtch") will suggest 'sched_switch'.
116 - SystemTap is now smarter about error reporting. Errors from the same
117 source are considered duplicates and suppressed. A message is
118 displayed on exit if any errors/warnings were suppressed.
120 - Statistics aggregate typed objects are now implemented locklessly,
121 if the translator finds that they are only ever read (using the
122 foreach / @count / etc. constructs) in a probe-begin/end/error.
124 - SystemTap now supports probing inside virtual machines using the
125 libvirt and unix schemes, e.g.
127 stap -ve 'probe timer.s(1) { printf("hello!\n") }' \
128 --remote=libvirt://MyVirtualMachine
130 Virtual machines managed by libvirt can be prepared using stapvirt.
131 See stapvirt(1) and the --remote option in stap(1) for more details.
133 - Systemtap now checks for and uses (when available) the .gnu_debugdata
134 section which contains a subset of debuginfo, useful for backtraces
137 - SystemTap map variables are now allocated with vmalloc() instead of
138 with kmalloc(), which should cause memory to be less fragmented.
140 - Although SystemTap itself requires elfutils 0.148+, staprun only
141 requires elfutils 0.142+, which could be useful with the
142 '--disable-translator' configure switch.
144 - Under FIPS mode (/proc/sys/crypto/fips_enabled=1), staprun will
145 refuse to load systemtap modules (since these are not normally
146 signed with the kernel's build-time keys). This protection may
147 be suppressed with the $STAP_FIPS_OVERRIDE environment variable.
149 - The stap-server client & server code now enable all SSL/TLS
150 ciphers rather than just the "export" subset.
152 - For systems with in-kernel utrace, 'process.end' and 'thread.end'
153 probes will hit before the target's parent process is notified of
154 the target's death. This matches the behavior of newer kernels
155 without in-kernel utrace.
157 * What's new in version 2.3, 2013-07-25
159 - More context-accessing functions throw systemtap exceptions upon a
160 failure, whereas in previous versions they might return non-error
161 sentinel values like "" or "<unknown>". Use try { } / catch { }
162 around these, or new wrapper functions such as user_string_{n_}quoted()
163 that internally absorb exceptions.
165 - java("org.my.MyApp") probes are now restricted to pre-existing jvm pid's with
166 a listing in jps -l output to avoid recursive calls
168 - The tapset [nd_]syscall.semop parameter tsops_uaddr is renamed sops_uaddr for
169 consistency with [nd_]syscall.semtimedop.
171 - The udp.stp tapset adds some ip-address/port variables.
173 - A new guru-mode-only tapset function raise() is available to send signals
176 - Support for the standard Posix ERE named character classes has been
177 added to the regexp engine, e.g. [:digit:], [:alpha:], ...
179 - A substantial internal overhaul of the regexp engine has resulted in
180 correct behaviour on further obscure edge cases. The regexp engine
181 now implements the ERE standard and correctly passes the testsuite
182 for the glibc regexp engine (minus portions corresponding to
183 unimplemented features -- i.e. subexpression capture and reuse).
185 - Alternative functions are now suggested when function probes could not be
186 resolved. For example, kernel.function("vfs_reads") will suggest vfs_read.
187 Other probes for which suggestions are made are module.function,
188 process.function, and process.library.function.
190 - Has life been a bit bland lately? Want to spice things up? Why not write a
191 few faulty probes and feast your eyes upon the myriad of colours adorning
192 your terminal as SystemTap softly whispers in your ear... 'parse error'.
193 Search for '--color' in 'man stap' for more info.
195 - The following tapset functions are deprecated in release 2.3:
196 'stap_NFS_CLIENT', '__getfh_inode', '_success_check',
197 '_sock_prot_num', '_sock_fam_num', '_sock_state_num',
198 '_sock_type_num', and '_sock_flags_num'.
200 * What's new in version 2.2.1, 2013-05-16
201 * What's new in version 2.2, 2013-05-14
203 - Experimental support has been added for probing Java methods using
204 Byteman 2.0 as a backend. Java method probes can target method entries,
205 returns, or specific statements in the method as specified by line number.
207 probe java("org.my.MyApp").class("^java.lang.Object").method("foo(int)")
210 See java/README for information on how to set up Java/Byteman
211 functionality. Set env STAPBM_VERBOSE=yes for more tracing.
213 - The stap -l output and pn() tapset function's return value may be slightly
214 different for complicated web of wildcarded/aliased probes.
216 - The dyninst backend has improved in several aspects:
218 - Setting custom values for global variables is now supported, both
219 with -G when compiling a script, and from the stapdyn command line
220 when loading a precompiled module.
222 - A high-performance shared-memory-based transport is used for
225 - A systemd service file and tmpfile have been added to allow
226 systemtap-server to be managed natively by systemd.
228 - Due to the removal of register_timer_hook in recent kernels, the
229 behaviour of timer.profile has been changed slightly. This probe is
230 now an alias which uses the old mechanism where possible, but falls
231 back to perf.sw.cpu_clock or another mechanism when the kernel timer
232 hook is not available.
234 To require the kernel timer hook mechanism in your script, use
235 timer.profile.tick instead of timer.profile.
237 - The following tapset variables are deprecated in release 2.2:
238 - The 'origin' variables in the 'generic.fop.llseek',
239 'generic.fop.llseek.return', and 'nfs.fop.llseek' probes. The
240 'origin' variable has been replaced by the 'whence' variable.
241 - The 'page_index' variable in the 'vfs.block_sync_page' and
242 'vfs.buffer_migrate_page' probe aliases.
243 - The 'write_from' and 'write_upto' variables in the
244 '_vfs.block_prepare_write' and '_vfs.block_prepare_write.return'
246 - The 'regs' variable in the 'syscall.sigaltstack',
247 'nd_syscall.sigaltstack', 'syscall.fork', and 'nd_syscall.fork'
249 - The 'first', 'second', 'third', and 'uptr_uaddr' variables in the
250 'syscall.compat_sys_shmat' and 'nd_syscall.compat_sys_shmat' probe
253 - The following tapset functions are deprecated in release 2.2:
254 'ppos_pos', '_dev_minor', and '_dev_major'
256 - The folowing tapset functions used to return error strings instead
257 of raising an error. The original behavior is deprecated in release
260 'ctime', 'probemod', 'modname'
262 * What's new in version 2.1, 2013-02-13
264 - EMACS and VIM editor modes for systemtap source files are included / updated.
266 - The translator now eliminates duplicate tapset files between its
267 preferred directory (as configured during the build with --prefix=/
268 or specified with the -I /path option), and files it may find under
269 $XDG_DATA_DIRS. This should eliminate a class of conflicts between
270 parallel system- and hand-built systemtap installations.
272 - The translator accepts a --suppress-time-limits option, which defeats
273 time-related constraints, to allows probe handlers to run for indefinite
274 periods. It requires the guru mode (-g) flag to work. Add the earlier
275 --suppress-handler-errors flag for a gung-ho "just-keep-going" attitude.
277 - Perf event probes may now be read on demand. The counter probe is
278 defined using the counter-name part:
279 probe perf.type(0).config(0).counter("NAME"). The counter is
280 read in a user space probe using @perf("NAME"), e.g.
281 process("PROCESS").statement("func@file") {stat <<< @perf("NAME")}
283 - Perf event probes may now be bound to a specific task using the
284 process-name part: probe perf.type(0).config(0).process("NAME") { }
285 If the probed process name is not specified, then it is inferred
286 from the -c CMD argument.
288 - Some error messages and warnings now refer to additional information
289 that is found in man pages. These are generally named
290 error::FOO or warning::BAR (in the 7stap man page section)
295 - The dyninst backend has improved in several aspects:
296 - The runtime now allows much more concurrency when probing multithreaded
297 processes, and will also follow probes across forks.
298 - Several new probe types are now supported, including timers, function
299 return, and process.begin/end and process.thread.begin/end.
300 - Semaphores for SDT probes are now set properly.
301 - Attaching to existing processes with -x PID now works.
303 - The foreach looping construct can now sort aggregate arrays by the user's
304 choice of aggregating function. Previously, @count was implied. e.g.:
305 foreach ([x,y] in array @sum +) { println(@sum(array[x,y])) }
307 - Proof of concept support for regular expression matching has been added:
308 if ("aqqqqqb" =~ "q*b") { ... }
309 if ("abc" !~ "q*b") { ... }
311 The eventual aim is to support roughly the same functionality as
312 the POSIX Extended Regular Expressions implemented by glibc.
313 Currently missing features include extraction of the matched string
314 and subexpressions, and named character classes ([:alpha:], [:digit:], &c).
316 Special thanks go to the re2c project, whose public domain code this
317 functionality has been based on. For more info on re2c, see:
318 http://sourceforge.net/projects/re2c/
320 - The folowing tapset variables are deprecated in release 2.1 and will
321 be removed in release 2.2:
322 - The 'send2queue' variable in the 'signal.send' probe.
323 - The 'oldset_addr' and 'regs' variables in the 'signal.handle' probe.
325 - The following tapset probes are deprecated in release 2.1 and will
326 be removed in release 2.2:
328 - signal.handle.return
330 * What's new in version 2.0, 2012-10-09
332 - Systemtap includes a new prototype backend, which uses Dyninst to instrument
333 a user's own processes at runtime. This backend does not use kernel modules,
334 and does not require root privileges, but is restricted with respect to the
335 kinds of probes and other constructs that a script may use.
337 Users from source should configure --with-dyninst and install a
338 fresh dyninst snapshot such as that in Fedora rawhide. It may be
339 necessary to disable conflicting selinux checks; systemtap will advise.
341 Select this new backend with the new stap option --runtime=dyninst
342 and a -c target process, along with normal options. (-x target
343 processes are not supported in this prototype version.) For example:
345 stap --runtime=dyninst -c 'stap -l begin' \
346 -e 'probe process.function("main") { println("hi from dyninst!") }'
348 - To aid diagnosis, when a kernel panic occurs systemtap now uses
349 the panic_notifier_list facility to dump a summary of its trace
350 buffers to the serial console.
352 - The systemtap preprocessor now has a simple macro facility as follows:
354 @define add(a,b) %( ((@a)+(@b)) %)
355 @define probegin(x) %(
361 @probegin( foo = @add(40, 2); print(foo) )
363 Macros defined in the user script and regular tapset .stp files are
364 local to the file. To get around this, the tapset library can define
365 globally visible 'library macros' inside .stpm files. (A .stpm file
366 must contain a series of @define directives and nothing else.)
368 The status of the feature is experimental; semantics of macroexpansion
369 may change (unlikely) or expand in the future.
371 - Systemtap probe aliases may be used with additional suffixes
372 attached. The suffixes are passed on to the underlying probe
373 point(s) as shown below:
375 probe foo = bar, baz { }
376 probe foo.subfoo.option("gronk") { }
377 // expands to: bar.subfoo.option("gronk"), baz.subfoo.option("gronk")
379 In practical terms, this allows us to specify additional options to
380 certain tapset probe aliases, by writing e.g.
381 probe syscall.open.return.maxactive(5) { ... }
383 - To support the possibility of separate kernel and dyninst backends,
384 the tapsets have been reorganized into separate folders according to
385 backend. Thus kernel-specific tapsets are located under linux/, the
386 dyninst-specific ones under dyninst/
388 - The backtrace/unwind tapsets have been expanded to allow random
389 access to individual elements of the backtrace. (A caching mechanism
390 ensures that the backtrace computation run at most once for each
391 time a probe fires, regardless of how many times or what order the
392 query functions are called in.) New tapset functions are:
393 stack/ustack - return n'th element of backtrace
394 callers/ucallers - return first n elements of backtrace
395 print_syms/print_usyms - print full information on a list of symbols
396 sprint_syms/sprint_usyms - as above, but return info as a string
398 The following existing functions have been superseded by print_syms()
399 et al.; new scripts are recommended to avoid using them:
405 - The probefunc() tapset function is now myproc-unprivileged, and can
406 now be used in unprivileged scripts for such things as profiling in
407 userspace programs. For instance, try running
408 systemtap.examples/general/para-callgraph.stp in unprivileged mode
409 with a stapusr-permitted probe. The previous implementation of
410 probefunc() is available with "stap --compatible=1.8".
412 - Preprocessor conditional to vary code based on script privilege level:
413 unprivileged -- %( systemtap_privilege == "stapusr" %? ... %)
414 privileged -- %( systemtap_privilege != "stapusr" %? ... %)
415 or, alternately %( systemtap_privilege == "stapsys"
416 || systemtap_privilege == "stapdev" %? ... %)
418 - To ease migration to the embedded-C locals syntax introduced in 1.8
419 (namely, STAP_ARG_* and STAP_RETVALUE), the old syntax can now be
420 re-enabled on a per-function basis using the /* unmangled */ pragma:
422 function add_foo:long(a:long, b:long) %{ /* unmangled */
423 THIS->__retvalue = THIS->a + STAP_ARG_b;
426 Note that both the old and the new syntax may be used in an
427 /* unmangled */ function. Functions not marked /* unmangled */
428 can only use the new syntax.
430 - Adjacent string literals are now glued together irrespective of
431 intervening whitespace or comments:
432 "foo " "bar" --> "foo bar"
433 "foo " /* comment */ "bar" --> "foo bar"
434 Previously, the first pair of literals would be glued correctly,
435 while the second would cause a syntax error.
437 * What's new in version 1.8, 2012-06-17
439 - staprun accepts a -T timeout option to allow less frequent wake-ups
440 to poll for low-throughput output from scripts.
442 - When invoked by systemtap, the kbuild $PATH environment is sanitized
443 (prefixed with /usr/bin:/bin:) in an attempt to exclude compilers
444 other than the one the kernel was presumed built with.
446 - Printf formats can now use "%#c" to escape non-printing characters.
448 - Pretty-printed bitfields use integers and chars use escaped formatting
451 - The systemtap compile-server and client now support IPv6 networks.
452 - IPv6 addresses may now be specified on the --use-server option and will
453 be displayed by --list-servers, if the avahi-daemon service is running and
455 - Automatic server selection will automatically choose IPv4 or IPv6 servers
456 according to the normal server selection criteria when avahi-daemon is
457 running. One is not preferred over the other.
458 - The compile-server will automatically listen on IPv6 addresses, if
460 - To enable IPv6 in avahi-daemon, ensure that /etc/avahi/avahi-daemon.conf
461 contains an active "use-ipv6=yes" line. After adding this line run
462 "service avahi-daemon restart" to activate IPv6 support.
463 - See man stap(1) for details on how to use IPv6 addresses with the
466 - Support for DWARF4 .debug_types sections (for executables and shared
467 libraries compiled with recent GCC's -gdwarf-4 / -fdebug-types-section).
468 PR12997. SystemTap now requires elfutils 0.148+, full .debug_types support
469 depends on elfutils 0.154+.
471 - Systemtap modules are somewhat smaller & faster to compile. Their
472 debuginfo is now suppressed by default; use -B CONFIG_DEBUG_INFO=y to
475 - @var now an alternative language syntax for accessing DWARF variables
476 in uprobe and kprobe handlers (process, kernel, module). @var("somevar")
477 can be used where $somevar can be used. The @var syntax also makes it
478 possible to access non-local, global compile unit (CU) variables by
479 specifying the CU source file as follows @var("somevar@some/src/file.c").
480 This will provide the target variable value of global "somevar" as defined
481 in the source file "some/src/file.c". The @var syntax combines with all
482 normal features of DWARF target variables like @defined(), @entry(),
483 [N] array indexing, field access through ->, taking the address with
484 the & prefix and shallow or deep pretty printing with a $ or $$ suffix.
486 - Stap now has resource limit options:
492 All resource limiting has been moved from the compile server to stap
493 itself. When running the server as "stap-server", default resource
494 limit values are specified in ~stap-server/.systemtap/rc.
496 - Bug CVE-2012-0875 (kernel panic when processing malformed DWARF unwind data)
499 - The systemtap compile-server now supports multiple concurrent connections.
500 Specify the desired maximum number of concurrent connections with
501 the new stap-server/stap-serverd --max-threads option. Specify a
502 value of '0' to tell the server not to spawn any new threads (handle
503 all connections serially in the main thread). The default value is
504 the number of processor cores on the host.
506 - The following tapset functions are deprecated in release 1.8 and will be
507 removed in release 1.9:
510 - SystemTap now mangles local variables to avoid collisions with C
511 headers included by tapsets. This required a change in how
512 embedded-C functions access local parameters and the return value slot.
514 Instead of THIS->foo in an embedded-C function, please use the newly
515 defined macro STAP_ARG_foo (substitute the actual name of the
516 argument for 'foo'); instead of THIS->__retvalue, use the newly
517 defined STAP_RETVALUE. All of the tapsets and test cases have been
518 adapted to use this new notation.
520 If you need to run code which uses the old THIS-> notation, run stap
521 with the --compatible=1.7 option.
523 - There is updated support for user-space probing against kernels >=
524 3.5, which have no utrace but do have the newer inode-uprobes work
525 by Srikar Dronamraju and colleagues. For kernels < 3.5, the
526 following 3 sets of kernel patches would need to be backported to
527 your kernel to use this preliminary user-space probing support:
529 - inode-uprobes patches:
530 - 2b144498350860b6ee9dc57ff27a93ad488de5dc
531 - 7b2d81d48a2d8e37efb6ce7b4d5ef58822b30d89
532 - a5f4374a9610fd7286c2164d4e680436727eff71
533 - 04a3d984d32e47983770d314cdb4e4d8f38fccb7
534 - 96379f60075c75b261328aa7830ef8aa158247ac
535 - 3ff54efdfaace9e9b2b7c1959a865be6b91de96c
536 - 35aa621b5ab9d08767f7bc8d209b696df281d715
537 - 900771a483ef28915a48066d7895d8252315607a
538 - e3343e6a2819ff5d0dfc4bb5c9fb7f9a4d04da73
539 - exec tracepoint kernel patch:
540 - 4ff16c25e2cc48cbe6956e356c38a25ac063a64d
541 - task_work_add kernel patches:
542 - e73f8959af0439d114847eab5a8a5ce48f1217c4
543 - 4d1d61a6b203d957777d73fcebf19d90b038b5b2
544 - 413cd3d9abeaef590e5ce00564f7a443165db238
545 - dea649b8ac1861107c5d91e1a71121434fc64193
546 - f23ca335462e3c84f13270b9e65f83936068ec2c
548 * What's new in version 1.7, 2012-02-01
550 - Map inserting and deleting is now significantly faster due to
551 improved hashing and larger hash tables. The hashes are also
552 now randomized to provide better protection against deliberate
555 - Formatted printing is faster by compiling the formatting directives
556 to C code rather than interpreting at run time.
558 - Systemtap loads extra command line options from $SYSTEMTAP_DIR/rc
559 ($HOME/.systemtap/rc by default) before the normal argc/argv. This
560 may be useful to activate site options such as --use-server or
561 --download-debuginfo or --modinfo.
563 - The stap-server has seen many improvements, and is no longer considered
566 - The stap-server service (initscript) now supports four new options:
571 These allow the specification of macro definitions to be passed to stap
572 by the server, the location of the log file, network port number and
573 NSS certificate database location respectively. These options are also
574 supported within individual server configuration files. See stap-server
575 and initscript/README.stap-server for details. The stap-server is no
576 longer activated by default.
578 - process("PATH").[library("PATH")].function("NAME").exported probes are now
579 supported to filter function() to only exported instances.
581 - The translator supports a new --suppress-handler-errors option, which
582 causes most runtime errors to be turned into quiet skipped probes. This
583 also disables the MAXERRORS and MAXSKIPPED limits.
585 - Translator warnings have been standardized and controlled by the -w / -W
588 - The translator supports a new --modinfo NAME=VALUE option to emit additional
589 MODULE_INFO(n,v) macros into the generated code.
591 - There is no more fixed maximum number of VMA pages that will be tracked
592 at runtime. This reduces memory use for those scripts that don't need any,
593 or only limited target process VMA tracking and allows easier system
594 wide probes inspecting shared library variables and/or user backtraces.
595 stap will now silently ignore -DTASK_FINDER_VMA_ENTRY_ITEMS.
597 - The tapset functions remote_id() and remote_uri() identify the member of a
598 swarm of "stap --remote FOO --remote BAR baz.stp" concurrent executions.
600 - Systemtap now supports a new privilege level and group, "stapsys", which
601 is equivalent to the privilege afforded by membership in the group "stapdev",
602 except that guru mode (-g) functionality may not be used. To support this, a
603 new option, --privilege=[stapusr|stapsys|stapdev] has been added.
604 --privilege=stapusr is equivalent to specifying the existing --unprivileged
605 option. --privilege=stapdev is the default. See man stap(1) for details.
607 - Scripts that use kernel.trace("...") probes compile much faster.
609 - The systemtap module cache is cleaned less frequently, governed by the
610 number of seconds in the $SYSTEMTAP_DIR/cache/cache_clean_interval_s file.
612 - SDT can now define up to 12 arguments in a probe point.
614 - Parse errors no longer generate a cascade of false errors. Instead, a
615 parse error skips the rest of the current probe or function, and resumes
616 at the next one. This should generate fewer and better messages.
618 - Global array wrapping is now supported for both associative and statistics typed
619 arrays using the '%' character to signify a wrapped array. For example,
620 'global foo%[100]' would allow the array 'foo' to be wrapped if more than 100
621 elements are inserted.
623 - process("PATH").library("PATH").plt("NAME") probes are now supported.
624 Wildcards are supported in the plt-name part, to refer to any function in the
625 program linkage table which matches the glob pattern and the rest of the
628 - A new option, --dump-probe-types, will dump a list of supported probe types.
629 If --unprivileged is also specified, the list will be limited to probe types
630 which are available to unprivileged users.
632 - Systemtap can now automatically download the required debuginfo
633 using abrt. The --download-debuginfo[=OPTION] can be used to
634 control this feature. Possible values are: 'yes', 'no', 'ask',
635 and a positive number representing the timeout desired. The
636 default behavior is to not automatically download the debuginfo.
638 - The translator has better support for probing C++ applications by
639 better undertanding of compilation units, nested types, templates,
640 as used in probe point and @cast constructs.
642 - On 2.6.29+ kernels, systemtap can now probe kernel modules that
643 arrive and/or depart during the run-time of a session. This allows
644 probing of device driver initialization functions, which had formerly been
647 - New tapset functions for cpu_clock and local_clock access were added.
649 - There is some limited preliminary support for user-space probing
650 against kernels such as linux-next, which have no utrace but do have
651 the newer inode-uprobes work by Srikar Dronamraju and colleagues.
653 - The following probe types are deprecated in release 1.7 and will be
654 removed in release 1.8:
655 kernel.function(number).inline
656 module(string).function(number).inline
657 process.function(number).inline
658 process.library(string).function(number).inline
659 process(string).function(number).inline
660 process(string).library(string).function(number).inline
662 - The systemtap-grapher is deprecated in release 1.7 and will be removed in
665 - The task_backtrace() tapset function was deprecated in 1.6 and has been
668 - MAXBACKTRACE did work in earlier releases, but has now been documented
669 in the stap 1 manual page.
671 - New tapset function probe_type(). Returns a short string describing
672 the low level probe handler type for the current probe point.
674 - Both unwind and symbol data is now only collected and emitted for
675 scripts actually using backtracing or function/data symbols.
676 Tapset functions are marked with /* pragma:symbols */ or
677 /* pragma:unwind */ to indicate they need the specific data.
679 - Kernel backtraces can now be generated for non-pt_regs probe context
680 if the kernel support dump_trace(). This enables backtraces from
681 certain timer probes and tracepoints.
683 - ubacktrace() should now also work for some kernel probes on x86 which can
684 use the dwarf unwinder to recover the user registers to provide
685 more accurate user backtraces.
687 - For s390x the systemtap runtime now properly splits kernel and user
688 addresses (which are in separate address spaces on that architecture)
689 which enable user space introspection.
691 - ppc and s390x now supports user backtraces through the DWARF unwinder.
693 - ppc now handles function descriptors as symbol names correctly.
695 - arm support kernel backtraces through the DWARF unwinder.
697 - arm now have a uprobes port which enables user probes. This still
698 requires some kernel patches (user_regsets and tracehook support for
701 - Starting in release 1.7, these old variables will be deprecated:
702 - The 'pid' variable in the 'kprocess.release' probe has been
703 deprecated in favor of the new 'released_pid' variable.
704 - The 'args' variable in the
705 '_sunrpc.clnt.create_client.rpc_new_client_inline' probe has been
706 deprecated in favor of the new internal-only '__args' variable.
708 - Experimental support for recent kernels without utrace has been
709 added for the following probe types:
712 process("PATH").begin
714 process(PID).thread.begin
715 process("PATH").thread.begin
720 process(PID).thread.end
721 process("PATH").thread.end
724 process("PATH").syscall
726 process(PID).syscall.return
727 process("PATH").syscall.return
728 process.syscall.return
730 - staprun disables kprobe-optimizations in recent kernels, as problems
731 were found. (PR13193)
733 * What's new in version 1.6, 2011-07-25
735 - Security fixes for CVE-2011-2503: read instead of mmap to load modules,
736 CVE-2011-2502: Don't allow path-based auth for uprobes
738 - The systemtap compile-server no longer uses the -k option when calling the
739 translator (stap). As a result, the server will now take advantage of the
740 module cache when compiling the same script more than once. You may observe
741 an improvement in the performance of the server in this situation.
743 - The systemtap compile-server and client now each check the version of the
744 other, allowing both to adapt when communicating with a down-level
745 counterpart. As a result, all version of the client can communicate
746 with all versions of the server and vice-versa. Client will prefer newer
747 servers when selecting a server automatically.
749 - SystemTap has improved support for the ARM architecture. The
750 kread() and kwrite() operations for ARM were corrected allowing many
751 of the tapsets probes and function to work properly on the ARM
754 - Staprun can now rename the module to a unique name with the '-R' option before
755 inserting it. Systemtap itself will also call staprun with '-R' by default.
756 This allows the same module to be inserted more than once, without conflicting
759 - Systemtap error messages now provide feedback when staprun or any other
760 process fails to launch. This also specifically covers when the user
761 doesn't have the proper permissions to launch staprun.
763 - Systemtap will now map - to _ in module names. Previously,
764 stap -L 'module("i2c-core").function("*")' would be empty. It now returns
765 a list had stap -L 'module("i2c_core").function("*") been specified.
767 - Systemtap now fills in missing process names to probe points, to
768 avoid having to name them twice twice:
769 % stap -e 'probe process("a.out").function("*") {}' -c 'a.out ...'
770 Now the probed process name is inferred from the -c CMD argument.
771 % stap -e 'probe process.function("*") {}' -c 'a.out ...'
773 - stap -L 'process("PATH").syscall' will now list context variables
775 - Depends on elfutils 0.142+.
777 - Deprecated task_backtrace:string (task:long). This function will go
778 away after 1.6. Please run your scripts with stap --check-version.
780 * What's new in version 1.5, 2011-05-23
782 - Security fixes for CVE-2011-1781, CVE-2011-1769: correct DW_OP_{mod,div}
785 - The compile server and its related tools (stap-gen-ert, stap-authorize-cert,
786 stap-sign-module) have been re-implemented in C++. Previously, these
787 components were a mix of bash scripts and C code. These changes should be
788 transparent to the end user with the exception of NSS certificate database
789 password prompting (see below). The old implementation would prompt more
790 than once for the same password in some situations.
792 - eventcount.stp now allows for event counting in the format of
793 'stap eventcount.stp process.end syscall.* ...', and also reports
794 corresponding event tid's.
796 - Systemtap checks that the build-id of the module being probed matches the
797 build-id saved in the systemtap module. Invoking systemtap with
798 -DSTP_NO_BUILDID_CHECK will bypass this build-id runtime verification. See
799 man ld(1) for info on --build-id.
801 - stapio will now report if a child process has an abnormal exit along with
802 the associated status or signal.
804 - Compiler optimization may sometimes result in systemtap not being able to
805 access a user-space probe argument. Compiling the application with
806 -DSTAP_SDT_ARG_CONSTRAINT=nr will force the argument to be an immediate or
807 register value which should enable systemtap to access the argument.
809 - GNU Gettext has now been intergrated with systemtap. Our translation
810 page can be found at http://www.transifex.net/projects/p/systemtap/ .
811 "make update-po" will generate the necessary files to use translated
812 messages. Please refer to the po/README file for more info and
813 please consider contributing to this I18N effort!
815 - The new addr() function returns the probe's instruction pointer.
817 - process("...").library("...") probes are now supported. Wildcards
818 are supported in the library-name part, to refer to any shared
819 library that is required by process-name, which matches the glob
820 pattern and the rest of the probe point.
822 - The "--remote USER@HOST" functionality can now be specified multiple times
823 to fan out on multiple targets. If the targets have distinct kernel and
824 architecture configurations, stap will automatically build the script
825 appropriately for each one. This option is also no longer considered
828 - The NSS certificate database generated for use by the compile server is now
829 generated with no password. Previously, a random password was generated and
830 used to access the database. This change should be transparent to most users.
831 However, if you are prompted for a password when using systemtap, then
832 running $libexecdir/stap-gen-cert should correct the problem.
834 - The timestamp tapset includes jiffies() and HZ() for lightweight approximate
837 - A powerful new command line option --version has been added.
839 - process.mark now supports $$parms for reading probe parameters.
841 - A new command line option, --use-server-on-error[=yes|no] is available
842 for stap. It instructs stap to retry compilation of a script using a
843 compile server if it fails on the local host. The default setting
846 - The following deprecated tools have been removed:
848 stap-authorize-server-cert
849 stap-authorize-signing-cert
850 stap-find-or-start-server
852 Use the --use-server, --trust-server and --list-servers options of stap
855 * What's new in version 1.4, 2011-01-17
857 - Security fixes for CVE-2010-4170, CVE-2010-4171: staprun module
860 - A new /* myproc-unprivileged */ marker is now available for embedded C
861 code and and expressions. Like the /* unprivileged */ marker, it makes
862 the code or expression available for use in unprivileged mode (see
863 --unprivileged). However, it also automatically adds a call to
864 assert_is_myproc() to the code or expression, thus, making it available
865 to the unprivileged user only if the target of the current probe is within
866 the user's own process.
868 - The experimental "--remote USER@HOST" option will run pass 5 on a given
869 ssh host, after building locally (or with --use-server) for that target.
871 - Warning messages from the script may now be suppressed with the stap
872 and/or staprun -w option. By default, duplicate warning messages are
873 suppressed (up to a certain limit). With stap --vp 00002 and above,
874 the duplicate elimination is defeated.
876 - The print_ubacktrace and usym* functions attempt to print the full
877 path of the user-space binaries' paths, instead of just the basename.
878 The maximum saved path length is set by -DTASK_FINDER_VMA_ENTRY_PATHLEN,
879 default 64. Warning messages are produced if unwinding fails due to
880 a missing 'stap -d MODULE' option, providing preloaded unwind data.
882 - The new tz_ctime() tapset function prints times in the local time zone.
884 - More kernel tracepoints are accessible to the kernel.trace("...") mechanism,
885 if kernel source trees or debuginfo are available. These formerly "hidden"
886 tracepoints are those that are declared somewhere other than the usual
887 include/linux/trace/ headers, such as xfs and kvm.
889 - debuginfo-based process("...").function/.statement/.mark probes support
890 wildcards in the process-name part, to refer to any executable files that
891 match the glob pattern and the rest of the probe point.
893 - The -t option now displays information per probe-point rather than a summary
894 for each probe. It also now shows the derivation chain for each probe-point.
896 - A rewrite of the sys/sdt.h header file provides zero-cost startup (few or
897 no ELF relocations) for the debuginfo-less near-zero-cost runtime probes.
898 Binaries compiled with earlier sdt.h versions remain supported. The
899 stap -L (listing) option now lists parameters for sys/sdt.h markers.
901 - The implementation of the integrated compile-server client has been
903 o --use-server now accepts an argument representing a particular server and
904 may be specified more than once.
905 o --list-servers now accepts an expanded range of arguments.
906 o a new --trust-servers option has been added to stap to replace several
907 old certificate-management scripts.
908 o The following tools are now deprecated and will be removed in release 1.5:
910 stap-authorize-server-cert
911 stap-authorize-signing-cert
912 stap-find-or-start-server
914 See man stap(1) for complete details.
916 - The compile-server now returns the uprobes.ko to the client when it is
917 required by the script being compiled. The integrated compile-server client
918 now makes it available to be loaded by staprun. The old (deprecated)
919 stap-client does not do this.
921 - process probes with scripts as the target are recognized by stap and the
922 interpreter would be selected for probing.
924 - Starting in release 1.5, these old variables/functions will be deprecated
925 and will only be available when the '--compatible=1.4' flag is used:
927 - In the 'syscall.add_key' probe, the 'description_auddr' variable
928 has been deprecated in favor of the new 'description_uaddr'
930 - In the 'syscall.fgetxattr', 'syscall.fsetxattr',
931 'syscall.getxattr', 'syscall.lgetxattr', and
932 'syscall.lremovexattr' probes, the 'name2' variable has been
933 deprecated in favor of the new 'name_str' variable.
934 - In the 'nd_syscall.accept' probe the 'flag_str' variable
935 has been deprecated in favor of the new 'flags_str' variable.
936 - In the 'nd_syscall.dup' probe the 'old_fd' variable has been
937 deprecated in favor of the new 'oldfd' variable.
938 - In the 'nd_syscall.fgetxattr', 'nd_syscall.fremovexattr',
939 'nd_syscall.fsetxattr', 'nd_syscall.getxattr', and
940 'nd_syscall.lremovexattr' probes, the 'name2' variable has been
941 deprecated in favor of the new 'name_str' variable.
942 - The tapset alias 'nd_syscall.compat_pselect7a' was misnamed. It should
943 have been 'nd_syscall.compat_pselect7' (without the trailing 'a').
944 - The tapset function 'cpuid' is deprecated in favor of the better known
946 - In the i386 'syscall.sigaltstack' probe, the 'ussp' variable has
947 been deprecated in favor of the new 'uss_uaddr' variable.
948 - In the ia64 'syscall.sigaltstack' probe, the 'ss_uaddr' and
949 'oss_uaddr' variables have been deprecated in favor of the new
950 'uss_uaddr' and 'uoss_uaddr' variables.
951 - The powerpc tapset alias 'syscall.compat_sysctl' was deprecated
952 and renamed 'syscall.sysctl32'.
953 - In the x86_64 'syscall.sigaltstack' probe, the 'regs_uaddr'
954 variable has been deprecated in favor of the new 'regs' variable.
956 * What's new in version 1.3, 2010-07-21
958 - The uprobes kernel module now has about half the overhead when probing
959 NOPs, which is particularly relevant for sdt.h markers.
961 - New stap option -G VAR=VALUE allows overriding global variables
962 by passing the settings to staprun as module options.
964 - The tapset alias 'syscall.compat_pselect7a' was misnamed. It should
965 have been 'syscall.compat_pselect7' (without the trailing 'a').
966 Starting in release 1.4, the old name will be deprecated and
967 will only be available when the '--compatible=1.3' flag is used.
969 - A new procfs parameter .umask(UMASK) which provides modification of
970 file permissions using the proper umask value. Default file
971 permissions for a read probe are 0400, 0200 for a write probe, and
972 0600 for a file with a read and write probe.
974 - It is now possible in some situations to use print_ubacktrace() to
975 get a user space stack trace from a kernel probe point. e.g. for
976 user backtraces when there is a pagefault:
977 $ stap -d /bin/sort --ldd -e 'probe vm.pagefault {
978 if (pid() == target()) {
979 printf("pagefault @0x%x\n", address); print_ubacktrace();
982 pagefault @0x7fea0595fa70
983 0x000000384f07f958 : __GI_strcmp+0x12b8/0x1440 [libc-2.12.so]
984 0x000000384f02824e : __gconv_lookup_cache+0xee/0x5a0 [libc-2.12.so]
985 0x000000384f021092 : __gconv_find_transform+0x92/0x2cf [libc-2.12.so]
986 0x000000384f094896 : __wcsmbs_load_conv+0x106/0x2b0 [libc-2.12.so]
987 0x000000384f08bd90 : mbrtowc+0x1b0/0x1c0 [libc-2.12.so]
988 0x0000000000404199 : ismbblank+0x39/0x90 [sort]
989 0x0000000000404a4f : inittables_mb+0xef/0x290 [sort]
990 0x0000000000406934 : main+0x174/0x2510 [sort]
991 0x000000384f01ec5d : __libc_start_main+0xfd/0x1d0 [libc-2.12.so]
992 0x0000000000402509 : _start+0x29/0x2c [sort]
995 - New tapset functions to get a string representation of a stack trace:
996 sprint_[u]backtrace() and sprint_[u]stack().
998 - New tapset function to get the module (shared library) name for a
999 user space address umodname:string(long). The module name will now
1000 also be in the output of usymdata() and in backtrace addresses even
1001 when they were not given with -d at the command line.
1003 - Kernel backtraces are now much faster (replaced a linear search
1004 with a binary search).
1006 - A new integrated compile-server client is now available as part of stap.
1008 o 'stap --use-server ...' is equivalent to 'stap-client ...'
1009 o 'stap --list-servers' is equivalent to 'stap-find-servers'
1010 o 'stap --list-servers=online' is equivalent to 'stap-find-servers --all'
1011 o stap-client and its related tools will soon be deprecated.
1012 o the nss-devel and avahi-devel packages are required for building stap with
1013 the integrated client (checked during configuration).
1014 o nss and avahi are required to run the integrated client.
1016 - A new operator @entry is available for automatically saving an expression
1017 at entry time for use in a .return probe.
1018 probe foo.return { println(get_cycles() - @entry(get_cycles())) }
1020 - Probe $target variables and @cast() can now use a suffix to print complex
1021 data types as strings. Use a single '$' for a shallow view, or '$$' for a
1022 deeper view that includes nested types. For example, with fs_struct:
1023 $fs$ : "{.users=%i, .lock={...}, .umask=%i,
1024 .in_exec=%i, .root={...}, .pwd={...}}"
1025 $fs$$ : "{.users=%i, .lock={.raw_lock={.lock=%u}}, .umask=%i, .in_exec=%i,
1026 .root={.mnt=%p, .dentry=%p}, .pwd={.mnt=%p, .dentry=%p}}"
1028 - The <sys/sdt.h> user-space markers no longer default to an implicit
1029 MARKER_NAME_ENABLED() semaphore check for each marker. To check for
1030 enabled markers use a .d declaration file, then:
1031 if (MARKER_NAME_ENABLED()) MARKER_NAME()
1033 - Hyphenated <sys/sdt.h> marker names such as process(...).mark("foo-bar")
1034 are now accepted in scripts. They are mapped to the double-underscore
1037 - More robust <sys/sdt.h> user-space markers support is included. For
1038 some platforms (x86*, ppc*), this can let systemtap probe the markers
1039 without debuginfo. This implementation also supports preserving
1040 the "provider" name associated with a marker:
1041 probe process("foo").provider("bar").mark("baz") to match
1042 STAP_PROBE<n>(bar, baz <...>)
1043 (Compile with -DSTAP_SDT_V1 to revert to the previous implementation.
1044 Systemtap supports pre-existing or new binaries using them.)
1046 - Embedded-C may be used within expressions as values, when in guru mode:
1047 num = %{ LINUX_VERSION_CODE %} // int64_t
1048 name = %{ /* string */ THIS_MODULE->name %} // const char*
1049 printf ("%s %x\n", name, num)
1050 The usual /* pure */, /* unprivileged */, and /* guru */ markers may be used
1051 as with embedded-C functions.
1053 - By default the systemtap-runtime RPM builds now include a shared
1054 library, staplog.so, that allows crash to extract systemtap data from
1057 - Iterating with "foreach" can now explicitly save the value for the loop.
1058 foreach(v = [i,j] in array)
1059 printf("array[%d,%s] = %d\n", i, j, v /* array[i,j] */)
1061 - The new "--ldd" option automatically adds any additional shared
1062 libraries needed by probed or -d-listed userspace binaries to the -d
1063 list, to enable symbolic backtracing through them. Similarly, the
1064 new "--all-modules" option automatically adds any currently loaded
1065 kernel modules (listed in /proc/modules) to the -d list.
1067 - A new family of set_kernel_* functions make it easier for gurus to write
1068 new values at arbitrary memory addresses.
1070 - Probe wildcards can now use '**' to cross the '.' separator.
1071 $ stap -l 'sys**open'
1075 - Backward compatibility flags (--compatible=VERSION, and matching
1076 script preprocessing predicate %( systemtap_v CMP "version" %)
1077 and a deprecation policy are being introduced, in case future
1078 tapset/language changes break valid scripts.
1080 * What's new in version 1.2, 2010-03-22
1082 - Prototype support for "perf events", where the kernel supports the
1083 2.6.33 in-kernel API. Probe points may refer to low-level
1084 perf_event_attr type/config numbers, or to a number of aliases
1085 defined in the new perf.stp tapset:
1086 probe perf.sw.cpu_clock, perf.type(0).config(4) { }
1088 - Type-casting can now use multiple headers to resolve codependencies.
1089 @cast(task, "task_struct",
1090 "kernel<linux/sched.h><linux/fs_struct.h>")->fs->umask
1092 - Tapset-related man pages have been renamed. 'man -k 3stap' should show
1093 the installed list, which due to prefixing should no longer collide over
1094 ordinary system functions.
1096 - User space marker arguments no longer use volatile if the version of gcc,
1097 which must be at least 4.5.0, supports richer DWARF debuginfo. Use cflags
1098 -DSTAP_SDT_VOLATILE=volatile or -DSTAP_SDT_VOLATILE= when building
1099 the sys/sdt.h application to override this one way or another.
1101 - A new construct for error handling is available. It is similar to c++
1102 exception catching, using try and catch as new keywords. Within a handler
1103 or function, the following is valid and may be nested:
1104 try { /* arbitrary statements */ }
1105 catch (er) { /* e.g. println("caught error ", er) */ }
1107 - A new command line flag '-W' forces systemtap to abort translation of
1108 a script if any warnings are produced. It is similar to gcc's -Werror.
1109 (If '-w' is also supplied to suppress warnings, it wins.)
1111 - A new predicate @defined is available for testing whether a
1112 particular $variable/expression is resolvable at translate time:
1113 probe foo { if (@defined($bar)) log ("$bar is available here") }
1115 - Adjacent string literals are glued together, making this
1117 probe process("/usr" @1 "/bin").function("*") { ... }
1119 - In order to limit potential impact from future security problems,
1120 the stap-server process does not permit its being launched as root.
1122 - On recent kernels, for some architectures/configurations, hardware
1123 breakpoint probes are supported. The probe point syntax is:
1125 probe kernel.data(ADDRESS).write
1126 probe kernel.data(ADDRESS).length(LEN).write
1127 probe kernel.data("SYMBOL_NAME").write
1129 * What's new in version 1.1, 2010-01-15
1131 - New tracepoint based tapset for memory subsystem.
1133 - The loading of signed modules by staprun is no longer allowed for
1134 ordinary, unprivileged users. This means that only root, members of
1135 the group 'stapdev' and members of the group 'stapusr' can load
1136 systemtap modules using staprun, stap or stap-client. The minimum
1137 privilege required to run arbitrary --unprivileged scripts is now
1138 'stapusr' membership.
1140 - The stap-server initscript is available. This initscript allows you
1141 to start systemtap compile servers as a system service and to manage
1142 these servers as a group or individually. The stap-server initscript
1143 is installed by the systemtap-server rpm. The build directory for
1144 the uprobes module (/usr/share/systemtap/runtime/uprobes) is made
1145 writable by the 'stap-server' group. All of the files generated when
1146 building the uprobes module, including the digital signature, are
1147 also writable by members of stap-server.
1149 See initscript/README.stap-server for details.
1151 - Some of the compile server client, server and certificate management
1152 tools have been moved from $bindir to $libexecdir/systemtap.
1153 You should use the new stap-server script or the stap-server initscript
1154 for server management where possible. The stap-server script provides the same
1155 functionality as the stap-server initscript except that the servers are
1156 run by the invoking user by default as opposed to servers started by the
1157 stap-server initscript which are run by the user stap-server
1158 by default. See stap-server(8) for more information.
1160 You may continue to use these tools by adding $libexecdir/systemtap to
1161 your path. You would need to do this, for example, if you are not root,
1162 you want to start a compile server and you are not running systemtap from a
1163 private installation. In this case you still need to use stap-start-server.
1165 - Any diagnostic output line that starts with "ERROR", as in
1166 error("foo"), will promote a "Pass 5: run failed", and the return
1169 - Systemtap now warns about global variables being referenced from other
1170 script files. This aims to protect against unintended local-vs-global
1171 namespace collisions such as:
1173 % cat some_tapset.stp
1174 probe baz.one = bar { foo = $foo; bar = $bar }
1175 % cat end_user_script.stp
1176 global foo # intended to be private variable
1177 probe timer.s(1) { foo ++ }
1178 probe baz.* { println(foo, pp()) }
1179 % stap end_user_script.stp
1180 WARNING: cross-file global variable reference to foo from some_tapset.stp
1182 - Preprocessor conditional for kernel configuration testing:
1183 %( CONFIG_foo == "y" %? ... %)
1185 - ftrace(msg:string) tapset function to send strings to the system-wide
1186 ftrace ring-buffer (if any).
1188 - Better support for richer DWARF debuginfo output from GCC 4.5
1189 (variable tracking assignments). Kernel modules are now always resolved
1190 against all their dependencies to find any info referring to missing
1191 symbols. DW_AT_const_value is now supported when no DW_AT_location
1194 * What's new in verson 1.0, 2009-09-22
1196 - process().mark() probes now use an enabling semaphore to reduce the
1197 computation overhead of dormant probes.
1199 - The function spec for dwarf probes now supports C++ scopes, so you can
1200 limit the probes to specific namespaces or classes. Multiple scopes
1201 can be specified, and they will be matched progressively outward.
1202 probe process("foo").function("std::vector<*>::*") { }
1203 probe process("foo").function("::global_function") { }
1205 - It is now possible to cross-compile systemtap scripts for foreign
1206 architectures, using the new '-a ARCH' and '-B OPT=VALUE' flags.
1207 For example, put arm-linux-gcc etc. into your $PATH, and point
1208 systemtap at the target kernel build tree with:
1209 stap -a arm -B CROSS_COMPILE=arm-linux- -r /build/tree [...]
1210 The -B option is passed to kbuild make. -r identifies the already
1211 configured/built kernel tree and -a its architecture (kbuild ARCH=...).
1212 Systemtap will infer -p4.
1214 - Cross compilation using the systemtap client and server
1215 - stap-start-server now accepts the -r, -R, -I, -B and -a options in
1216 order to start a cross compiling server. The server will correctly
1217 advertise itself with respect to the kernel release and architecture
1218 that it compiles for.
1219 - When specified on stap-client, the -r and -a options will be
1220 considered when searching for a suitable server.
1222 - When using the systemtap client and server udp port 5353 must be open
1223 in your firewall in order for the client to find servers using
1224 avahi-browse. Also the systemtap server will choose a random port in
1225 the range 1024-63999 for accepting ssl connections.
1227 - Support for unprivileged users:
1228 ***********************************************************************
1229 * WARNING!!!!!!!!!! *
1230 * This feature is EXPERIMENTAL at this time and should be used with *
1231 * care. This feature allows systemtap kernel modules to be loaded by *
1232 * unprivileged users. The user interface and restrictions will change *
1233 * as this feature evolves. *
1234 ***********************************************************************
1235 - Systemtap modules generated from scripts which use a restricted
1236 subset of the features available may be loaded by staprun for
1237 unprivileged users. Previously, staprun would load modules only for
1238 root or for members of the groups stapdev and stapusr.
1239 - Using the --unprivileged option on stap enables translation-time
1240 checking for use by unprivileged users (see restrictions below).
1241 - All modules deemed suitable for use by unprivileged users will be
1242 signed by the systemtap server when --unprivileged is specified on
1243 stap-client. See module signing in release 0.9.8 and stap-server in
1245 - Modules signed by trusted signers (servers) and verified by staprun
1246 will be loaded by staprun regardless of the user's privilege level.
1247 - The system administrator asserts the trustworthiness of a signer
1248 (server) by running stap-authorize-signing-cert <cert-file> as root,
1249 where the <cert-file> can be found in
1250 ~<user>/.systemtap/ssl/server/stap.cert for servers started by
1251 ordinary users and in $sysconfdir/systemtap/ssl/server/stap.cert for
1252 servers started by root.
1253 - Restrictions are intentionally strict at this time and may be
1254 relaxed in the future:
1255 - probe points are restricted to:
1256 begin, begin(n), end, end(n), error, error(n), never,
1257 timer.{jiffies,s,sec,ms,msec,us,usec,ns,nsec}(n)*, timer.hz(n),
1258 process.* (for processes owned by the user).
1259 - use of embedded C code is not allowed.
1260 - use of tapset functions is restricted.
1261 - some tapset functions may not be used at all. A message will be
1262 generated at module compilation time.
1263 - some actions by allowed tapset functions may only be performed
1264 in the context of the user's own process. A runtime fault will
1265 occur in these situations, for example, direct memory access.
1266 - The is_myproc() tapset function has been provided so that
1267 tapset writers for unprivileged users can check that the
1268 context is of the users own process before attempting these
1270 - accessing the kernel memory space is not allowed.
1271 - The following command line options may not be used by stap-client
1273 - The following environment variables are ignored by stap-client:
1274 SYSTEMTAP_RUNTIME, SYSTEMTAP_TAPSET, SYSTEMTAP_DEBUGINFO_PATH
1275 - nss and nss-tools are required to use this feature.
1277 - Support output file switching by SIGUSR2. Users can command running
1278 stapio to switch output file by sending SIGUSR2.
1280 - Memory consumption for scripts involving many uprobes has been
1281 dramatically reduced.
1283 - The preprocessor now supports || and && in the conditions.
1284 e.g. %( arch == "x86_64" || arch == "ia64" %: ... %)
1286 - The systemtap notion of "architecture" now matches the kernel's, rather
1287 than that of "uname -m". This means that 32-bit i386 family are all
1288 known as "i386" rather than "i386" or "i686"; "ppc64" as "powerpc";
1289 "s390x" as "s390", and so on. This is consistent between the new
1290 "-a ARCH" flag and the script-level %( arch ... %) conditional.
1292 - It is now possible to define multiple probe aliases with the same name.
1293 A probe will expand to all matching aliases.
1296 probe foo { } # expands twice, once to bar and once to baz
1298 - A new experimental transport mechanism, using ftrace's ring_buffer,
1299 has been added. This may become the default transport mechanism in
1300 future versions of systemtap. To test this new transport mechanism,
1301 define 'STP_USE_RING_BUFFER'.
1303 - Support for recognizing DW_OP_{stack,implicit}_value DWARF expressions
1304 as emitted by GCC 4.5.
1306 * What's new in version 0.9.9, 2009-08-04
1308 - Systemwide kernel .function.return (kretprobe) maxactive defaults may
1309 be overridden with the -DKRETACTIVE=nnn parameter.
1311 - Translation pass 2 is significantly faster by avoiding unnecessary
1312 searching through a kernel build/module directory tree.
1314 - When compiled against elfutils 0.142 systemtap now handles the new
1315 DW_OP_call_frame_CFA generated by by GCC.
1317 - uprobes and ustack() are more robust when used on applications that
1318 depend on prelinked/separate debuginfo shared libraries.
1320 - User space PROBE marks are not always found with or without separate
1321 debuginfo. The .probes section itself is now always put in the main
1322 elf file and marked as allocated. When building pic code the section
1323 is marked writable. The selinux memory check problems seen with
1324 programs using STAP_PROBES is fixed.
1326 - statement() probes can now override "address not at start of statement"
1327 errors in guru mode. They also provide alternative addresses to use
1330 - The stapgraph application can generate graphs of data and events
1331 emitted by systemtap scripts in real time. Run "stapgraph
1332 testsuite/systemtap.examples/general/grapher.stp" for an example of
1333 graphing the system load average and keyboard events.
1335 - Dwarf probes now show parameters and local variables in the verbose
1338 - Symbol aliases are now resolved to their canonical dwarf names. For
1339 example, probing "malloc" in libc resolves to "__libc_malloc".
1341 - The syntax for dereferencing $target variables and @cast() gained new
1343 - Array indexes can now be arbitrary numeric expressions.
1344 - Array subscripts are now supported on pointer types.
1345 - An '&' operator before a @cast or $target returns the address of the
1346 final component, especially useful for nested structures.
1348 - For reading all probe variables, kernel.mark now supports $$vars and
1349 $$parms, and process.syscall now supports $$vars.
1351 - The SNMP tapset provides probes and functions for many network
1352 statistics. See stapprobes.snmp(3stap) for more details.
1354 - The dentry tapset provides functions to map kernel VFS directory entries
1355 to file or full path names: d_path(), d_name() and reverse_path_walk().
1357 - SystemTap now has userspace markers in its own binaries, and the stap
1358 tapset provides the available probepoints and local variables.
1360 - Miscellaneous new tapset functions:
1361 - pgrp() returns the process group ID of the current process
1362 - str_replace() performs string replacement
1364 * What's new in version 0.9.8, 2009-06-11
1366 - Miscellaneous new tapset functions:
1367 - sid() returns the session ID of the current process
1368 - stringat() indexes a single character from a string.
1370 - Using %M in print formats for hex dumps can now print entire buffers,
1371 instead of just small numbers.
1373 - Dwarfless syscalls: The nd_syscalls tapset is now available to probe
1374 system calls without requiring kernel debugging information. All of
1375 the same probepoints in the normal syscalls tapset are available with
1376 an "nd_" prefix, e.g. syscall.open becomes nd_syscall.open. Most
1377 syscall arguments are also available by name in nd_syscalls.
1379 - Module signing: If the appropriate nss libraries are available on your
1380 system, stap-server will sign each compiled module using a self-generated
1381 certificate. This is the first step toward extending authority to
1382 load certain modules to unprivileged users. For now, if the system
1383 administrator adds a certificate to a database of trusted signers
1384 (stap-authorize-signing-cert), modules signed using that certificate
1385 will be verified by staprun against tampering. Otherwise, you should
1386 notice no difference in the operation of stap or staprun.
1388 * What's new in version 0.9.7, 2009-04-23
1390 - @cast can now determine its type information using an explicit header
1391 specification. For example:
1392 @cast(tv, "timeval", "<sys/time.h>")->tv_sec
1393 @cast(task, "task_struct", "kernel<linux/sched.h>")->tgid
1395 - The overlapping process.* tapsets are now separated. Those probe points
1396 documented in stapprobes(3stap) remain the same. Those that were formerly
1397 in stapprobes.process(3stap) have been renamed to kprocess, to reflect
1398 their kernel perspective on processes.
1400 - The --skip-badvars option now also suppresses run-time error
1401 messages that would otherwise result from erroneous memory accesses.
1402 Such accesses can originate from $context expressions fueled by
1403 erroneous debug data, or by kernel_{long,string,...}() tapset calls.
1405 - New probes kprobe.function(FUNCTION) and kprobe.function(FUNCTION).return
1406 for dwarfless probing. These postpone function address resolution to
1407 run-time and use the kprobe symbol-resolution mechanism.
1408 Probing of absolute statements can be done using the
1409 kprobe.statement(ADDRESS).absolute construct.
1411 - EXPERIMENTAL support for user process unwinding. A new collection of
1412 tapset functions have been added to handle user space backtraces from
1413 probe points that support them (currently process and timer probes -
1414 for timer probes test whether or not in user space first with the
1415 already existing user_mode() function). The new tapset functions are:
1416 uaddr - User space address of current running task.
1417 usymname - Return the symbol of an address in the current task.
1418 usymdata - Return the symbol and module offset of an address.
1419 print_ustack - Print out stack for the current task from string.
1420 print_ubacktrace - Print stack back trace for current task.
1421 ubacktrace - Hex backtrace of current task stack.
1422 Please read http://sourceware.org/ml/systemtap/2009-q2/msg00364.html
1423 on the current restrictions and possible changes in the future and
1424 give feedback if you want to influence future developments.
1426 * What's new in version 0.9.5, 2009-03-27
1428 - New probes process().insn and process().insn.block that allows
1429 inspection of the process after each instruction or block of
1430 instructions executed. So to count the total number of instructions
1431 a process executes during a run do something like:
1432 $ stap -e 'global steps; probe process("/bin/ls").insn {steps++}
1433 probe end {printf("Total instructions: %d\n", steps);}' \
1435 This feature can slow down execution of a process somewhat.
1437 - Systemtap probes and function man pages extracted from the tapsets
1438 are now available under 3stap. To show the page for probe vm.pagefault
1439 or the stap function pexecname do:
1440 $ man 3stap vm.pagefault
1441 $ man 3stap pexecname
1443 - Kernel tracepoints are now supported for probing predefined kernel
1444 events without any debuginfo. Tracepoints incur less overhead than
1445 kprobes, and context parameters are available with full type
1446 information. Any kernel 2.6.28 and later should have defined
1447 tracepoints. Try the following to see what's available:
1448 $ stap -L 'kernel.trace("*")'
1450 - Typecasting with @cast now supports modules search paths, which is
1451 useful in case there are multiple places where the type definition
1452 may be found. For example:
1453 @cast(sdev, "scsi_device", "kernel:scsi_mod")->sdev_state
1455 - On-file flight recorder is supported. It allows stap to record huge
1456 trace log on the disk and to run in background.
1457 Passing -F option with -o option runs stap in background mode. In this
1458 mode, staprun is detached from console, and stap itself shows staprun's
1460 Specifying the max size and the max number of log files are also available
1461 by passing -S option. This option has one or two arguments seperated by
1462 a comma. The first argument is the max size of a log file in MB. If the
1463 size of a log file exceeds it, stap switches to the next log file
1464 automatically. The second is how many files are kept on the disk. If the
1465 number of log files exceeds it, the oldest log file is removed
1466 automatically. The second argument can be omitted.
1468 For example, this will record output on log files each of them is smaller
1469 than 1024MB and keep last 3 logs, in background.
1470 % stap -F -o /tmp/staplog -S 1024,3 script.stp
1472 - In guru mode (-g), the kernel probing blacklist is disabled, leaving
1473 only a subset - the kernel's own internal kprobe blacklist - to attempt
1474 to filter out areas unsafe to probe. The differences may be enough to
1475 probe more interrupt handlers.
1477 - Variables unavailable in current context may be skipped by setting a
1478 session level flag with command line option --skip-badvars now available.
1479 This replaces any dwarf $variable expressions that could not be resolved
1480 with literal numeric zeros, along with a warning message.
1482 - Both kernel markers and kernel tracepoint support argument listing
1483 through stap -L 'kernel.mark("*")' or stap -L 'kernel.trace("*")'
1485 - Users can use -DINTERRUPTIBLE=0 to prevent interrupt reentrancy in
1486 their script, at the cost of a bit more overhead to toggle the
1489 - Added reentrancy debugging. If stap is run with the arguments
1490 "-t -DDEBUG_REENTRANCY", additional warnings will be printed for
1491 every reentrancy event, including the probe points of the
1492 resident and interloper probes.
1494 - Default to --disable-pie for configure.
1495 Use --enable-pie to turn it back on.
1497 - Improved sdt.h compatibility and test suite for static dtrace
1498 compatible user space markers.
1500 - Some architectures now use syscall wrappers (HAVE_SYSCALL_WRAPPERS).
1501 The syscall tapset has been enhanced to take care of the syscall
1502 wrappers in this release.
1504 - Security fix for CVE-2009-0784: stapusr module-path checking race.
1506 * What's new in version 0.9, 2009-02-19
1508 - Typecasting is now supported using the @cast operator. A script can
1509 define a pointer type for a "long" value, and then access type members
1510 using the same syntax as with $target variables. For example, this will
1511 retrieve the parent pid from a kernel task_struct:
1512 @cast(pointer, "task_struct", "kernel")->parent->pid
1514 - process().mark() probes are now possible to trace static user space
1515 markers put in programs with the STAP_PROBE macro using the new
1516 sys/sdt.h include file. This also provides dtrace compatible markers
1517 through DTRACE_PROBE and an associated python 'dtrace' script that
1518 can be used in builds based on dtrace that need dtrace -h or -G
1521 - For those that really want to run stap from the build tree there is
1522 now the 'run-stap' script in the top-level build directory that sets
1523 up the SYSTEMTAP_TAPSET, SYSTEMTAP_RUNTIME, SYSTEMTAP_STAPRUN, and
1524 SYSTEMTAP_STAPIO environment variables (installing systemtap, in a
1525 local prefix, is still recommended for common use).
1527 - Systemtap now comes with a new Beginners Guide that walks the user
1528 through their first steps setting up stap, understanding how it all
1529 works, introduces some useful scripts and describes some common
1530 pitfalls. It isn't created by default since it needs a Publican
1531 setup, but full build instructions can be found in the wiki:
1532 http://sourceware.org/systemtap/wiki/PublicanQuikHowto
1533 An online version can be found at:
1534 http://sourceware.org/systemtap/SystemTap_Beginners_Guide.pdf
1536 - Standard tapsets included with Systemtap were modified to include
1537 extractable documentation information based on the kernel-doc
1538 infrastructure. When configured --enabled-docs a HTML and PDF
1539 version of the Tapset Reference Manual is produced explaining probes
1540 defined in each tapset.
1542 - The systemtap client and compile server are now available.
1543 These allow you to compile a systemtap module on a host other than
1544 the one which it will be run, providing the client and server
1545 are compatible. Other than using a server for passes 1 through
1546 4, the client behaves like the 'stap' front end itself. This
1547 means, among other things, that the client will automatically
1548 load the resulting module on the local host unless -p[1234]
1549 was specified. See stap-server(8) for more details.
1550 The client/server now use SSL for network connection security and
1553 The systemtap client and server are prototypes only. Interfaces, options
1554 and usage may change at any time.
1556 - function("func").label("label") probes are now supported to allow matching
1557 the label of a function.
1559 - Systemtap initscript is available. This initscript allows you to run
1560 systemtap scripts as system services (in flight recorder mode) and
1561 control those scripts individually.
1562 See README.systemtap for details.
1564 - The stap "-r DIR" option may be used to identify a hand-made kernel
1565 build directory. The tool determines the appropriate release string
1566 automatically from the directory.
1568 - Serious problems associated with user-space probing in shared libraries
1569 were corrected, making it now possible to experiment with probe shared
1570 libraries. Assuming dwarf debugging information is installed, use this
1571 twist on the normal syntax:
1573 probe process("/lib64/libc-2.8.so").function("....") { ... }
1575 This would probe all threads that call into that library. Running
1576 "stap -c CMD" or "stap -x PID" naturally restricts this to the target
1577 command+descendants only. $$vars etc. may be used.
1579 - For scripts that sometimes terminate with excessive "skipped" probes,
1580 rerunning the script with "-t" (timing) will print more details about
1581 the skippage reasons.
1583 - Symbol tables and unwind (backtracing) data support were formerly
1584 compiled in for all probed modules as identified by the script
1585 (kernel; module("name"); process("file")) plus those listed by the
1586 stap "-d BINARY" option. Now, this data is included only if the systemtap
1587 script uses tapset functions like probefunc() or backtrace() that require
1588 such information. This shrinks the probe modules considerably for the rest.
1590 - Per-pass verbosity control is available with the new "--vp {N}+" option.
1591 "stap --vp 040" adds 4 units of -v verbosity only to pass 2. This is useful
1592 for diagnosing errors from one pass without excessive verbosity from others.
1594 - Most probe handlers now run with interrupts enabled, for improved
1595 system responsiveness and less probing overhead. This may result
1596 in more skipped probes, for example if a reentrant probe handler
1597 is attempted from within an interrupt handler. It may also make the
1598 systemtap overload detection facility more likely to be triggered, as
1599 interrupt handlers' run time would be included in the self-assessed
1600 overhead of running probe handlers.
1602 * What's new in version 0.8, 2008-11-13
1604 - Cache limiting is now available. If the compiled module cache size is
1605 over a limit specified in the $SYSTEMTAP_DIR/cache/cache_mb_limit file,
1606 some old cache entries will be unlinked. See man stap(1) for more.
1608 - Error and warning messages are now followed by source context displaying
1609 the erroneous line/s and a handy '^' in the following line pointing to the
1612 - A bug reporting tool "stap-report" is now available which will quickly
1613 retrieve much of the information requested here:
1614 http://sourceware.org/systemtap/wiki/HowToReportBugs
1616 - The translator can resolve members of anonymous structs / unions:
1617 given struct { int foo; struct { int bar; }; } *p;
1618 this now works: $p->bar
1620 - The stap "-F" flag activates "flight recorder" mode, which consists of
1621 translating the given script as usual, but implicitly launching it into
1622 the background with staprun's existing "-L" (launch) option. A user
1623 can later reattach to the module with "staprun -A MODULENAME".
1625 - Additional context variables are available on user-space syscall probes.
1626 - $argN ($arg1, $arg2, ... $arg6) in process(PATH_OR_PID).syscall
1627 gives you the argument of the system call.
1628 - $return in process(PATH_OR_PID).syscall.return gives you the return
1629 value of the system call.
1631 - Target process mode (stap -c CMD or -x PID) now implicitly restricts all
1632 "process.*" probes to the given child process. (It does not affect
1633 kernel.* or other probe types.) The CMD string is normally run directly,
1634 rather than via a /bin/sh -c subshell, since then utrace/uprobe probes
1635 receive a fairly "clean" event stream. If metacharacters like
1636 redirection operators were present in CMD, then "sh -c CMD" is still
1637 used, and utrace/uprobe probes will receive events from the shell.
1639 % stap -e 'probe process.syscall, process.end {
1640 printf("%s %d %s\n", execname(), pid(), pp())}'\
1642 ls 2323 process.syscall
1643 ls 2323 process.syscall
1646 - Probe listing mode is improved: "-L" lists available script-level variables
1648 % stap -L 'syscall.*open*'
1649 syscall.mq_open name:string name_uaddr:long filename:string mode:long u_attr_uaddr:long oflag:long argstr:string
1650 syscall.open name:string filename:string flags:long mode:long argstr:string
1651 syscall.openat name:string filename:string flags:long mode:long argstr:string
1653 - All user-space-related probes support $PATH-resolved executable
1656 probe process("ls").syscall {}
1657 probe process("./a.out").syscall {}
1659 work now, instead of just
1661 probe process("/bin/ls").syscall {}
1662 probe process("/my/directory/a.out").syscall {}
1664 - Prototype symbolic user-space probing support:
1666 # stap -e 'probe process("ls").function("*").call {
1667 log (probefunc()." ".$$parms)
1672 - debugging information for the named program
1673 - a version of utrace in the kernel that is compatible with the "uprobes"
1674 kernel module prototype. This includes RHEL5 and older Fedora, but not
1675 yet current lkml-track utrace; a "pass 4a"-time build failure means
1676 your system cannot use this yet.
1678 - Global variables which are written to but never read are now
1679 automatically displayed when the session does a shutdown. For example:
1681 global running_tasks
1682 probe timer.profile {running_tasks[pid(),tid()] = execname()}
1683 probe timer.ms(8000) {exit()}
1685 - A formatted string representation of the variables, parameters, or local
1686 variables at a probe point is now supported via the special $$vars,
1687 $$parms, and $$locals context variables, which expand to a string
1688 containing a list "var1=0xdead var2=0xbeef var3=?". (Here, var3 exists
1689 but is for some reason unavailable.) In return probes only, $$return
1690 expands to an empty string for a void function, or "return=0xf00".
1693 * What's new in version 0.7, 2008-07-15
1695 - .statement("func@file:*") and .statement("func@file:M-N") probes are now
1696 supported to allow matching a range of lines in a function. This allows
1697 tracing the execution of a function.
1699 - Scripts relying on probe point wildcards like "syscall.*" that expand
1700 to distinct kprobes are processed significantly faster than before.
1702 - The vector of script command line arguments is available in a
1703 tapset-provided global array argv[]. It is indexed 1 ... argc,
1704 another global. This can substitute for of preprocessor
1705 directives @NNN that fail at parse time if there are not
1708 printf("argv: %s %s %s", argv[1], argv[2], argv[3])
1710 - .statement("func@file+line") probes are now supported to allow a
1711 match relative to the entry of the function incremented by line
1712 number. This allows using the same systemtap script if the rest
1713 of the file.c source only changes slightly.
1715 - A probe listing mode is available.
1724 - More user-space probe types are added:
1726 probe process(PID).begin { }
1727 probe process("PATH").begin { }
1728 probe process(PID).thread.begin { }
1729 probe process("PATH").thread.begin { }
1730 probe process(PID).end { }
1731 probe process("PATH").end { }
1732 probe process(PID).thread.end { }
1733 probe process("PATH").thread.end { }
1734 probe process(PID).syscall { }
1735 probe process("PATH").syscall { }
1736 probe process(PID).syscall.return { }
1737 probe process("PATH").syscall.return { }
1739 - Globals now accept ; terminators
1742 global little[10], big[5];
1744 * What's new in version 0.6, 2007-12-15
1746 - A copy of the systemtap tutorial and language reference guide
1749 - There is a new format specifier, %m, for the printf family of
1750 functions. It functions like %s, except that it does not stop when
1751 a nul ('\0') byte is encountered. The number of bytes output is
1752 determined by the precision specifier. The default precision is 1.
1755 printf ("%m", "My String") // prints one character: M
1756 printf ("%.5", myString) // prints 5 bytes beginning at the start
1759 - The %b format specifier for the printf family of functions has been enhanced
1762 1) When the width and precision are both unspecified, the default is %8.8b.
1763 2) When only one of the width or precision is specified, the other defaults
1764 to the same value. For example, %4b == %.4b == %4.4b
1765 3) Nul ('\0') bytes are used for field width padding. For example,
1767 printf ("%b", 0x1111deadbeef2222) // prints all eight bytes
1768 printf ("%4.2b", 0xdeadbeef) // prints \0\0\xbe\xef
1770 - Dynamic width and precision are now supported for all printf family format
1771 specifiers. For example:
1775 printf ("%*.*b", four, two, 0xdeadbbeef) // prints \0\0\xbe\xef
1776 printf ("%*d", four, two) // prints <space><space><space>2
1778 - Preprocessor conditional expressions can now include wildcard style
1779 matches on kernel versions.
1780 %( kernel_vr != "*xen" %? foo %: bar %)
1782 - Prototype support for user-space probing is showing some progress.
1783 No symbolic notations are supported yet (so no probing by function names,
1784 file names, process names, and no access to $context variables), but at
1785 least it's something:
1787 probe process(PID).statement(ADDRESS).absolute { }
1789 This will set a uprobe on the given process-id and given virtual address.
1790 The proble handler runs in kernel-space as usual, and can generally use
1791 existing tapset functions.
1793 - Crash utility can retrieve systemtap's relay buffer from a kernel dump
1794 image by using staplog which is a crash extension module. To use this
1795 feature, type commands as below from crash(8)'s command line:
1797 crash> extend staplog.so
1798 crash> help systemtaplog
1800 Then, you can see more precise help message.
1802 - You can share a relay buffer amoung several scripts and merge outputs from
1803 several scripts by using "-DRELAY_HOST" and "-DRELAY_GUEST" options.
1807 % stap -ve 'probe begin{}' -o merged.out -DRELAY_HOST &
1808 # wait until starting the host.
1809 % stap -ve 'probe begin{print("hello ");exit()}' -DRELAY_GUEST
1810 % stap -ve 'probe begin{print("world\n");exit()}' -DRELAY_GUEST
1812 Then, you'll see "hello world" in merged.out.
1814 - You can add a conditional statement for each probe point or aliase, which
1815 is evaluated when the probe point is hit. If the condition is false, the
1816 whole probe body(including aliases) is skipped. For example:
1819 probe syscall.* if (switch) { ... }
1820 probe procfs.write {switch = strtol($value,10)} /* enable/disable ctrl */
1822 - Systemtap will warn you if your script contains unused variables or
1823 functions. This is helpful in case of misspelled variables. If it
1824 doth protest too much, turn it off with "stap -w ...".
1826 - You can add error-handling probes to a script, which are run if a
1827 script was stopped due to errors. In such a case, "end" probes are
1828 not run, but "error" ones are.
1830 probe error { println ("oops, errors encountered; here's a report anyway")
1831 foreach (coin in mint) { println (coin) } }
1833 - In a related twist, one may list probe points in order of preference,
1834 and mark any of them as "sufficient" beyond just "optional". Probe
1835 point sequence expansion stops if a sufficient-marked probe point has a hit.
1836 This is useful for probes on functions that may be in a module (CONFIG_FOO=m)
1837 or may have been compiled into the kernel (CONFIG_FOO=y), but we don't know
1840 probe module("sd").function("sd_init_command") ? ,
1841 kernel.function("sd_init_command") ? { ... }
1843 which might match neither, now one can write this:
1845 probe module("sd").function("sd_init_command") ! , /* <-- note excl. mark */
1846 kernel.function("sd_init_command") { ... }
1848 - New security model. To install a systemtap kernel module, a user
1849 must be one of the following: the root user; a member of the
1850 'stapdev' group; or a member of the 'stapusr' group. Members of the
1851 stapusr group can only use modules located in the
1852 /lib/modules/VERSION/systemtap directory (where VERSION is the
1853 output of "uname -r").
1855 - .statement("...@file:line") probes now apply heuristics to allow an
1856 approximate match for the line number. This works similarly to gdb,
1857 where a breakpoint placed on an empty source line is automatically
1858 moved to the next statement. A silly bug that made many $target
1859 variables inaccessible to .statement() probes was also fixed.
1861 - LKET has been retired. Please let us know on <systemtap@sourceware.org>
1862 if you have been a user of the tapset/tools, so we can help you find
1865 - New families of printing functions println() and printd() have been added.
1866 println() is like print() but adds a newline at the end;
1867 printd() is like a sequence of print()s, with a specified field delimiter.
1869 * What's new since version 0.5.14?, 2007-07-03
1871 - The way in which command line arguments for scripts are substituted has
1872 changed. Previously, $1 etc. would interpret the corresponding command
1873 line argument as an numeric literal, and @1 as a string literal. Now,
1874 the command line arguments are pasted uninterpreted wherever $1 etc.
1875 appears at the beginning of a token. @1 is similar, but is quoted as
1876 a string. This change does not modify old scripts, but has the effect
1877 of permitting substitution of arbitrary token sequences.
1879 # This worked before, and still does:
1880 % stap -e 'probe timer.s($1) {}' 5
1881 # Now this also works:
1882 % stap -e 'probe syscall.$1 {log(@1)}' open
1883 # This won't crash, just signal a recursion error:
1885 # As before, $1... is recognized only at the beginning of a token
1886 % stap -e 'probe begin {foo$1=5}'
1888 * What's new since version 0.5.13?, 2007-03-26
1890 - The way in which systemtap resolves function/inline probes has changed:
1891 .function(...) - now refers to all functions, inlined or not
1892 .inline(...) - is deprecated, use instead:
1893 .function(...).inline - filters function() to only inlined instances
1894 .function(...).call - filters function() to only non-inlined instances
1895 .function(...).return - as before, but now pairs best with .function().call
1896 .statement() is unchanged.
1898 * What's new since version 0.5.12?, 2007-01-01
1900 - When running in -p4 (compile-only) mode, the compiled .ko file name
1901 is printed on standard output.
1903 - An array element with a null value such as zero or an empty string
1904 is now preserved, and will show up in a "foreach" loop or "in" test.
1905 To delete such an element, the scripts needs to use an explicit
1906 "delete array[idx]" statement rather than something like "array[idx]=0".
1908 - The new "-P" option controls whether prologue searching heuristics
1909 will be activated for function probes. This was needed to get correct
1910 debugging information (dwarf location list) data for $target variables.
1911 Modern compilers (gcc 4.1+) tend not to need this heuristic, so it is
1912 no longer default. A new configure flag (--enable-prologues) restores
1913 it as a default setting, and is appropriate for older compilers (gcc 3.*).
1915 - Each systemtap module prints a one-line message to the kernel informational
1916 log when it starts. This line identifies the translator version, base
1917 address of the probe module, a broken-down memory consumption estimate, and
1918 the total number of probes. This is meant as a debugging / auditing aid.
1920 - Begin/end probes are run with interrupts enabled (but with
1921 preemption disabled). This will allow begin/end probes to be
1922 longer, to support generating longer reports.
1924 - The numeric forms of kernel.statement() and kernel.function() probe points
1925 are now interpreted as relocatable values - treated as relative to the
1926 _stext symbol in that kernel binary. Since some modern kernel images
1927 are relocated to a different virtual address at startup, such addresses
1928 may shift up or down when actually inserted into a running kernel.
1930 kernel.statement(0xdeadbeef): validated, interpreted relative to _stext,
1931 may map to 0xceadbeef at run time.
1933 In order to specify unrelocated addresses, use the new ".absolute"
1934 probe point suffix for such numeric addresses. These are only
1935 allowed in guru mode, and provide access to no $target variables.
1936 They don't use debugging information at all, actually.
1938 kernel.statement(0xfeedface).absolute: raw, unvalidated, guru mode only
1940 * What's new since version 0.5.10?, 2006-10-19
1942 - Offline processing of debugging information, enabling general
1943 cross-compilation of probe scripts to remote hosts, without
1944 requiring identical module/memory layout. This slows down
1945 compilation/translation somewhat.
1947 - Kernel symbol table data is loaded by staprun at startup time
1948 rather than compiled into the module.
1950 - Support the "limit" keyword for foreach iterations:
1951 foreach ([x,y] in ary limit 5) { ... }
1952 This implicitly exits after the fifth iteration. It also enables
1953 more efficient key/value sorting.
1955 - Support the "maxactive" keyword for return probes:
1956 probe kernel.function("sdfsdf").maxactive(848) { ... }
1957 This allows up to 848 concurrently outstanding entries to
1958 the sdfsdf function before one returns. The default maxactive
1959 number is smaller, and can result in missed return probes.
1961 - Support accessing of saved function arguments from within
1962 return probes. These values are saved by a synthesized
1963 function-entry probe.
1965 - Add substantial version/architecture checking in compiled probes to
1966 assert correct installation of debugging information and correct
1967 execution on a compatible kernel.
1969 - Add probe-time checking for sufficient free stack space when probe
1970 handlers are invoked, as a safety improvement.
1972 - Add an optional numeric parameter for begin/end probe specifications,
1973 to order their execution.
1974 probe begin(10) { } /* comes after */ probe begin(-10) {}
1976 - Add an optional array size declaration, which is handy for very small
1978 global little[5], big[20000]
1980 - Include some example scripts along with the documentation.
1982 - Change the start-time allocation of probe memory to avoid causing OOM
1983 situations, and to abort cleanly if free kernel memory is short.
1985 - Automatically use the kernel DWARF unwinder, if present, for stack
1988 - Many minor bug fixes, performance, tapset, and error message