1 * What's new in version 2.9
3 - Callee probe points now support '.return' and '.call' suffix.
6 process("proc").function("foo").callee("bar").return
8 will fire upon returning from bar when called by foo.
10 process("proc").function("foo").callee("bar").call
12 will only fire for non-inlined callees.
14 - The following tapset variables and functions are deprecated in
16 - The '__int32_compat' library macro got deprecated in favor of
17 new '__compat_long' library macro.
18 - The 'uargs' convenience variable of the 'seccomp' syscall probe
19 got deprecated in favor of new 'uargs_uaddr' variable.
21 * What's new in version 2.8, 2015-06-17
23 - SystemTap has improved support for probing golang programs. Work has been
24 done to be able to handle DWARF information, reporting file names, line
25 numbers, and column numbers, and tolerance of odd characters in symbol names.
27 - The function::*, probe::* and new macro::* man pages cross-references the
28 enclosing tapset::* man page. For example:
30 function::pn(3stap) mentions tapset::pn(3stap) in the SEE ALSO section
32 - New stapref(1) man page provides a reference for the scripting language. The
33 stapref page contains an overview of the features available in the language,
34 such as keywords, data types, operators and more.
36 - The @task macro performs the very common @cast to a task_struct.
38 The embedded-C bodies of task_current() and pid2task() are now wrapped
39 by @task, which gives them a debuginfo type on the return value. With
40 autocast type propagation, this removes the need for any explicit @cast
43 Other places which take untyped task pointers as parameters, for
44 instance, now use @task as well to simplify their code.
46 - New namespace-aware tapset functions [task_]ns_*() and ia new option
47 --target-namespaces=PID to denote a target set of namespaces corresponding to
48 the PID's namespaces. The namespace-aware tapsets will return values
49 relative to the target namespaces if specified, or the stap process' namespaces.
51 - Netfilter probes now attempt to decode Spanning Tree Protocol packets
52 into local variables: probe netfilter.bridge.*, br_* variables,
53 stp_dump.stp sample script.
55 - Colorization of error string tokens is made more robust, especially
56 in presence of $N/@N substitution.
58 - The following tapset variables and functions are deprecated in
60 - The 'hostname_uaddr' variable in the syscall.setdomainname and
61 nd_syscall.setdomainname probe aliases have been deprecated in
62 favor of the new 'domainname_uaddr' variable.
63 - The 'fd' and 'fd_str' variables in the syscall.execveat and
64 nd_syscall.execveat probe aliases have been deprecated in favor of
65 the new 'dirfd' and 'dirfd_str' variables.
67 * What's new in version 2.7, 2015-02-18
69 - Some systemtap sample scripts are now identified with the "_best" keyword,
70 because they are generally useful or educational. They are now promoted
71 within the generated index files.
73 - Passing strings to and from functions has become faster due to optimization
74 (passing some strings by reference instead of by value/copy). It may
75 be disabled by using the unoptimize flag (-u).
77 To make embedded-C functions eligible for the same optimization, use the pragma
78 /* unmodified-fnargs */ to indicate that the function body will not modify
79 the function arguments. Remember to use MAXSTRINGLEN for string length,
80 rather than sizeof(string_arg) (which might now be a pointer).
82 - SystemTap now allows .function probes to be specified by their full function
83 name, file, and declaration line number. Use the .statement probe to probe a
86 - Tracepoint probes can now also be specified by the target subsystem. For
87 example, the following are all supported:
89 probe kernel.trace("sched:sched_switch") --> probe sched_switch found in the
91 probe kernel.trace("sched:*") --> probe all tracepoints in sched subsystem
93 As a result, tapset functions such as pn() will now return a different string
94 than before. To retain the previous behaviour, use '--compatible=2.6'.
96 - The following functions are deprecated in release 2.7:
97 - _adjtx_mode_str(), _statfs_f_type_str(), _waitid_opt_str(),
98 _internal_wait_opt_str(), and _epoll_events_str().
100 - New tapset functions [u]symfileline(), [u]symfile() and [u]symline() will
101 return a string containing the specified portion of the filename:linenumber
102 match from a given address.
104 Using these functions may result in large generated modules from stored
105 address->file:line information.
107 * What's new in version 2.6, 2014-09-05
109 - SystemTap now supports on-the-fly arming/disarming of certain probe types:
110 kprobes, uprobes, and timer.*s(NUM) probes. For example, this probe
112 probe kernel.function("vfs_read") if (i > 4) { ... }
114 will automatically register/unregister the associated kprobe on vfs_read
115 whenever the value of the condition changes (as some probe handler
116 modifies 'i'). This allows us to avoid probe overhead when we're not
117 interested. If the arming capability is not relevant/useful, nest the
118 condition in the normal probe handler:
120 probe kernel.function("vfs_read") { if (i > 4) { ... } }
122 - statement("*@file:NNN").nearest probes now available to let systemtap
123 translate probe to nearest probe-able line to one given if necessary
125 - process("PATH").library("PATH").plt("NAME").return probes are now supported.
127 - SystemTap now supports SDT probes with operands that refer to symbols.
129 - While in listing mode (-l/-L), probes printed are now more consistent
132 - Statement probes now support enumerated linenos to probe discontiguous
133 linenos using the form:
135 process.statement("foo@file.c:3,5-7,9")
137 - Statement counting is now suppressed in the generated c code for probes that
138 are non-recursive and loop-free. Statement counting can be turned back on in
139 unoptimize mode (-u).
141 - SystemTap now asserts that the PID provided for a process probe corresponds
142 to a running process.
144 - DWARF process probes can be bound to a specific process using the form:
146 process(PID).function("*")
148 - SystemTap now accepts additional scripts through the new -E SCRIPT option.
149 There still needs to be a main script specified through -e or file in order
150 to provide an additional script. This makes it feasible to have scripts in
151 the $HOME/.systemtap/rc file. For example:
153 -E 'probe begin, end, error { log("systemtap script " . pn()) }'
154 -E 'probe timer.s(30) { error ("timeout") }
156 The -E SCRIPT option can also be used in listing mode (-l/-L), such that
157 probe points for the additional scripts will not listed, but other parts of
158 the script are still available, such as macros or aliases.
160 - SystemTap now supports array slicing within foreach loop conditions, delete
161 statements and membership tests. Wildcards are represented by "*". Examples
162 of the expressions are:
164 foreach ([a,b,c] in val[*,2,*])
168 - Integer expressions which are derived from DWARF values, like context $vars,
169 @cast, and @var, will now carry that type information into subsequent reads.
170 Such expressions can now use "->" and "[]" operators, as can local variables
171 which were assigned such values.
173 foo = $param->foo; printf("x:%d y:%d\n", foo->x, foo->y)
174 printf("my value is %d\n", ($type == 42 ? $foo : $bar)->value)
175 printf("my parent pid is %d\n", task_parent(task_current())->tgid)
177 * What's new in version 2.5, 2014-04-30
179 - Systemtap now supports backtracing through its own, invoking module.
181 - Java probes now support backtracing using the print_java_backtrace()
182 and sprint_java_backtrace() functions.
184 - Statement probes (e.g. process.statement) are now faster to resolve,
185 more precise, and work better with inlined functions.
187 - New switches have been added to help inspect the contents of installed
190 stap --dump-functions --> list all library functions and their args
191 stap --dump-probe-aliases --> list all library probe aliases
193 - The heuristic algorithms used to search for function-prologue
194 endings were improved, to cover more optimization (or
195 lack-of-optimization, or incorrect-debuginfo) cases. These
196 heuristics are necessary to find $context parameters for some
197 function-call/entry probes. We recommend programs be built with
198 CFLAGS+=-grecord-gcc-switches to feed information to the heuristics.
200 - The stap --use-server option now more correctly supports address:port
201 type parametrization, for manual use in the absence of avahi.
203 - A new probe alias "oneshot" allows a single quick script fragment to run,
206 - The argv tapset now merges translate-time and run-time positional
207 arguments, so all of these work:
209 stap -e 'probe oneshot {println(argv[1]," ",argv[2])}' hello world
211 stap -e 'probe oneshot {println(argv[1]," ",argv[2])}' \
212 -G argv_1=hello -G argv_2=world
214 staprun hello.ko argv_1=hello argv_2=world
216 - SystemTap now falls back on the symbol table for probing
217 functions in processes if the debuginfo is not available.
219 - SystemTap now supports a %( guru_mode == 0 /* or 1 */ %)
220 conditional for making dual-use scripts.
222 - SystemTap now supports UEFI/SecureBoot systems, via
223 machine-owner-keys maintained by a trusted stap-server on the
224 network. (Key enrollment requires a one-time reboot and BIOS
226 https://sourceware.org/systemtap/wiki/SecureBoot
228 - SystemTap now reports more accurate and succinct errors on type
231 - Embedded-C functions may use STAP_RETURN(value) instead of the
232 more wordy STAP_RETVALUE assignment followed by a "goto out".
233 The macro supports numeric or string values as appropriate.
234 STAP_ERROR(...) is available to return with a (catchable) error.
236 - Some struct-sockaddr fields are now individually decoded for
237 socket-related syscalls:
238 probe syscall.connect { println (uaddr_af, ":", uaddr_ip) }
240 - The documentation for the SystemTap initscript service and the
241 SystemTap compile-server service have been completely converted from
242 README files to man pages (see systemtap(8) and stap-server(8)).
244 - SystemTap is now capable of inserting modules early during the boot
245 process on dracut-based systems. See the 'onboot' command in
246 systemtap(8) for more information.
248 - DWARF probes can now use the '.callee[s]' variants, which allow more
249 precise function probing. For example, the probe point
251 process("myproc").function("foo").callee("bar")
253 will fire upon entering bar() from foo(). A '.callees' probe will
254 instead place probes on all callees of foo().
255 Note that this also means that probe point wildcards should be used
256 with more care. For example, use signal.*.return rather than
257 signal.*.*, which would also match '.callees'. See stapprobes(3stap)
258 for more info. This feature requires at least GCC 4.7.
260 - A few new functions in the task_time tapsets, as well as a new tapset
261 function task_ancestry(), which prints out the parentage of a process.
263 - The kprocess.exec probe has been updated to use syscall.execve, which
264 allows access to the new process' arguments (through the new 'argstr'
265 or 'args' variables) as well as giving better support across kernel
266 versions. Note also that the 'filename' variable now holds the
267 filename (quoted), or the address (unquoted) if it couldn't be
270 - The [s]println() function can now be called without any arguments to
271 simply print a newline.
273 - Suggestions are now provided when markers could not be resolved. For
274 example, process("stap").mark("benchmart") will suggest 'benchmark'.
276 - SystemTap colors can now be turned off by simply setting
277 SYSTEMTAP_COLORS to be empty, rather than having to make it invalid.
279 - There is a new context tapset function, pnlabel(), which returns the
280 name of the label which fired.
282 - The following tapset variables and functions are deprecated in
284 - The 'clone_flags', 'stack_start', 'stack_size',
285 'parent_tid_uaddr', and 'child_tid_uaddr' variables in the
286 'syscall.fork' and 'nd_syscall.fork' probe aliases.
287 - The '_sendflags_str()' and '_recvflags_str()' functions have been
288 deprecated in favor of the new '_msg_flags_str()' function.
289 - The 'flags' and 'flags_str' variables in the 'syscall.accept' and
290 'nd_syscall.accept' probe alias.
291 - The 'first', 'second', and 'uptr_uaddr' variables in the
292 'syscall.compat_sys_shmctl', and 'nd_syscall.compat_sys_shmctl'
293 probe aliases have been deprecated in favor of the new 'shmid',
294 'cmd', and 'buf_uaddr' variables.
296 * What's new in version 2.4, 2013-11-06
298 - Better suggestions are given in many of the semantic errors in which
299 alternatives are provided. Additionally, suggestions are now provided
300 when plt and trace probes could not be resolved. For example,
301 kernel.trace("sched_siwtch") will suggest 'sched_switch'.
303 - SystemTap is now smarter about error reporting. Errors from the same
304 source are considered duplicates and suppressed. A message is
305 displayed on exit if any errors/warnings were suppressed.
307 - Statistics aggregate typed objects are now implemented locklessly,
308 if the translator finds that they are only ever read (using the
309 foreach / @count / etc. constructs) in a probe-begin/end/error.
311 - SystemTap now supports probing inside virtual machines using the
312 libvirt and unix schemes, e.g.
314 stap -ve 'probe timer.s(1) { printf("hello!\n") }' \
315 --remote=libvirt://MyVirtualMachine
317 Virtual machines managed by libvirt can be prepared using stapvirt.
318 See stapvirt(1) and the --remote option in stap(1) for more details.
320 - Systemtap now checks for and uses (when available) the .gnu_debugdata
321 section which contains a subset of debuginfo, useful for backtraces
324 - SystemTap map variables are now allocated with vmalloc() instead of
325 with kmalloc(), which should cause memory to be less fragmented.
327 - Although SystemTap itself requires elfutils 0.148+, staprun only
328 requires elfutils 0.142+, which could be useful with the
329 '--disable-translator' configure switch.
331 - Under FIPS mode (/proc/sys/crypto/fips_enabled=1), staprun will
332 refuse to load systemtap modules (since these are not normally
333 signed with the kernel's build-time keys). This protection may
334 be suppressed with the $STAP_FIPS_OVERRIDE environment variable.
336 - The stap-server client & server code now enable all SSL/TLS
337 ciphers rather than just the "export" subset.
339 - For systems with in-kernel utrace, 'process.end' and 'thread.end'
340 probes will hit before the target's parent process is notified of
341 the target's death. This matches the behavior of newer kernels
342 without in-kernel utrace.
344 * What's new in version 2.3, 2013-07-25
346 - More context-accessing functions throw systemtap exceptions upon a
347 failure, whereas in previous versions they might return non-error
348 sentinel values like "" or "<unknown>". Use try { } / catch { }
349 around these, or new wrapper functions such as user_string_{n_}quoted()
350 that internally absorb exceptions.
352 - java("org.my.MyApp") probes are now restricted to pre-existing jvm pid's with
353 a listing in jps -l output to avoid recursive calls
355 - The tapset [nd_]syscall.semop parameter tsops_uaddr is renamed sops_uaddr for
356 consistency with [nd_]syscall.semtimedop.
358 - The udp.stp tapset adds some ip-address/port variables.
360 - A new guru-mode-only tapset function raise() is available to send signals
363 - Support for the standard Posix ERE named character classes has been
364 added to the regexp engine, e.g. [:digit:], [:alpha:], ...
366 - A substantial internal overhaul of the regexp engine has resulted in
367 correct behaviour on further obscure edge cases. The regexp engine
368 now implements the ERE standard and correctly passes the testsuite
369 for the glibc regexp engine (minus portions corresponding to
370 unimplemented features -- i.e. subexpression capture and reuse).
372 - Alternative functions are now suggested when function probes could not be
373 resolved. For example, kernel.function("vfs_reads") will suggest vfs_read.
374 Other probes for which suggestions are made are module.function,
375 process.function, and process.library.function.
377 - Has life been a bit bland lately? Want to spice things up? Why not write a
378 few faulty probes and feast your eyes upon the myriad of colours adorning
379 your terminal as SystemTap softly whispers in your ear... 'parse error'.
380 Search for '--color' in 'man stap' for more info.
382 - The following tapset functions are deprecated in release 2.3:
383 'stap_NFS_CLIENT', '__getfh_inode', '_success_check',
384 '_sock_prot_num', '_sock_fam_num', '_sock_state_num',
385 '_sock_type_num', and '_sock_flags_num'.
387 * What's new in version 2.2.1, 2013-05-16
388 * What's new in version 2.2, 2013-05-14
390 - Experimental support has been added for probing Java methods using
391 Byteman 2.0 as a backend. Java method probes can target method entries,
392 returns, or specific statements in the method as specified by line number.
394 probe java("org.my.MyApp").class("^java.lang.Object").method("foo(int)")
397 See java/README for information on how to set up Java/Byteman
398 functionality. Set env STAPBM_VERBOSE=yes for more tracing.
400 - The stap -l output and pn() tapset function's return value may be slightly
401 different for complicated web of wildcarded/aliased probes.
403 - The dyninst backend has improved in several aspects:
405 - Setting custom values for global variables is now supported, both
406 with -G when compiling a script, and from the stapdyn command line
407 when loading a precompiled module.
409 - A high-performance shared-memory-based transport is used for
412 - A systemd service file and tmpfile have been added to allow
413 systemtap-server to be managed natively by systemd.
415 - Due to the removal of register_timer_hook in recent kernels, the
416 behaviour of timer.profile has been changed slightly. This probe is
417 now an alias which uses the old mechanism where possible, but falls
418 back to perf.sw.cpu_clock or another mechanism when the kernel timer
419 hook is not available.
421 To require the kernel timer hook mechanism in your script, use
422 timer.profile.tick instead of timer.profile.
424 - The following tapset variables are deprecated in release 2.2:
425 - The 'origin' variables in the 'generic.fop.llseek',
426 'generic.fop.llseek.return', and 'nfs.fop.llseek' probes. The
427 'origin' variable has been replaced by the 'whence' variable.
428 - The 'page_index' variable in the 'vfs.block_sync_page' and
429 'vfs.buffer_migrate_page' probe aliases.
430 - The 'write_from' and 'write_upto' variables in the
431 '_vfs.block_prepare_write' and '_vfs.block_prepare_write.return'
433 - The 'regs' variable in the 'syscall.sigaltstack',
434 'nd_syscall.sigaltstack', 'syscall.fork', and 'nd_syscall.fork'
436 - The 'first', 'second', 'third', and 'uptr_uaddr' variables in the
437 'syscall.compat_sys_shmat' and 'nd_syscall.compat_sys_shmat' probe
440 - The following tapset functions are deprecated in release 2.2:
441 'ppos_pos', '_dev_minor', and '_dev_major'
443 - The folowing tapset functions used to return error strings instead
444 of raising an error. The original behavior is deprecated in release
447 'ctime', 'probemod', 'modname'
449 * What's new in version 2.1, 2013-02-13
451 - EMACS and VIM editor modes for systemtap source files are included / updated.
453 - The translator now eliminates duplicate tapset files between its
454 preferred directory (as configured during the build with --prefix=/
455 or specified with the -I /path option), and files it may find under
456 $XDG_DATA_DIRS. This should eliminate a class of conflicts between
457 parallel system- and hand-built systemtap installations.
459 - The translator accepts a --suppress-time-limits option, which defeats
460 time-related constraints, to allows probe handlers to run for indefinite
461 periods. It requires the guru mode (-g) flag to work. Add the earlier
462 --suppress-handler-errors flag for a gung-ho "just-keep-going" attitude.
464 - Perf event probes may now be read on demand. The counter probe is
465 defined using the counter-name part:
466 probe perf.type(0).config(0).counter("NAME"). The counter is
467 read in a user space probe using @perf("NAME"), e.g.
468 process("PROCESS").statement("func@file") {stat <<< @perf("NAME")}
470 - Perf event probes may now be bound to a specific task using the
471 process-name part: probe perf.type(0).config(0).process("NAME") { }
472 If the probed process name is not specified, then it is inferred
473 from the -c CMD argument.
475 - Some error messages and warnings now refer to additional information
476 that is found in man pages. These are generally named
477 error::FOO or warning::BAR (in the 7stap man page section)
482 - The dyninst backend has improved in several aspects:
483 - The runtime now allows much more concurrency when probing multithreaded
484 processes, and will also follow probes across forks.
485 - Several new probe types are now supported, including timers, function
486 return, and process.begin/end and process.thread.begin/end.
487 - Semaphores for SDT probes are now set properly.
488 - Attaching to existing processes with -x PID now works.
490 - The foreach looping construct can now sort aggregate arrays by the user's
491 choice of aggregating function. Previously, @count was implied. e.g.:
492 foreach ([x,y] in array @sum +) { println(@sum(array[x,y])) }
494 - Proof of concept support for regular expression matching has been added:
495 if ("aqqqqqb" =~ "q*b") { ... }
496 if ("abc" !~ "q*b") { ... }
498 The eventual aim is to support roughly the same functionality as
499 the POSIX Extended Regular Expressions implemented by glibc.
500 Currently missing features include extraction of the matched string
501 and subexpressions, and named character classes ([:alpha:], [:digit:], &c).
503 Special thanks go to the re2c project, whose public domain code this
504 functionality has been based on. For more info on re2c, see:
505 http://sourceforge.net/projects/re2c/
507 - The folowing tapset variables are deprecated in release 2.1 and will
508 be removed in release 2.2:
509 - The 'send2queue' variable in the 'signal.send' probe.
510 - The 'oldset_addr' and 'regs' variables in the 'signal.handle' probe.
512 - The following tapset probes are deprecated in release 2.1 and will
513 be removed in release 2.2:
515 - signal.handle.return
517 * What's new in version 2.0, 2012-10-09
519 - Systemtap includes a new prototype backend, which uses Dyninst to instrument
520 a user's own processes at runtime. This backend does not use kernel modules,
521 and does not require root privileges, but is restricted with respect to the
522 kinds of probes and other constructs that a script may use.
524 Users from source should configure --with-dyninst and install a
525 fresh dyninst snapshot such as that in Fedora rawhide. It may be
526 necessary to disable conflicting selinux checks; systemtap will advise.
528 Select this new backend with the new stap option --runtime=dyninst
529 and a -c target process, along with normal options. (-x target
530 processes are not supported in this prototype version.) For example:
532 stap --runtime=dyninst -c 'stap -l begin' \
533 -e 'probe process.function("main") { println("hi from dyninst!") }'
535 - To aid diagnosis, when a kernel panic occurs systemtap now uses
536 the panic_notifier_list facility to dump a summary of its trace
537 buffers to the serial console.
539 - The systemtap preprocessor now has a simple macro facility as follows:
541 @define add(a,b) %( ((@a)+(@b)) %)
542 @define probegin(x) %(
548 @probegin( foo = @add(40, 2); print(foo) )
550 Macros defined in the user script and regular tapset .stp files are
551 local to the file. To get around this, the tapset library can define
552 globally visible 'library macros' inside .stpm files. (A .stpm file
553 must contain a series of @define directives and nothing else.)
555 The status of the feature is experimental; semantics of macroexpansion
556 may change (unlikely) or expand in the future.
558 - Systemtap probe aliases may be used with additional suffixes
559 attached. The suffixes are passed on to the underlying probe
560 point(s) as shown below:
562 probe foo = bar, baz { }
563 probe foo.subfoo.option("gronk") { }
564 // expands to: bar.subfoo.option("gronk"), baz.subfoo.option("gronk")
566 In practical terms, this allows us to specify additional options to
567 certain tapset probe aliases, by writing e.g.
568 probe syscall.open.return.maxactive(5) { ... }
570 - To support the possibility of separate kernel and dyninst backends,
571 the tapsets have been reorganized into separate folders according to
572 backend. Thus kernel-specific tapsets are located under linux/, the
573 dyninst-specific ones under dyninst/
575 - The backtrace/unwind tapsets have been expanded to allow random
576 access to individual elements of the backtrace. (A caching mechanism
577 ensures that the backtrace computation run at most once for each
578 time a probe fires, regardless of how many times or what order the
579 query functions are called in.) New tapset functions are:
580 stack/ustack - return n'th element of backtrace
581 callers/ucallers - return first n elements of backtrace
582 print_syms/print_usyms - print full information on a list of symbols
583 sprint_syms/sprint_usyms - as above, but return info as a string
585 The following existing functions have been superseded by print_syms()
586 et al.; new scripts are recommended to avoid using them:
592 - The probefunc() tapset function is now myproc-unprivileged, and can
593 now be used in unprivileged scripts for such things as profiling in
594 userspace programs. For instance, try running
595 systemtap.examples/general/para-callgraph.stp in unprivileged mode
596 with a stapusr-permitted probe. The previous implementation of
597 probefunc() is available with "stap --compatible=1.8".
599 - Preprocessor conditional to vary code based on script privilege level:
600 unprivileged -- %( systemtap_privilege == "stapusr" %? ... %)
601 privileged -- %( systemtap_privilege != "stapusr" %? ... %)
602 or, alternately %( systemtap_privilege == "stapsys"
603 || systemtap_privilege == "stapdev" %? ... %)
605 - To ease migration to the embedded-C locals syntax introduced in 1.8
606 (namely, STAP_ARG_* and STAP_RETVALUE), the old syntax can now be
607 re-enabled on a per-function basis using the /* unmangled */ pragma:
609 function add_foo:long(a:long, b:long) %{ /* unmangled */
610 THIS->__retvalue = THIS->a + STAP_ARG_b;
613 Note that both the old and the new syntax may be used in an
614 /* unmangled */ function. Functions not marked /* unmangled */
615 can only use the new syntax.
617 - Adjacent string literals are now glued together irrespective of
618 intervening whitespace or comments:
619 "foo " "bar" --> "foo bar"
620 "foo " /* comment */ "bar" --> "foo bar"
621 Previously, the first pair of literals would be glued correctly,
622 while the second would cause a syntax error.
624 * What's new in version 1.8, 2012-06-17
626 - staprun accepts a -T timeout option to allow less frequent wake-ups
627 to poll for low-throughput output from scripts.
629 - When invoked by systemtap, the kbuild $PATH environment is sanitized
630 (prefixed with /usr/bin:/bin:) in an attempt to exclude compilers
631 other than the one the kernel was presumed built with.
633 - Printf formats can now use "%#c" to escape non-printing characters.
635 - Pretty-printed bitfields use integers and chars use escaped formatting
638 - The systemtap compile-server and client now support IPv6 networks.
639 - IPv6 addresses may now be specified on the --use-server option and will
640 be displayed by --list-servers, if the avahi-daemon service is running and
642 - Automatic server selection will automatically choose IPv4 or IPv6 servers
643 according to the normal server selection criteria when avahi-daemon is
644 running. One is not preferred over the other.
645 - The compile-server will automatically listen on IPv6 addresses, if
647 - To enable IPv6 in avahi-daemon, ensure that /etc/avahi/avahi-daemon.conf
648 contains an active "use-ipv6=yes" line. After adding this line run
649 "service avahi-daemon restart" to activate IPv6 support.
650 - See man stap(1) for details on how to use IPv6 addresses with the
653 - Support for DWARF4 .debug_types sections (for executables and shared
654 libraries compiled with recent GCC's -gdwarf-4 / -fdebug-types-section).
655 PR12997. SystemTap now requires elfutils 0.148+, full .debug_types support
656 depends on elfutils 0.154+.
658 - Systemtap modules are somewhat smaller & faster to compile. Their
659 debuginfo is now suppressed by default; use -B CONFIG_DEBUG_INFO=y to
662 - @var now an alternative language syntax for accessing DWARF variables
663 in uprobe and kprobe handlers (process, kernel, module). @var("somevar")
664 can be used where $somevar can be used. The @var syntax also makes it
665 possible to access non-local, global compile unit (CU) variables by
666 specifying the CU source file as follows @var("somevar@some/src/file.c").
667 This will provide the target variable value of global "somevar" as defined
668 in the source file "some/src/file.c". The @var syntax combines with all
669 normal features of DWARF target variables like @defined(), @entry(),
670 [N] array indexing, field access through ->, taking the address with
671 the & prefix and shallow or deep pretty printing with a $ or $$ suffix.
673 - Stap now has resource limit options:
679 All resource limiting has been moved from the compile server to stap
680 itself. When running the server as "stap-server", default resource
681 limit values are specified in ~stap-server/.systemtap/rc.
683 - Bug CVE-2012-0875 (kernel panic when processing malformed DWARF unwind data)
686 - The systemtap compile-server now supports multiple concurrent connections.
687 Specify the desired maximum number of concurrent connections with
688 the new stap-server/stap-serverd --max-threads option. Specify a
689 value of '0' to tell the server not to spawn any new threads (handle
690 all connections serially in the main thread). The default value is
691 the number of processor cores on the host.
693 - The following tapset functions are deprecated in release 1.8 and will be
694 removed in release 1.9:
697 - SystemTap now mangles local variables to avoid collisions with C
698 headers included by tapsets. This required a change in how
699 embedded-C functions access local parameters and the return value slot.
701 Instead of THIS->foo in an embedded-C function, please use the newly
702 defined macro STAP_ARG_foo (substitute the actual name of the
703 argument for 'foo'); instead of THIS->__retvalue, use the newly
704 defined STAP_RETVALUE. All of the tapsets and test cases have been
705 adapted to use this new notation.
707 If you need to run code which uses the old THIS-> notation, run stap
708 with the --compatible=1.7 option.
710 - There is updated support for user-space probing against kernels >=
711 3.5, which have no utrace but do have the newer inode-uprobes work
712 by Srikar Dronamraju and colleagues. For kernels < 3.5, the
713 following 3 sets of kernel patches would need to be backported to
714 your kernel to use this preliminary user-space probing support:
716 - inode-uprobes patches:
717 - 2b144498350860b6ee9dc57ff27a93ad488de5dc
718 - 7b2d81d48a2d8e37efb6ce7b4d5ef58822b30d89
719 - a5f4374a9610fd7286c2164d4e680436727eff71
720 - 04a3d984d32e47983770d314cdb4e4d8f38fccb7
721 - 96379f60075c75b261328aa7830ef8aa158247ac
722 - 3ff54efdfaace9e9b2b7c1959a865be6b91de96c
723 - 35aa621b5ab9d08767f7bc8d209b696df281d715
724 - 900771a483ef28915a48066d7895d8252315607a
725 - e3343e6a2819ff5d0dfc4bb5c9fb7f9a4d04da73
726 - exec tracepoint kernel patch:
727 - 4ff16c25e2cc48cbe6956e356c38a25ac063a64d
728 - task_work_add kernel patches:
729 - e73f8959af0439d114847eab5a8a5ce48f1217c4
730 - 4d1d61a6b203d957777d73fcebf19d90b038b5b2
731 - 413cd3d9abeaef590e5ce00564f7a443165db238
732 - dea649b8ac1861107c5d91e1a71121434fc64193
733 - f23ca335462e3c84f13270b9e65f83936068ec2c
735 * What's new in version 1.7, 2012-02-01
737 - Map inserting and deleting is now significantly faster due to
738 improved hashing and larger hash tables. The hashes are also
739 now randomized to provide better protection against deliberate
742 - Formatted printing is faster by compiling the formatting directives
743 to C code rather than interpreting at run time.
745 - Systemtap loads extra command line options from $SYSTEMTAP_DIR/rc
746 ($HOME/.systemtap/rc by default) before the normal argc/argv. This
747 may be useful to activate site options such as --use-server or
748 --download-debuginfo or --modinfo.
750 - The stap-server has seen many improvements, and is no longer considered
753 - The stap-server service (initscript) now supports four new options:
758 These allow the specification of macro definitions to be passed to stap
759 by the server, the location of the log file, network port number and
760 NSS certificate database location respectively. These options are also
761 supported within individual server configuration files. See stap-server
762 and initscript/README.stap-server for details. The stap-server is no
763 longer activated by default.
765 - process("PATH").[library("PATH")].function("NAME").exported probes are now
766 supported to filter function() to only exported instances.
768 - The translator supports a new --suppress-handler-errors option, which
769 causes most runtime errors to be turned into quiet skipped probes. This
770 also disables the MAXERRORS and MAXSKIPPED limits.
772 - Translator warnings have been standardized and controlled by the -w / -W
775 - The translator supports a new --modinfo NAME=VALUE option to emit additional
776 MODULE_INFO(n,v) macros into the generated code.
778 - There is no more fixed maximum number of VMA pages that will be tracked
779 at runtime. This reduces memory use for those scripts that don't need any,
780 or only limited target process VMA tracking and allows easier system
781 wide probes inspecting shared library variables and/or user backtraces.
782 stap will now silently ignore -DTASK_FINDER_VMA_ENTRY_ITEMS.
784 - The tapset functions remote_id() and remote_uri() identify the member of a
785 swarm of "stap --remote FOO --remote BAR baz.stp" concurrent executions.
787 - Systemtap now supports a new privilege level and group, "stapsys", which
788 is equivalent to the privilege afforded by membership in the group "stapdev",
789 except that guru mode (-g) functionality may not be used. To support this, a
790 new option, --privilege=[stapusr|stapsys|stapdev] has been added.
791 --privilege=stapusr is equivalent to specifying the existing --unprivileged
792 option. --privilege=stapdev is the default. See man stap(1) for details.
794 - Scripts that use kernel.trace("...") probes compile much faster.
796 - The systemtap module cache is cleaned less frequently, governed by the
797 number of seconds in the $SYSTEMTAP_DIR/cache/cache_clean_interval_s file.
799 - SDT can now define up to 12 arguments in a probe point.
801 - Parse errors no longer generate a cascade of false errors. Instead, a
802 parse error skips the rest of the current probe or function, and resumes
803 at the next one. This should generate fewer and better messages.
805 - Global array wrapping is now supported for both associative and statistics typed
806 arrays using the '%' character to signify a wrapped array. For example,
807 'global foo%[100]' would allow the array 'foo' to be wrapped if more than 100
808 elements are inserted.
810 - process("PATH").library("PATH").plt("NAME") probes are now supported.
811 Wildcards are supported in the plt-name part, to refer to any function in the
812 program linkage table which matches the glob pattern and the rest of the
815 - A new option, --dump-probe-types, will dump a list of supported probe types.
816 If --unprivileged is also specified, the list will be limited to probe types
817 which are available to unprivileged users.
819 - Systemtap can now automatically download the required debuginfo
820 using abrt. The --download-debuginfo[=OPTION] can be used to
821 control this feature. Possible values are: 'yes', 'no', 'ask',
822 and a positive number representing the timeout desired. The
823 default behavior is to not automatically download the debuginfo.
825 - The translator has better support for probing C++ applications by
826 better undertanding of compilation units, nested types, templates,
827 as used in probe point and @cast constructs.
829 - On 2.6.29+ kernels, systemtap can now probe kernel modules that
830 arrive and/or depart during the run-time of a session. This allows
831 probing of device driver initialization functions, which had formerly been
834 - New tapset functions for cpu_clock and local_clock access were added.
836 - There is some limited preliminary support for user-space probing
837 against kernels such as linux-next, which have no utrace but do have
838 the newer inode-uprobes work by Srikar Dronamraju and colleagues.
840 - The following probe types are deprecated in release 1.7 and will be
841 removed in release 1.8:
842 kernel.function(number).inline
843 module(string).function(number).inline
844 process.function(number).inline
845 process.library(string).function(number).inline
846 process(string).function(number).inline
847 process(string).library(string).function(number).inline
849 - The systemtap-grapher is deprecated in release 1.7 and will be removed in
852 - The task_backtrace() tapset function was deprecated in 1.6 and has been
855 - MAXBACKTRACE did work in earlier releases, but has now been documented
856 in the stap 1 manual page.
858 - New tapset function probe_type(). Returns a short string describing
859 the low level probe handler type for the current probe point.
861 - Both unwind and symbol data is now only collected and emitted for
862 scripts actually using backtracing or function/data symbols.
863 Tapset functions are marked with /* pragma:symbols */ or
864 /* pragma:unwind */ to indicate they need the specific data.
866 - Kernel backtraces can now be generated for non-pt_regs probe context
867 if the kernel support dump_trace(). This enables backtraces from
868 certain timer probes and tracepoints.
870 - ubacktrace() should now also work for some kernel probes on x86 which can
871 use the dwarf unwinder to recover the user registers to provide
872 more accurate user backtraces.
874 - For s390x the systemtap runtime now properly splits kernel and user
875 addresses (which are in separate address spaces on that architecture)
876 which enable user space introspection.
878 - ppc and s390x now supports user backtraces through the DWARF unwinder.
880 - ppc now handles function descriptors as symbol names correctly.
882 - arm support kernel backtraces through the DWARF unwinder.
884 - arm now have a uprobes port which enables user probes. This still
885 requires some kernel patches (user_regsets and tracehook support for
888 - Starting in release 1.7, these old variables will be deprecated:
889 - The 'pid' variable in the 'kprocess.release' probe has been
890 deprecated in favor of the new 'released_pid' variable.
891 - The 'args' variable in the
892 '_sunrpc.clnt.create_client.rpc_new_client_inline' probe has been
893 deprecated in favor of the new internal-only '__args' variable.
895 - Experimental support for recent kernels without utrace has been
896 added for the following probe types:
899 process("PATH").begin
901 process(PID).thread.begin
902 process("PATH").thread.begin
907 process(PID).thread.end
908 process("PATH").thread.end
911 process("PATH").syscall
913 process(PID).syscall.return
914 process("PATH").syscall.return
915 process.syscall.return
917 - staprun disables kprobe-optimizations in recent kernels, as problems
918 were found. (PR13193)
920 * What's new in version 1.6, 2011-07-25
922 - Security fixes for CVE-2011-2503: read instead of mmap to load modules,
923 CVE-2011-2502: Don't allow path-based auth for uprobes
925 - The systemtap compile-server no longer uses the -k option when calling the
926 translator (stap). As a result, the server will now take advantage of the
927 module cache when compiling the same script more than once. You may observe
928 an improvement in the performance of the server in this situation.
930 - The systemtap compile-server and client now each check the version of the
931 other, allowing both to adapt when communicating with a down-level
932 counterpart. As a result, all version of the client can communicate
933 with all versions of the server and vice-versa. Client will prefer newer
934 servers when selecting a server automatically.
936 - SystemTap has improved support for the ARM architecture. The
937 kread() and kwrite() operations for ARM were corrected allowing many
938 of the tapsets probes and function to work properly on the ARM
941 - Staprun can now rename the module to a unique name with the '-R' option before
942 inserting it. Systemtap itself will also call staprun with '-R' by default.
943 This allows the same module to be inserted more than once, without conflicting
946 - Systemtap error messages now provide feedback when staprun or any other
947 process fails to launch. This also specifically covers when the user
948 doesn't have the proper permissions to launch staprun.
950 - Systemtap will now map - to _ in module names. Previously,
951 stap -L 'module("i2c-core").function("*")' would be empty. It now returns
952 a list had stap -L 'module("i2c_core").function("*") been specified.
954 - Systemtap now fills in missing process names to probe points, to
955 avoid having to name them twice twice:
956 % stap -e 'probe process("a.out").function("*") {}' -c 'a.out ...'
957 Now the probed process name is inferred from the -c CMD argument.
958 % stap -e 'probe process.function("*") {}' -c 'a.out ...'
960 - stap -L 'process("PATH").syscall' will now list context variables
962 - Depends on elfutils 0.142+.
964 - Deprecated task_backtrace:string (task:long). This function will go
965 away after 1.6. Please run your scripts with stap --check-version.
967 * What's new in version 1.5, 2011-05-23
969 - Security fixes for CVE-2011-1781, CVE-2011-1769: correct DW_OP_{mod,div}
972 - The compile server and its related tools (stap-gen-ert, stap-authorize-cert,
973 stap-sign-module) have been re-implemented in C++. Previously, these
974 components were a mix of bash scripts and C code. These changes should be
975 transparent to the end user with the exception of NSS certificate database
976 password prompting (see below). The old implementation would prompt more
977 than once for the same password in some situations.
979 - eventcount.stp now allows for event counting in the format of
980 'stap eventcount.stp process.end syscall.* ...', and also reports
981 corresponding event tid's.
983 - Systemtap checks that the build-id of the module being probed matches the
984 build-id saved in the systemtap module. Invoking systemtap with
985 -DSTP_NO_BUILDID_CHECK will bypass this build-id runtime verification. See
986 man ld(1) for info on --build-id.
988 - stapio will now report if a child process has an abnormal exit along with
989 the associated status or signal.
991 - Compiler optimization may sometimes result in systemtap not being able to
992 access a user-space probe argument. Compiling the application with
993 -DSTAP_SDT_ARG_CONSTRAINT=nr will force the argument to be an immediate or
994 register value which should enable systemtap to access the argument.
996 - GNU Gettext has now been intergrated with systemtap. Our translation
997 page can be found at http://www.transifex.net/projects/p/systemtap/ .
998 "make update-po" will generate the necessary files to use translated
999 messages. Please refer to the po/README file for more info and
1000 please consider contributing to this I18N effort!
1002 - The new addr() function returns the probe's instruction pointer.
1004 - process("...").library("...") probes are now supported. Wildcards
1005 are supported in the library-name part, to refer to any shared
1006 library that is required by process-name, which matches the glob
1007 pattern and the rest of the probe point.
1009 - The "--remote USER@HOST" functionality can now be specified multiple times
1010 to fan out on multiple targets. If the targets have distinct kernel and
1011 architecture configurations, stap will automatically build the script
1012 appropriately for each one. This option is also no longer considered
1015 - The NSS certificate database generated for use by the compile server is now
1016 generated with no password. Previously, a random password was generated and
1017 used to access the database. This change should be transparent to most users.
1018 However, if you are prompted for a password when using systemtap, then
1019 running $libexecdir/stap-gen-cert should correct the problem.
1021 - The timestamp tapset includes jiffies() and HZ() for lightweight approximate
1024 - A powerful new command line option --version has been added.
1026 - process.mark now supports $$parms for reading probe parameters.
1028 - A new command line option, --use-server-on-error[=yes|no] is available
1029 for stap. It instructs stap to retry compilation of a script using a
1030 compile server if it fails on the local host. The default setting
1033 - The following deprecated tools have been removed:
1035 stap-authorize-server-cert
1036 stap-authorize-signing-cert
1037 stap-find-or-start-server
1039 Use the --use-server, --trust-server and --list-servers options of stap
1042 * What's new in version 1.4, 2011-01-17
1044 - Security fixes for CVE-2010-4170, CVE-2010-4171: staprun module
1047 - A new /* myproc-unprivileged */ marker is now available for embedded C
1048 code and and expressions. Like the /* unprivileged */ marker, it makes
1049 the code or expression available for use in unprivileged mode (see
1050 --unprivileged). However, it also automatically adds a call to
1051 assert_is_myproc() to the code or expression, thus, making it available
1052 to the unprivileged user only if the target of the current probe is within
1053 the user's own process.
1055 - The experimental "--remote USER@HOST" option will run pass 5 on a given
1056 ssh host, after building locally (or with --use-server) for that target.
1058 - Warning messages from the script may now be suppressed with the stap
1059 and/or staprun -w option. By default, duplicate warning messages are
1060 suppressed (up to a certain limit). With stap --vp 00002 and above,
1061 the duplicate elimination is defeated.
1063 - The print_ubacktrace and usym* functions attempt to print the full
1064 path of the user-space binaries' paths, instead of just the basename.
1065 The maximum saved path length is set by -DTASK_FINDER_VMA_ENTRY_PATHLEN,
1066 default 64. Warning messages are produced if unwinding fails due to
1067 a missing 'stap -d MODULE' option, providing preloaded unwind data.
1069 - The new tz_ctime() tapset function prints times in the local time zone.
1071 - More kernel tracepoints are accessible to the kernel.trace("...") mechanism,
1072 if kernel source trees or debuginfo are available. These formerly "hidden"
1073 tracepoints are those that are declared somewhere other than the usual
1074 include/linux/trace/ headers, such as xfs and kvm.
1076 - debuginfo-based process("...").function/.statement/.mark probes support
1077 wildcards in the process-name part, to refer to any executable files that
1078 match the glob pattern and the rest of the probe point.
1080 - The -t option now displays information per probe-point rather than a summary
1081 for each probe. It also now shows the derivation chain for each probe-point.
1083 - A rewrite of the sys/sdt.h header file provides zero-cost startup (few or
1084 no ELF relocations) for the debuginfo-less near-zero-cost runtime probes.
1085 Binaries compiled with earlier sdt.h versions remain supported. The
1086 stap -L (listing) option now lists parameters for sys/sdt.h markers.
1088 - The implementation of the integrated compile-server client has been
1090 o --use-server now accepts an argument representing a particular server and
1091 may be specified more than once.
1092 o --list-servers now accepts an expanded range of arguments.
1093 o a new --trust-servers option has been added to stap to replace several
1094 old certificate-management scripts.
1095 o The following tools are now deprecated and will be removed in release 1.5:
1097 stap-authorize-server-cert
1098 stap-authorize-signing-cert
1099 stap-find-or-start-server
1101 See man stap(1) for complete details.
1103 - The compile-server now returns the uprobes.ko to the client when it is
1104 required by the script being compiled. The integrated compile-server client
1105 now makes it available to be loaded by staprun. The old (deprecated)
1106 stap-client does not do this.
1108 - process probes with scripts as the target are recognized by stap and the
1109 interpreter would be selected for probing.
1111 - Starting in release 1.5, these old variables/functions will be deprecated
1112 and will only be available when the '--compatible=1.4' flag is used:
1114 - In the 'syscall.add_key' probe, the 'description_auddr' variable
1115 has been deprecated in favor of the new 'description_uaddr'
1117 - In the 'syscall.fgetxattr', 'syscall.fsetxattr',
1118 'syscall.getxattr', 'syscall.lgetxattr', and
1119 'syscall.lremovexattr' probes, the 'name2' variable has been
1120 deprecated in favor of the new 'name_str' variable.
1121 - In the 'nd_syscall.accept' probe the 'flag_str' variable
1122 has been deprecated in favor of the new 'flags_str' variable.
1123 - In the 'nd_syscall.dup' probe the 'old_fd' variable has been
1124 deprecated in favor of the new 'oldfd' variable.
1125 - In the 'nd_syscall.fgetxattr', 'nd_syscall.fremovexattr',
1126 'nd_syscall.fsetxattr', 'nd_syscall.getxattr', and
1127 'nd_syscall.lremovexattr' probes, the 'name2' variable has been
1128 deprecated in favor of the new 'name_str' variable.
1129 - The tapset alias 'nd_syscall.compat_pselect7a' was misnamed. It should
1130 have been 'nd_syscall.compat_pselect7' (without the trailing 'a').
1131 - The tapset function 'cpuid' is deprecated in favor of the better known
1133 - In the i386 'syscall.sigaltstack' probe, the 'ussp' variable has
1134 been deprecated in favor of the new 'uss_uaddr' variable.
1135 - In the ia64 'syscall.sigaltstack' probe, the 'ss_uaddr' and
1136 'oss_uaddr' variables have been deprecated in favor of the new
1137 'uss_uaddr' and 'uoss_uaddr' variables.
1138 - The powerpc tapset alias 'syscall.compat_sysctl' was deprecated
1139 and renamed 'syscall.sysctl32'.
1140 - In the x86_64 'syscall.sigaltstack' probe, the 'regs_uaddr'
1141 variable has been deprecated in favor of the new 'regs' variable.
1143 * What's new in version 1.3, 2010-07-21
1145 - The uprobes kernel module now has about half the overhead when probing
1146 NOPs, which is particularly relevant for sdt.h markers.
1148 - New stap option -G VAR=VALUE allows overriding global variables
1149 by passing the settings to staprun as module options.
1151 - The tapset alias 'syscall.compat_pselect7a' was misnamed. It should
1152 have been 'syscall.compat_pselect7' (without the trailing 'a').
1153 Starting in release 1.4, the old name will be deprecated and
1154 will only be available when the '--compatible=1.3' flag is used.
1156 - A new procfs parameter .umask(UMASK) which provides modification of
1157 file permissions using the proper umask value. Default file
1158 permissions for a read probe are 0400, 0200 for a write probe, and
1159 0600 for a file with a read and write probe.
1161 - It is now possible in some situations to use print_ubacktrace() to
1162 get a user space stack trace from a kernel probe point. e.g. for
1163 user backtraces when there is a pagefault:
1164 $ stap -d /bin/sort --ldd -e 'probe vm.pagefault {
1165 if (pid() == target()) {
1166 printf("pagefault @0x%x\n", address); print_ubacktrace();
1169 pagefault @0x7fea0595fa70
1170 0x000000384f07f958 : __GI_strcmp+0x12b8/0x1440 [libc-2.12.so]
1171 0x000000384f02824e : __gconv_lookup_cache+0xee/0x5a0 [libc-2.12.so]
1172 0x000000384f021092 : __gconv_find_transform+0x92/0x2cf [libc-2.12.so]
1173 0x000000384f094896 : __wcsmbs_load_conv+0x106/0x2b0 [libc-2.12.so]
1174 0x000000384f08bd90 : mbrtowc+0x1b0/0x1c0 [libc-2.12.so]
1175 0x0000000000404199 : ismbblank+0x39/0x90 [sort]
1176 0x0000000000404a4f : inittables_mb+0xef/0x290 [sort]
1177 0x0000000000406934 : main+0x174/0x2510 [sort]
1178 0x000000384f01ec5d : __libc_start_main+0xfd/0x1d0 [libc-2.12.so]
1179 0x0000000000402509 : _start+0x29/0x2c [sort]
1182 - New tapset functions to get a string representation of a stack trace:
1183 sprint_[u]backtrace() and sprint_[u]stack().
1185 - New tapset function to get the module (shared library) name for a
1186 user space address umodname:string(long). The module name will now
1187 also be in the output of usymdata() and in backtrace addresses even
1188 when they were not given with -d at the command line.
1190 - Kernel backtraces are now much faster (replaced a linear search
1191 with a binary search).
1193 - A new integrated compile-server client is now available as part of stap.
1195 o 'stap --use-server ...' is equivalent to 'stap-client ...'
1196 o 'stap --list-servers' is equivalent to 'stap-find-servers'
1197 o 'stap --list-servers=online' is equivalent to 'stap-find-servers --all'
1198 o stap-client and its related tools will soon be deprecated.
1199 o the nss-devel and avahi-devel packages are required for building stap with
1200 the integrated client (checked during configuration).
1201 o nss and avahi are required to run the integrated client.
1203 - A new operator @entry is available for automatically saving an expression
1204 at entry time for use in a .return probe.
1205 probe foo.return { println(get_cycles() - @entry(get_cycles())) }
1207 - Probe $target variables and @cast() can now use a suffix to print complex
1208 data types as strings. Use a single '$' for a shallow view, or '$$' for a
1209 deeper view that includes nested types. For example, with fs_struct:
1210 $fs$ : "{.users=%i, .lock={...}, .umask=%i,
1211 .in_exec=%i, .root={...}, .pwd={...}}"
1212 $fs$$ : "{.users=%i, .lock={.raw_lock={.lock=%u}}, .umask=%i, .in_exec=%i,
1213 .root={.mnt=%p, .dentry=%p}, .pwd={.mnt=%p, .dentry=%p}}"
1215 - The <sys/sdt.h> user-space markers no longer default to an implicit
1216 MARKER_NAME_ENABLED() semaphore check for each marker. To check for
1217 enabled markers use a .d declaration file, then:
1218 if (MARKER_NAME_ENABLED()) MARKER_NAME()
1220 - Hyphenated <sys/sdt.h> marker names such as process(...).mark("foo-bar")
1221 are now accepted in scripts. They are mapped to the double-underscore
1224 - More robust <sys/sdt.h> user-space markers support is included. For
1225 some platforms (x86*, ppc*), this can let systemtap probe the markers
1226 without debuginfo. This implementation also supports preserving
1227 the "provider" name associated with a marker:
1228 probe process("foo").provider("bar").mark("baz") to match
1229 STAP_PROBE<n>(bar, baz <...>)
1230 (Compile with -DSTAP_SDT_V1 to revert to the previous implementation.
1231 Systemtap supports pre-existing or new binaries using them.)
1233 - Embedded-C may be used within expressions as values, when in guru mode:
1234 num = %{ LINUX_VERSION_CODE %} // int64_t
1235 name = %{ /* string */ THIS_MODULE->name %} // const char*
1236 printf ("%s %x\n", name, num)
1237 The usual /* pure */, /* unprivileged */, and /* guru */ markers may be used
1238 as with embedded-C functions.
1240 - By default the systemtap-runtime RPM builds now include a shared
1241 library, staplog.so, that allows crash to extract systemtap data from
1244 - Iterating with "foreach" can now explicitly save the value for the loop.
1245 foreach(v = [i,j] in array)
1246 printf("array[%d,%s] = %d\n", i, j, v /* array[i,j] */)
1248 - The new "--ldd" option automatically adds any additional shared
1249 libraries needed by probed or -d-listed userspace binaries to the -d
1250 list, to enable symbolic backtracing through them. Similarly, the
1251 new "--all-modules" option automatically adds any currently loaded
1252 kernel modules (listed in /proc/modules) to the -d list.
1254 - A new family of set_kernel_* functions make it easier for gurus to write
1255 new values at arbitrary memory addresses.
1257 - Probe wildcards can now use '**' to cross the '.' separator.
1258 $ stap -l 'sys**open'
1262 - Backward compatibility flags (--compatible=VERSION, and matching
1263 script preprocessing predicate %( systemtap_v CMP "version" %)
1264 and a deprecation policy are being introduced, in case future
1265 tapset/language changes break valid scripts.
1267 * What's new in version 1.2, 2010-03-22
1269 - Prototype support for "perf events", where the kernel supports the
1270 2.6.33 in-kernel API. Probe points may refer to low-level
1271 perf_event_attr type/config numbers, or to a number of aliases
1272 defined in the new perf.stp tapset:
1273 probe perf.sw.cpu_clock, perf.type(0).config(4) { }
1275 - Type-casting can now use multiple headers to resolve codependencies.
1276 @cast(task, "task_struct",
1277 "kernel<linux/sched.h><linux/fs_struct.h>")->fs->umask
1279 - Tapset-related man pages have been renamed. 'man -k 3stap' should show
1280 the installed list, which due to prefixing should no longer collide over
1281 ordinary system functions.
1283 - User space marker arguments no longer use volatile if the version of gcc,
1284 which must be at least 4.5.0, supports richer DWARF debuginfo. Use cflags
1285 -DSTAP_SDT_VOLATILE=volatile or -DSTAP_SDT_VOLATILE= when building
1286 the sys/sdt.h application to override this one way or another.
1288 - A new construct for error handling is available. It is similar to c++
1289 exception catching, using try and catch as new keywords. Within a handler
1290 or function, the following is valid and may be nested:
1291 try { /* arbitrary statements */ }
1292 catch (er) { /* e.g. println("caught error ", er) */ }
1294 - A new command line flag '-W' forces systemtap to abort translation of
1295 a script if any warnings are produced. It is similar to gcc's -Werror.
1296 (If '-w' is also supplied to suppress warnings, it wins.)
1298 - A new predicate @defined is available for testing whether a
1299 particular $variable/expression is resolvable at translate time:
1300 probe foo { if (@defined($bar)) log ("$bar is available here") }
1302 - Adjacent string literals are glued together, making this
1304 probe process("/usr" @1 "/bin").function("*") { ... }
1306 - In order to limit potential impact from future security problems,
1307 the stap-server process does not permit its being launched as root.
1309 - On recent kernels, for some architectures/configurations, hardware
1310 breakpoint probes are supported. The probe point syntax is:
1312 probe kernel.data(ADDRESS).write
1313 probe kernel.data(ADDRESS).length(LEN).write
1314 probe kernel.data("SYMBOL_NAME").write
1316 * What's new in version 1.1, 2010-01-15
1318 - New tracepoint based tapset for memory subsystem.
1320 - The loading of signed modules by staprun is no longer allowed for
1321 ordinary, unprivileged users. This means that only root, members of
1322 the group 'stapdev' and members of the group 'stapusr' can load
1323 systemtap modules using staprun, stap or stap-client. The minimum
1324 privilege required to run arbitrary --unprivileged scripts is now
1325 'stapusr' membership.
1327 - The stap-server initscript is available. This initscript allows you
1328 to start systemtap compile servers as a system service and to manage
1329 these servers as a group or individually. The stap-server initscript
1330 is installed by the systemtap-server rpm. The build directory for
1331 the uprobes module (/usr/share/systemtap/runtime/uprobes) is made
1332 writable by the 'stap-server' group. All of the files generated when
1333 building the uprobes module, including the digital signature, are
1334 also writable by members of stap-server.
1336 See initscript/README.stap-server for details.
1338 - Some of the compile server client, server and certificate management
1339 tools have been moved from $bindir to $libexecdir/systemtap.
1340 You should use the new stap-server script or the stap-server initscript
1341 for server management where possible. The stap-server script provides the same
1342 functionality as the stap-server initscript except that the servers are
1343 run by the invoking user by default as opposed to servers started by the
1344 stap-server initscript which are run by the user stap-server
1345 by default. See stap-server(8) for more information.
1347 You may continue to use these tools by adding $libexecdir/systemtap to
1348 your path. You would need to do this, for example, if you are not root,
1349 you want to start a compile server and you are not running systemtap from a
1350 private installation. In this case you still need to use stap-start-server.
1352 - Any diagnostic output line that starts with "ERROR", as in
1353 error("foo"), will promote a "Pass 5: run failed", and the return
1356 - Systemtap now warns about global variables being referenced from other
1357 script files. This aims to protect against unintended local-vs-global
1358 namespace collisions such as:
1360 % cat some_tapset.stp
1361 probe baz.one = bar { foo = $foo; bar = $bar }
1362 % cat end_user_script.stp
1363 global foo # intended to be private variable
1364 probe timer.s(1) { foo ++ }
1365 probe baz.* { println(foo, pp()) }
1366 % stap end_user_script.stp
1367 WARNING: cross-file global variable reference to foo from some_tapset.stp
1369 - Preprocessor conditional for kernel configuration testing:
1370 %( CONFIG_foo == "y" %? ... %)
1372 - ftrace(msg:string) tapset function to send strings to the system-wide
1373 ftrace ring-buffer (if any).
1375 - Better support for richer DWARF debuginfo output from GCC 4.5
1376 (variable tracking assignments). Kernel modules are now always resolved
1377 against all their dependencies to find any info referring to missing
1378 symbols. DW_AT_const_value is now supported when no DW_AT_location
1381 * What's new in verson 1.0, 2009-09-22
1383 - process().mark() probes now use an enabling semaphore to reduce the
1384 computation overhead of dormant probes.
1386 - The function spec for dwarf probes now supports C++ scopes, so you can
1387 limit the probes to specific namespaces or classes. Multiple scopes
1388 can be specified, and they will be matched progressively outward.
1389 probe process("foo").function("std::vector<*>::*") { }
1390 probe process("foo").function("::global_function") { }
1392 - It is now possible to cross-compile systemtap scripts for foreign
1393 architectures, using the new '-a ARCH' and '-B OPT=VALUE' flags.
1394 For example, put arm-linux-gcc etc. into your $PATH, and point
1395 systemtap at the target kernel build tree with:
1396 stap -a arm -B CROSS_COMPILE=arm-linux- -r /build/tree [...]
1397 The -B option is passed to kbuild make. -r identifies the already
1398 configured/built kernel tree and -a its architecture (kbuild ARCH=...).
1399 Systemtap will infer -p4.
1401 - Cross compilation using the systemtap client and server
1402 - stap-start-server now accepts the -r, -R, -I, -B and -a options in
1403 order to start a cross compiling server. The server will correctly
1404 advertise itself with respect to the kernel release and architecture
1405 that it compiles for.
1406 - When specified on stap-client, the -r and -a options will be
1407 considered when searching for a suitable server.
1409 - When using the systemtap client and server udp port 5353 must be open
1410 in your firewall in order for the client to find servers using
1411 avahi-browse. Also the systemtap server will choose a random port in
1412 the range 1024-63999 for accepting ssl connections.
1414 - Support for unprivileged users:
1415 ***********************************************************************
1416 * WARNING!!!!!!!!!! *
1417 * This feature is EXPERIMENTAL at this time and should be used with *
1418 * care. This feature allows systemtap kernel modules to be loaded by *
1419 * unprivileged users. The user interface and restrictions will change *
1420 * as this feature evolves. *
1421 ***********************************************************************
1422 - Systemtap modules generated from scripts which use a restricted
1423 subset of the features available may be loaded by staprun for
1424 unprivileged users. Previously, staprun would load modules only for
1425 root or for members of the groups stapdev and stapusr.
1426 - Using the --unprivileged option on stap enables translation-time
1427 checking for use by unprivileged users (see restrictions below).
1428 - All modules deemed suitable for use by unprivileged users will be
1429 signed by the systemtap server when --unprivileged is specified on
1430 stap-client. See module signing in release 0.9.8 and stap-server in
1432 - Modules signed by trusted signers (servers) and verified by staprun
1433 will be loaded by staprun regardless of the user's privilege level.
1434 - The system administrator asserts the trustworthiness of a signer
1435 (server) by running stap-authorize-signing-cert <cert-file> as root,
1436 where the <cert-file> can be found in
1437 ~<user>/.systemtap/ssl/server/stap.cert for servers started by
1438 ordinary users and in $sysconfdir/systemtap/ssl/server/stap.cert for
1439 servers started by root.
1440 - Restrictions are intentionally strict at this time and may be
1441 relaxed in the future:
1442 - probe points are restricted to:
1443 begin, begin(n), end, end(n), error, error(n), never,
1444 timer.{jiffies,s,sec,ms,msec,us,usec,ns,nsec}(n)*, timer.hz(n),
1445 process.* (for processes owned by the user).
1446 - use of embedded C code is not allowed.
1447 - use of tapset functions is restricted.
1448 - some tapset functions may not be used at all. A message will be
1449 generated at module compilation time.
1450 - some actions by allowed tapset functions may only be performed
1451 in the context of the user's own process. A runtime fault will
1452 occur in these situations, for example, direct memory access.
1453 - The is_myproc() tapset function has been provided so that
1454 tapset writers for unprivileged users can check that the
1455 context is of the users own process before attempting these
1457 - accessing the kernel memory space is not allowed.
1458 - The following command line options may not be used by stap-client
1460 - The following environment variables are ignored by stap-client:
1461 SYSTEMTAP_RUNTIME, SYSTEMTAP_TAPSET, SYSTEMTAP_DEBUGINFO_PATH
1462 - nss and nss-tools are required to use this feature.
1464 - Support output file switching by SIGUSR2. Users can command running
1465 stapio to switch output file by sending SIGUSR2.
1467 - Memory consumption for scripts involving many uprobes has been
1468 dramatically reduced.
1470 - The preprocessor now supports || and && in the conditions.
1471 e.g. %( arch == "x86_64" || arch == "ia64" %: ... %)
1473 - The systemtap notion of "architecture" now matches the kernel's, rather
1474 than that of "uname -m". This means that 32-bit i386 family are all
1475 known as "i386" rather than "i386" or "i686"; "ppc64" as "powerpc";
1476 "s390x" as "s390", and so on. This is consistent between the new
1477 "-a ARCH" flag and the script-level %( arch ... %) conditional.
1479 - It is now possible to define multiple probe aliases with the same name.
1480 A probe will expand to all matching aliases.
1483 probe foo { } # expands twice, once to bar and once to baz
1485 - A new experimental transport mechanism, using ftrace's ring_buffer,
1486 has been added. This may become the default transport mechanism in
1487 future versions of systemtap. To test this new transport mechanism,
1488 define 'STP_USE_RING_BUFFER'.
1490 - Support for recognizing DW_OP_{stack,implicit}_value DWARF expressions
1491 as emitted by GCC 4.5.
1493 * What's new in version 0.9.9, 2009-08-04
1495 - Systemwide kernel .function.return (kretprobe) maxactive defaults may
1496 be overridden with the -DKRETACTIVE=nnn parameter.
1498 - Translation pass 2 is significantly faster by avoiding unnecessary
1499 searching through a kernel build/module directory tree.
1501 - When compiled against elfutils 0.142 systemtap now handles the new
1502 DW_OP_call_frame_CFA generated by by GCC.
1504 - uprobes and ustack() are more robust when used on applications that
1505 depend on prelinked/separate debuginfo shared libraries.
1507 - User space PROBE marks are not always found with or without separate
1508 debuginfo. The .probes section itself is now always put in the main
1509 elf file and marked as allocated. When building pic code the section
1510 is marked writable. The selinux memory check problems seen with
1511 programs using STAP_PROBES is fixed.
1513 - statement() probes can now override "address not at start of statement"
1514 errors in guru mode. They also provide alternative addresses to use
1517 - The stapgraph application can generate graphs of data and events
1518 emitted by systemtap scripts in real time. Run "stapgraph
1519 testsuite/systemtap.examples/general/grapher.stp" for an example of
1520 graphing the system load average and keyboard events.
1522 - Dwarf probes now show parameters and local variables in the verbose
1525 - Symbol aliases are now resolved to their canonical dwarf names. For
1526 example, probing "malloc" in libc resolves to "__libc_malloc".
1528 - The syntax for dereferencing $target variables and @cast() gained new
1530 - Array indexes can now be arbitrary numeric expressions.
1531 - Array subscripts are now supported on pointer types.
1532 - An '&' operator before a @cast or $target returns the address of the
1533 final component, especially useful for nested structures.
1535 - For reading all probe variables, kernel.mark now supports $$vars and
1536 $$parms, and process.syscall now supports $$vars.
1538 - The SNMP tapset provides probes and functions for many network
1539 statistics. See stapprobes.snmp(3stap) for more details.
1541 - The dentry tapset provides functions to map kernel VFS directory entries
1542 to file or full path names: d_path(), d_name() and reverse_path_walk().
1544 - SystemTap now has userspace markers in its own binaries, and the stap
1545 tapset provides the available probepoints and local variables.
1547 - Miscellaneous new tapset functions:
1548 - pgrp() returns the process group ID of the current process
1549 - str_replace() performs string replacement
1551 * What's new in version 0.9.8, 2009-06-11
1553 - Miscellaneous new tapset functions:
1554 - sid() returns the session ID of the current process
1555 - stringat() indexes a single character from a string.
1557 - Using %M in print formats for hex dumps can now print entire buffers,
1558 instead of just small numbers.
1560 - Dwarfless syscalls: The nd_syscalls tapset is now available to probe
1561 system calls without requiring kernel debugging information. All of
1562 the same probepoints in the normal syscalls tapset are available with
1563 an "nd_" prefix, e.g. syscall.open becomes nd_syscall.open. Most
1564 syscall arguments are also available by name in nd_syscalls.
1566 - Module signing: If the appropriate nss libraries are available on your
1567 system, stap-server will sign each compiled module using a self-generated
1568 certificate. This is the first step toward extending authority to
1569 load certain modules to unprivileged users. For now, if the system
1570 administrator adds a certificate to a database of trusted signers
1571 (stap-authorize-signing-cert), modules signed using that certificate
1572 will be verified by staprun against tampering. Otherwise, you should
1573 notice no difference in the operation of stap or staprun.
1575 * What's new in version 0.9.7, 2009-04-23
1577 - @cast can now determine its type information using an explicit header
1578 specification. For example:
1579 @cast(tv, "timeval", "<sys/time.h>")->tv_sec
1580 @cast(task, "task_struct", "kernel<linux/sched.h>")->tgid
1582 - The overlapping process.* tapsets are now separated. Those probe points
1583 documented in stapprobes(3stap) remain the same. Those that were formerly
1584 in stapprobes.process(3stap) have been renamed to kprocess, to reflect
1585 their kernel perspective on processes.
1587 - The --skip-badvars option now also suppresses run-time error
1588 messages that would otherwise result from erroneous memory accesses.
1589 Such accesses can originate from $context expressions fueled by
1590 erroneous debug data, or by kernel_{long,string,...}() tapset calls.
1592 - New probes kprobe.function(FUNCTION) and kprobe.function(FUNCTION).return
1593 for dwarfless probing. These postpone function address resolution to
1594 run-time and use the kprobe symbol-resolution mechanism.
1595 Probing of absolute statements can be done using the
1596 kprobe.statement(ADDRESS).absolute construct.
1598 - EXPERIMENTAL support for user process unwinding. A new collection of
1599 tapset functions have been added to handle user space backtraces from
1600 probe points that support them (currently process and timer probes -
1601 for timer probes test whether or not in user space first with the
1602 already existing user_mode() function). The new tapset functions are:
1603 uaddr - User space address of current running task.
1604 usymname - Return the symbol of an address in the current task.
1605 usymdata - Return the symbol and module offset of an address.
1606 print_ustack - Print out stack for the current task from string.
1607 print_ubacktrace - Print stack back trace for current task.
1608 ubacktrace - Hex backtrace of current task stack.
1609 Please read http://sourceware.org/ml/systemtap/2009-q2/msg00364.html
1610 on the current restrictions and possible changes in the future and
1611 give feedback if you want to influence future developments.
1613 * What's new in version 0.9.5, 2009-03-27
1615 - New probes process().insn and process().insn.block that allows
1616 inspection of the process after each instruction or block of
1617 instructions executed. So to count the total number of instructions
1618 a process executes during a run do something like:
1619 $ stap -e 'global steps; probe process("/bin/ls").insn {steps++}
1620 probe end {printf("Total instructions: %d\n", steps);}' \
1622 This feature can slow down execution of a process somewhat.
1624 - Systemtap probes and function man pages extracted from the tapsets
1625 are now available under 3stap. To show the page for probe vm.pagefault
1626 or the stap function pexecname do:
1627 $ man 3stap vm.pagefault
1628 $ man 3stap pexecname
1630 - Kernel tracepoints are now supported for probing predefined kernel
1631 events without any debuginfo. Tracepoints incur less overhead than
1632 kprobes, and context parameters are available with full type
1633 information. Any kernel 2.6.28 and later should have defined
1634 tracepoints. Try the following to see what's available:
1635 $ stap -L 'kernel.trace("*")'
1637 - Typecasting with @cast now supports modules search paths, which is
1638 useful in case there are multiple places where the type definition
1639 may be found. For example:
1640 @cast(sdev, "scsi_device", "kernel:scsi_mod")->sdev_state
1642 - On-file flight recorder is supported. It allows stap to record huge
1643 trace log on the disk and to run in background.
1644 Passing -F option with -o option runs stap in background mode. In this
1645 mode, staprun is detached from console, and stap itself shows staprun's
1647 Specifying the max size and the max number of log files are also available
1648 by passing -S option. This option has one or two arguments seperated by
1649 a comma. The first argument is the max size of a log file in MB. If the
1650 size of a log file exceeds it, stap switches to the next log file
1651 automatically. The second is how many files are kept on the disk. If the
1652 number of log files exceeds it, the oldest log file is removed
1653 automatically. The second argument can be omitted.
1655 For example, this will record output on log files each of them is smaller
1656 than 1024MB and keep last 3 logs, in background.
1657 % stap -F -o /tmp/staplog -S 1024,3 script.stp
1659 - In guru mode (-g), the kernel probing blacklist is disabled, leaving
1660 only a subset - the kernel's own internal kprobe blacklist - to attempt
1661 to filter out areas unsafe to probe. The differences may be enough to
1662 probe more interrupt handlers.
1664 - Variables unavailable in current context may be skipped by setting a
1665 session level flag with command line option --skip-badvars now available.
1666 This replaces any dwarf $variable expressions that could not be resolved
1667 with literal numeric zeros, along with a warning message.
1669 - Both kernel markers and kernel tracepoint support argument listing
1670 through stap -L 'kernel.mark("*")' or stap -L 'kernel.trace("*")'
1672 - Users can use -DINTERRUPTIBLE=0 to prevent interrupt reentrancy in
1673 their script, at the cost of a bit more overhead to toggle the
1676 - Added reentrancy debugging. If stap is run with the arguments
1677 "-t -DDEBUG_REENTRANCY", additional warnings will be printed for
1678 every reentrancy event, including the probe points of the
1679 resident and interloper probes.
1681 - Default to --disable-pie for configure.
1682 Use --enable-pie to turn it back on.
1684 - Improved sdt.h compatibility and test suite for static dtrace
1685 compatible user space markers.
1687 - Some architectures now use syscall wrappers (HAVE_SYSCALL_WRAPPERS).
1688 The syscall tapset has been enhanced to take care of the syscall
1689 wrappers in this release.
1691 - Security fix for CVE-2009-0784: stapusr module-path checking race.
1693 * What's new in version 0.9, 2009-02-19
1695 - Typecasting is now supported using the @cast operator. A script can
1696 define a pointer type for a "long" value, and then access type members
1697 using the same syntax as with $target variables. For example, this will
1698 retrieve the parent pid from a kernel task_struct:
1699 @cast(pointer, "task_struct", "kernel")->parent->pid
1701 - process().mark() probes are now possible to trace static user space
1702 markers put in programs with the STAP_PROBE macro using the new
1703 sys/sdt.h include file. This also provides dtrace compatible markers
1704 through DTRACE_PROBE and an associated python 'dtrace' script that
1705 can be used in builds based on dtrace that need dtrace -h or -G
1708 - For those that really want to run stap from the build tree there is
1709 now the 'run-stap' script in the top-level build directory that sets
1710 up the SYSTEMTAP_TAPSET, SYSTEMTAP_RUNTIME, SYSTEMTAP_STAPRUN, and
1711 SYSTEMTAP_STAPIO environment variables (installing systemtap, in a
1712 local prefix, is still recommended for common use).
1714 - Systemtap now comes with a new Beginners Guide that walks the user
1715 through their first steps setting up stap, understanding how it all
1716 works, introduces some useful scripts and describes some common
1717 pitfalls. It isn't created by default since it needs a Publican
1718 setup, but full build instructions can be found in the wiki:
1719 http://sourceware.org/systemtap/wiki/PublicanQuikHowto
1720 An online version can be found at:
1721 http://sourceware.org/systemtap/SystemTap_Beginners_Guide.pdf
1723 - Standard tapsets included with Systemtap were modified to include
1724 extractable documentation information based on the kernel-doc
1725 infrastructure. When configured --enabled-docs a HTML and PDF
1726 version of the Tapset Reference Manual is produced explaining probes
1727 defined in each tapset.
1729 - The systemtap client and compile server are now available.
1730 These allow you to compile a systemtap module on a host other than
1731 the one which it will be run, providing the client and server
1732 are compatible. Other than using a server for passes 1 through
1733 4, the client behaves like the 'stap' front end itself. This
1734 means, among other things, that the client will automatically
1735 load the resulting module on the local host unless -p[1234]
1736 was specified. See stap-server(8) for more details.
1737 The client/server now use SSL for network connection security and
1740 The systemtap client and server are prototypes only. Interfaces, options
1741 and usage may change at any time.
1743 - function("func").label("label") probes are now supported to allow matching
1744 the label of a function.
1746 - Systemtap initscript is available. This initscript allows you to run
1747 systemtap scripts as system services (in flight recorder mode) and
1748 control those scripts individually.
1749 See README.systemtap for details.
1751 - The stap "-r DIR" option may be used to identify a hand-made kernel
1752 build directory. The tool determines the appropriate release string
1753 automatically from the directory.
1755 - Serious problems associated with user-space probing in shared libraries
1756 were corrected, making it now possible to experiment with probe shared
1757 libraries. Assuming dwarf debugging information is installed, use this
1758 twist on the normal syntax:
1760 probe process("/lib64/libc-2.8.so").function("....") { ... }
1762 This would probe all threads that call into that library. Running
1763 "stap -c CMD" or "stap -x PID" naturally restricts this to the target
1764 command+descendants only. $$vars etc. may be used.
1766 - For scripts that sometimes terminate with excessive "skipped" probes,
1767 rerunning the script with "-t" (timing) will print more details about
1768 the skippage reasons.
1770 - Symbol tables and unwind (backtracing) data support were formerly
1771 compiled in for all probed modules as identified by the script
1772 (kernel; module("name"); process("file")) plus those listed by the
1773 stap "-d BINARY" option. Now, this data is included only if the systemtap
1774 script uses tapset functions like probefunc() or backtrace() that require
1775 such information. This shrinks the probe modules considerably for the rest.
1777 - Per-pass verbosity control is available with the new "--vp {N}+" option.
1778 "stap --vp 040" adds 4 units of -v verbosity only to pass 2. This is useful
1779 for diagnosing errors from one pass without excessive verbosity from others.
1781 - Most probe handlers now run with interrupts enabled, for improved
1782 system responsiveness and less probing overhead. This may result
1783 in more skipped probes, for example if a reentrant probe handler
1784 is attempted from within an interrupt handler. It may also make the
1785 systemtap overload detection facility more likely to be triggered, as
1786 interrupt handlers' run time would be included in the self-assessed
1787 overhead of running probe handlers.
1789 * What's new in version 0.8, 2008-11-13
1791 - Cache limiting is now available. If the compiled module cache size is
1792 over a limit specified in the $SYSTEMTAP_DIR/cache/cache_mb_limit file,
1793 some old cache entries will be unlinked. See man stap(1) for more.
1795 - Error and warning messages are now followed by source context displaying
1796 the erroneous line/s and a handy '^' in the following line pointing to the
1799 - A bug reporting tool "stap-report" is now available which will quickly
1800 retrieve much of the information requested here:
1801 http://sourceware.org/systemtap/wiki/HowToReportBugs
1803 - The translator can resolve members of anonymous structs / unions:
1804 given struct { int foo; struct { int bar; }; } *p;
1805 this now works: $p->bar
1807 - The stap "-F" flag activates "flight recorder" mode, which consists of
1808 translating the given script as usual, but implicitly launching it into
1809 the background with staprun's existing "-L" (launch) option. A user
1810 can later reattach to the module with "staprun -A MODULENAME".
1812 - Additional context variables are available on user-space syscall probes.
1813 - $argN ($arg1, $arg2, ... $arg6) in process(PATH_OR_PID).syscall
1814 gives you the argument of the system call.
1815 - $return in process(PATH_OR_PID).syscall.return gives you the return
1816 value of the system call.
1818 - Target process mode (stap -c CMD or -x PID) now implicitly restricts all
1819 "process.*" probes to the given child process. (It does not affect
1820 kernel.* or other probe types.) The CMD string is normally run directly,
1821 rather than via a /bin/sh -c subshell, since then utrace/uprobe probes
1822 receive a fairly "clean" event stream. If metacharacters like
1823 redirection operators were present in CMD, then "sh -c CMD" is still
1824 used, and utrace/uprobe probes will receive events from the shell.
1826 % stap -e 'probe process.syscall, process.end {
1827 printf("%s %d %s\n", execname(), pid(), pp())}'\
1829 ls 2323 process.syscall
1830 ls 2323 process.syscall
1833 - Probe listing mode is improved: "-L" lists available script-level variables
1835 % stap -L 'syscall.*open*'
1836 syscall.mq_open name:string name_uaddr:long filename:string mode:long u_attr_uaddr:long oflag:long argstr:string
1837 syscall.open name:string filename:string flags:long mode:long argstr:string
1838 syscall.openat name:string filename:string flags:long mode:long argstr:string
1840 - All user-space-related probes support $PATH-resolved executable
1843 probe process("ls").syscall {}
1844 probe process("./a.out").syscall {}
1846 work now, instead of just
1848 probe process("/bin/ls").syscall {}
1849 probe process("/my/directory/a.out").syscall {}
1851 - Prototype symbolic user-space probing support:
1853 # stap -e 'probe process("ls").function("*").call {
1854 log (probefunc()." ".$$parms)
1859 - debugging information for the named program
1860 - a version of utrace in the kernel that is compatible with the "uprobes"
1861 kernel module prototype. This includes RHEL5 and older Fedora, but not
1862 yet current lkml-track utrace; a "pass 4a"-time build failure means
1863 your system cannot use this yet.
1865 - Global variables which are written to but never read are now
1866 automatically displayed when the session does a shutdown. For example:
1868 global running_tasks
1869 probe timer.profile {running_tasks[pid(),tid()] = execname()}
1870 probe timer.ms(8000) {exit()}
1872 - A formatted string representation of the variables, parameters, or local
1873 variables at a probe point is now supported via the special $$vars,
1874 $$parms, and $$locals context variables, which expand to a string
1875 containing a list "var1=0xdead var2=0xbeef var3=?". (Here, var3 exists
1876 but is for some reason unavailable.) In return probes only, $$return
1877 expands to an empty string for a void function, or "return=0xf00".
1880 * What's new in version 0.7, 2008-07-15
1882 - .statement("func@file:*") and .statement("func@file:M-N") probes are now
1883 supported to allow matching a range of lines in a function. This allows
1884 tracing the execution of a function.
1886 - Scripts relying on probe point wildcards like "syscall.*" that expand
1887 to distinct kprobes are processed significantly faster than before.
1889 - The vector of script command line arguments is available in a
1890 tapset-provided global array argv[]. It is indexed 1 ... argc,
1891 another global. This can substitute for of preprocessor
1892 directives @NNN that fail at parse time if there are not
1895 printf("argv: %s %s %s", argv[1], argv[2], argv[3])
1897 - .statement("func@file+line") probes are now supported to allow a
1898 match relative to the entry of the function incremented by line
1899 number. This allows using the same systemtap script if the rest
1900 of the file.c source only changes slightly.
1902 - A probe listing mode is available.
1911 - More user-space probe types are added:
1913 probe process(PID).begin { }
1914 probe process("PATH").begin { }
1915 probe process(PID).thread.begin { }
1916 probe process("PATH").thread.begin { }
1917 probe process(PID).end { }
1918 probe process("PATH").end { }
1919 probe process(PID).thread.end { }
1920 probe process("PATH").thread.end { }
1921 probe process(PID).syscall { }
1922 probe process("PATH").syscall { }
1923 probe process(PID).syscall.return { }
1924 probe process("PATH").syscall.return { }
1926 - Globals now accept ; terminators
1929 global little[10], big[5];
1931 * What's new in version 0.6, 2007-12-15
1933 - A copy of the systemtap tutorial and language reference guide
1936 - There is a new format specifier, %m, for the printf family of
1937 functions. It functions like %s, except that it does not stop when
1938 a nul ('\0') byte is encountered. The number of bytes output is
1939 determined by the precision specifier. The default precision is 1.
1942 printf ("%m", "My String") // prints one character: M
1943 printf ("%.5", myString) // prints 5 bytes beginning at the start
1946 - The %b format specifier for the printf family of functions has been enhanced
1949 1) When the width and precision are both unspecified, the default is %8.8b.
1950 2) When only one of the width or precision is specified, the other defaults
1951 to the same value. For example, %4b == %.4b == %4.4b
1952 3) Nul ('\0') bytes are used for field width padding. For example,
1954 printf ("%b", 0x1111deadbeef2222) // prints all eight bytes
1955 printf ("%4.2b", 0xdeadbeef) // prints \0\0\xbe\xef
1957 - Dynamic width and precision are now supported for all printf family format
1958 specifiers. For example:
1962 printf ("%*.*b", four, two, 0xdeadbbeef) // prints \0\0\xbe\xef
1963 printf ("%*d", four, two) // prints <space><space><space>2
1965 - Preprocessor conditional expressions can now include wildcard style
1966 matches on kernel versions.
1967 %( kernel_vr != "*xen" %? foo %: bar %)
1969 - Prototype support for user-space probing is showing some progress.
1970 No symbolic notations are supported yet (so no probing by function names,
1971 file names, process names, and no access to $context variables), but at
1972 least it's something:
1974 probe process(PID).statement(ADDRESS).absolute { }
1976 This will set a uprobe on the given process-id and given virtual address.
1977 The proble handler runs in kernel-space as usual, and can generally use
1978 existing tapset functions.
1980 - Crash utility can retrieve systemtap's relay buffer from a kernel dump
1981 image by using staplog which is a crash extension module. To use this
1982 feature, type commands as below from crash(8)'s command line:
1984 crash> extend staplog.so
1985 crash> help systemtaplog
1987 Then, you can see more precise help message.
1989 - You can share a relay buffer amoung several scripts and merge outputs from
1990 several scripts by using "-DRELAY_HOST" and "-DRELAY_GUEST" options.
1994 % stap -ve 'probe begin{}' -o merged.out -DRELAY_HOST &
1995 # wait until starting the host.
1996 % stap -ve 'probe begin{print("hello ");exit()}' -DRELAY_GUEST
1997 % stap -ve 'probe begin{print("world\n");exit()}' -DRELAY_GUEST
1999 Then, you'll see "hello world" in merged.out.
2001 - You can add a conditional statement for each probe point or aliase, which
2002 is evaluated when the probe point is hit. If the condition is false, the
2003 whole probe body(including aliases) is skipped. For example:
2006 probe syscall.* if (switch) { ... }
2007 probe procfs.write {switch = strtol($value,10)} /* enable/disable ctrl */
2009 - Systemtap will warn you if your script contains unused variables or
2010 functions. This is helpful in case of misspelled variables. If it
2011 doth protest too much, turn it off with "stap -w ...".
2013 - You can add error-handling probes to a script, which are run if a
2014 script was stopped due to errors. In such a case, "end" probes are
2015 not run, but "error" ones are.
2017 probe error { println ("oops, errors encountered; here's a report anyway")
2018 foreach (coin in mint) { println (coin) } }
2020 - In a related twist, one may list probe points in order of preference,
2021 and mark any of them as "sufficient" beyond just "optional". Probe
2022 point sequence expansion stops if a sufficient-marked probe point has a hit.
2023 This is useful for probes on functions that may be in a module (CONFIG_FOO=m)
2024 or may have been compiled into the kernel (CONFIG_FOO=y), but we don't know
2027 probe module("sd").function("sd_init_command") ? ,
2028 kernel.function("sd_init_command") ? { ... }
2030 which might match neither, now one can write this:
2032 probe module("sd").function("sd_init_command") ! , /* <-- note excl. mark */
2033 kernel.function("sd_init_command") { ... }
2035 - New security model. To install a systemtap kernel module, a user
2036 must be one of the following: the root user; a member of the
2037 'stapdev' group; or a member of the 'stapusr' group. Members of the
2038 stapusr group can only use modules located in the
2039 /lib/modules/VERSION/systemtap directory (where VERSION is the
2040 output of "uname -r").
2042 - .statement("...@file:line") probes now apply heuristics to allow an
2043 approximate match for the line number. This works similarly to gdb,
2044 where a breakpoint placed on an empty source line is automatically
2045 moved to the next statement. A silly bug that made many $target
2046 variables inaccessible to .statement() probes was also fixed.
2048 - LKET has been retired. Please let us know on <systemtap@sourceware.org>
2049 if you have been a user of the tapset/tools, so we can help you find
2052 - New families of printing functions println() and printd() have been added.
2053 println() is like print() but adds a newline at the end;
2054 printd() is like a sequence of print()s, with a specified field delimiter.
2056 * What's new since version 0.5.14?, 2007-07-03
2058 - The way in which command line arguments for scripts are substituted has
2059 changed. Previously, $1 etc. would interpret the corresponding command
2060 line argument as an numeric literal, and @1 as a string literal. Now,
2061 the command line arguments are pasted uninterpreted wherever $1 etc.
2062 appears at the beginning of a token. @1 is similar, but is quoted as
2063 a string. This change does not modify old scripts, but has the effect
2064 of permitting substitution of arbitrary token sequences.
2066 # This worked before, and still does:
2067 % stap -e 'probe timer.s($1) {}' 5
2068 # Now this also works:
2069 % stap -e 'probe syscall.$1 {log(@1)}' open
2070 # This won't crash, just signal a recursion error:
2072 # As before, $1... is recognized only at the beginning of a token
2073 % stap -e 'probe begin {foo$1=5}'
2075 * What's new since version 0.5.13?, 2007-03-26
2077 - The way in which systemtap resolves function/inline probes has changed:
2078 .function(...) - now refers to all functions, inlined or not
2079 .inline(...) - is deprecated, use instead:
2080 .function(...).inline - filters function() to only inlined instances
2081 .function(...).call - filters function() to only non-inlined instances
2082 .function(...).return - as before, but now pairs best with .function().call
2083 .statement() is unchanged.
2085 * What's new since version 0.5.12?, 2007-01-01
2087 - When running in -p4 (compile-only) mode, the compiled .ko file name
2088 is printed on standard output.
2090 - An array element with a null value such as zero or an empty string
2091 is now preserved, and will show up in a "foreach" loop or "in" test.
2092 To delete such an element, the scripts needs to use an explicit
2093 "delete array[idx]" statement rather than something like "array[idx]=0".
2095 - The new "-P" option controls whether prologue searching heuristics
2096 will be activated for function probes. This was needed to get correct
2097 debugging information (dwarf location list) data for $target variables.
2098 Modern compilers (gcc 4.1+) tend not to need this heuristic, so it is
2099 no longer default. A new configure flag (--enable-prologues) restores
2100 it as a default setting, and is appropriate for older compilers (gcc 3.*).
2102 - Each systemtap module prints a one-line message to the kernel informational
2103 log when it starts. This line identifies the translator version, base
2104 address of the probe module, a broken-down memory consumption estimate, and
2105 the total number of probes. This is meant as a debugging / auditing aid.
2107 - Begin/end probes are run with interrupts enabled (but with
2108 preemption disabled). This will allow begin/end probes to be
2109 longer, to support generating longer reports.
2111 - The numeric forms of kernel.statement() and kernel.function() probe points
2112 are now interpreted as relocatable values - treated as relative to the
2113 _stext symbol in that kernel binary. Since some modern kernel images
2114 are relocated to a different virtual address at startup, such addresses
2115 may shift up or down when actually inserted into a running kernel.
2117 kernel.statement(0xdeadbeef): validated, interpreted relative to _stext,
2118 may map to 0xceadbeef at run time.
2120 In order to specify unrelocated addresses, use the new ".absolute"
2121 probe point suffix for such numeric addresses. These are only
2122 allowed in guru mode, and provide access to no $target variables.
2123 They don't use debugging information at all, actually.
2125 kernel.statement(0xfeedface).absolute: raw, unvalidated, guru mode only
2127 * What's new since version 0.5.10?, 2006-10-19
2129 - Offline processing of debugging information, enabling general
2130 cross-compilation of probe scripts to remote hosts, without
2131 requiring identical module/memory layout. This slows down
2132 compilation/translation somewhat.
2134 - Kernel symbol table data is loaded by staprun at startup time
2135 rather than compiled into the module.
2137 - Support the "limit" keyword for foreach iterations:
2138 foreach ([x,y] in ary limit 5) { ... }
2139 This implicitly exits after the fifth iteration. It also enables
2140 more efficient key/value sorting.
2142 - Support the "maxactive" keyword for return probes:
2143 probe kernel.function("sdfsdf").maxactive(848) { ... }
2144 This allows up to 848 concurrently outstanding entries to
2145 the sdfsdf function before one returns. The default maxactive
2146 number is smaller, and can result in missed return probes.
2148 - Support accessing of saved function arguments from within
2149 return probes. These values are saved by a synthesized
2150 function-entry probe.
2152 - Add substantial version/architecture checking in compiled probes to
2153 assert correct installation of debugging information and correct
2154 execution on a compatible kernel.
2156 - Add probe-time checking for sufficient free stack space when probe
2157 handlers are invoked, as a safety improvement.
2159 - Add an optional numeric parameter for begin/end probe specifications,
2160 to order their execution.
2161 probe begin(10) { } /* comes after */ probe begin(-10) {}
2163 - Add an optional array size declaration, which is handy for very small
2165 global little[5], big[20000]
2167 - Include some example scripts along with the documentation.
2169 - Change the start-time allocation of probe memory to avoid causing OOM
2170 situations, and to abort cleanly if free kernel memory is short.
2172 - Automatically use the kernel DWARF unwinder, if present, for stack
2175 - Many minor bug fixes, performance, tapset, and error message