1 * What's new in version 4.0
3 - Runtime/tapsets were ported to include up to kernel version PRERELEASE.
5 - When a systemtap module is loaded, the name of the original stap script
6 is now printed to dmesg by the kernel runtime.
8 - Embedded-C functions marked /* guru */ may now be invoked from other
9 tapset probes / functions, while still being invalid for normal call
10 from an unprivileged user script.
12 - The syscall tapset now includes an automatic fallback alias to the
13 sys_enter / sys_exit kernel tracepoints, if no other kprobe-based
16 - New script language operators @kderef/@uderef and @kregister/@uregister
18 @kderef/@uderef (size,address) can be used to dereference integers and
19 @kregister/@uregister (dwarf#) can be used to access register values.
21 * What's new in version 3.3, 2018-06-08
23 - A new "stap --example FOO.stp" mode searches the example scripts
24 distributed with systemtap for a file named FOO.stp, so its whole
25 path does not need to be typed in.
27 - Systemtap's runtime has learned to deal with several of the
28 collateral damage from kernel hardening after meltdown/spectre,
29 including more pointer hiding and relocation. The kptr_restrict
30 procfs flag is forced on if running on a new enough kernel.
32 - The "stap --sysroot /PATH" option has received a revamp, so it
33 works much better against cross-compiled environments.
35 - The eBPF backend has learned to perform loops - at least in the
36 userspace "begin/end" probe contexts, so one can iterate across BPF
37 arrays for reporting. (The linux kernel eBPF interpreter precludes
38 loops and string processing.). It can also handle much larger probe
39 handler bodies, with a smarter register spiller/allocator.
41 - The eBPF backend now supports uprobes, perf counter, timer, and
44 - An rpm macro %_systemtap_tapsetdir is now defined, to make it
45 easier for third party packages to add .stp files into the standard
48 - Several low level locking-related fixes were added to the runtime
49 that used uprobes/tracepoint apis, in order to work more reliably on
50 rt kernels and on high-cpu-count machines.
52 - Runtime/tapsets were ported to include up to kernel version 4.16.
53 (The syscall tapsets are broken on kernel 4.17-rc, and will be fixed
54 in a next release coming soon; PR23160.)
56 * What's new in version 3.2, 2017-10-18
58 - SystemTap now includes an extended Berkeley Packet Filter (eBPF)
59 backend. This experimental backend does not use kernel modules
60 and instead produces eBPF programs that are verified by the kernel
61 and executed by an in-kernel virtual machine. Select this backend
62 with the new stap option '--runtime=bpf'. For example:
64 stap --runtime=bpf -e \
65 'probe kernel.function("sys_open") { printf("hi from stapbpf!\n") }'
67 Please see the stapbpf(8) man page for more information.
69 - The regular expression engine now supports extraction of the matched
70 string and subexpressions using the matched() tapset function:
72 if ("regexculpicator" =~ "reg(ex.*p).*r") log(matched(1))
75 - The translator produces better diagnostics for common/annoying case
76 of missing debuginfo that blocks use of context $variables.
78 - "stap -k" build trees in $TMPDIR now also include a preprocessed .i form
79 of the generated module .c code, for problem diagnostics purposes.
81 - Runtime/tapsets were ported to include up to kernel version PRERELEASE.
83 - The syscall.execve probes now provide a decoded env_str string vector,
84 just like the argument vector. Because of this, the unused
85 __count_envp() and __count_compat_evenp() functions have been
88 - The task_exe_file() Function has been deprecated and replaced by the
89 current_exe_file() function.
91 - A new probe alias input.char allows scripts to access input from stdin
94 * What's new in version 3.1, 2017-02-17
96 - Systemtap now needs C++11 to build.
98 - Syscall and nd_syscall tapsets have been merged in a way that either
99 dwarf-based, or non-dwarf probe gets automatically used based on
100 debuginfo availability (e.g. probe syscall.open).
102 To force use the dwarf based probe, a dw_syscall has been introduced
103 (e.g. probe dw_syscall.open) and the non-dwarf syscall probes were
104 left untouched (e.g. nd_syscall.open).
106 - The syscall tapset files have been reorganized in a way that original
107 big tapset files carrying many syscall probes were split into smaller
108 'sysc_' prefixed tapset files. This should reduce the syscall tapset
111 - The powerpc variant of syscall.compat_sysctl got deprecated on favor of
112 syscall.sysctl32. This aligns the syscall to its respective nd_syscall and
113 to ia64/s390/x86_64 variants too.
115 - The syscall.compat_pselect7a (this was actually a typo, but still available
116 for compatibility purposes with --compatible 1.3) has beed deprecated.
118 - The 'description_auddr' convenience variable of syscall.add_key has been
121 - Support has been added for probing python 2 and 3 functions using a
122 custom python helper module. Python function probes can target
123 function entry, returns, or specific line numbers.
125 probe python2.module("myscript").function("foo")
128 To run with the custom python helper module, you'd use python's '-m'
129 option like the following:
131 stap myscript.stp -c "python -m HelperSDT myscript.py"
133 - Java method probes now convert all types of java parameters to
134 strings using the java toString() method before passing them to
135 systemtap probes; new argN variables copy them into string
136 variables. Previously, only numeric types were passed, and only by
137 casting to integers. The previous behaviour is available with
140 3.1: probe java(...).class(...).method(...) { printf("%s", arg1) }
141 3.0: probe java(...).class(...).method(...) { printf("%d", $arg1) }
143 - An older defensive measure to suppress kernel kprobes optimizations
144 since the 3.x era has been disabled for recent kernels. This improves
145 the performance of kernel function probes. In case of related problems,
146 please report and work around with:
147 # echo 0 > /proc/sys/debug/kprobes-optimization
149 - Context variables in .return probes should be accessed with @entry($var)
150 rather than $var, to make it clear that entry-time snapshots are being
151 used. The latter construct now generates a warning. Availability testing
152 with either @defined(@entry($var)) or @defined($var) works.
154 - Tapsets containing process probes may now be placed in the special
155 $prefix/share/systemtap/tapset/PATH/ directory to have their process parameter
156 prefixed with the location of the tapset. For example,
158 process("foo").function("NAME") expands to process("/usr/bin/foo").function("NAME")
159 when placed in $prefix/share/systemtap/tapset/PATH/usr/bin/
161 This is intended to help write more reusable tapsets for userspace binaries.
163 - The implementation of "var <<< X" for each aggregate variable is now
164 specially compiled to compute only the script-requested @op(var) values,
165 not all potential ones. This speeds up the <<< operations.
167 - Systemtap now warns if script arguments given on the command line are unused,
168 instead of mentioned by the script with $n/@n.
170 - Netfilter tapsets now provide variables data_hex and data_str to display packet
171 contents in hexadecimal and ASCII respectively.
173 - Translator now accepts new @const() operator for convenient expressing
174 constants in tapset code, or guru-mode scripts. See stap(1) for details.
176 - New -T option allows the script to be terminated after a specified number
177 of seconds. This is a shortcut for adding the probe, timer {exit()}.
179 - New installcheck-parallel testsuite feature allows running the tests in
180 parallel in order to save time. See testsuite/README for details.
182 - New tapset functions set_user_string(), set_user_string_n(), set_user_long()
183 set_user_int(), set_user_short(), set_user_char() and set_user_pointer() to
184 write a value of specified type directly to a user space address.
186 - New tapset functions user_buffer_quoted(), user_buffer_quoted_error(),
187 kernel_buffer_quoted(), and kernel_buffer_quoted_error() to print a
188 buffer of an exact length. These functions can handle '\0' characters
191 - New statistics @variance() operator using the Welford's online algorithm
192 for per-cpu computation, and the Total Variance formula authored by
193 Niranjan Kamat and Arnab Nandi from the Ohio State University for the
194 cross-cpu aggregation.
196 - New command within interactive mode, sample. Allows you to search through
197 all included example scripts to load for further editing or running. Sample
198 and example scripts have been moved to /usr/share/systemtap/examples. A
199 symlink in the former location under $docdir links to it.
201 * What's new in version 3.0, 2016-03-27
203 - The new experimental "interactive" mode, specified by "stap -i",
204 drops you into a command-line prompt where you can build up a script,
205 run it, edit it, run it again, etc. Type "help" for a list of commands.
207 - New experimental --monitor[=INTERVAL] option similar to unix "top". This
208 allows users to see statistics about the running module(uptime, module name,
209 invoker uid, memory sizes, global variables, and the current probe list
210 along with their statistics).
212 An interface is also provided to allow control over the running
213 module(resetting global variables, sorting the list of probes, deactivating
214 and reactivating probes).
216 - The performance of associative arrays have been dramatically
217 improved, especially for densely filled tables and for multiple
218 indexes. The hash tables behind these arrays is now sized as a
219 function of the array maximum size with an optional MAPHASHBIAS
220 space/time tradeoff knob.
222 - Add macros @prints to print a scalar aggregate variable, @prints[1-9]
223 to print an array aggregate (of given index-arity), formatted similarly
224 to the automatic printing of written-only global variables.
227 probe oneshot { a <<< 1; b[tid()] <<< 2 }
228 probe end { @prints(a); @prints1(b) }
230 - Functions may now be overloaded during module runtime using the "next"
231 statement in script functions and STAP_NEXT macro for embedded-C functions.
232 They may also be overloaded by number of parameters during compile time.
236 function f() { if (condition) next; print("first function") }
237 function f() %{ STAP_NEXT; print("second function") %}
238 function f() { print("third function") }
240 For the given functions above, a functioncall f(), will execute the body of the
241 third function if condition evaluates to true and print "third function".
242 Note that the second function is unconditionally nexted.
244 Parameter overloading:
245 function g() { print("first function") }
246 function g(x) { print("second function") }
247 g() -> "first function"
248 g(1) -> "second function"
250 Note that runtime overloading does not occur in the above example as the number
251 of parameters of the functions differ. The use of a next statement inside a function
252 while no more overloads remain will trigger a runtime exception. The function
253 candidates are selected at compile time and is determined by the number of arguments
254 provided for the functioncall.
256 - Add Czech version of manual pages.
258 - The stap compile server will log the stap client's options that are passed
259 to the server. The options that get logged on the server will include the
260 script name or the -e script, depending on which is used by the client.
262 - Embedded-C functions and blocks may now access script level global variables
263 using the STAP_GLOBAL_GET_* and STAP_GLOBAL_SET_* macros.
265 To read or write the script global var, the /* pragma:read:var */ or
266 /* pragma:write:var */ marker must be placed in the embedded-C function or block.
267 The written type must match the type inferred at script level.
270 STAP_GLOBAL_SET_var(STAP_GLOBAL_GET_var()+1) -> increments script global var by 1
271 STAP_GLOBAL_SET_var("hello")
274 STAP_GLOBAL_GET_var(index-1, ..., index-n)
275 STAP_GLOBAL_SET_var(index-1, ..., index-n, new value)
277 - Probe point brace expansion is now supported to improve brevity in
278 specifying probe points. For example,
280 process.{function("a"), function("b").{call,return}}
281 => process.function("a"), process.function("b").call, process.function("b").return
283 process.{function("*").callees,plt}?
284 => process.function("*").callees?, process.plt?
286 {kernel,module("nfs")}.function("nfs*")!
287 => kernel.function("nfs*")!, module("nfs").function("nfs*")!
289 - Profiling timers at arbitrary frequencies are now provided and perf probes
290 now support a frequency field as an alternative to sampling counts.
292 probe timer.profile.freq.hz(N)
293 probe perf.type(N).config(M).hz(X)
295 The specified frequency is only accurate up to around 100hz. You may
296 need to provide a higher value to achieve the desired rate.
298 - Added support for private global variables and private functions. The scope
299 of these is limited to the tapset file they are defined in (PR19136).
301 - New tapset function string_quoted() to quote and \-escape general strings.
302 String $context variables that are pretty-printed are now processed with
303 such a quotation engine, falling back to a 0x%x (hex pointer) on errors.
305 - Functions get_mmap_args() and get_32mmap_args() got deprecated.
307 * What's new in version 2.9, 2015-10-08
309 - SystemTap now uses symbols from /proc/kallsyms when kernel debuginfo is not
312 - New --prologue-searching[=WHEN] option has been added to stap with '-P' being
313 its short counterpart. Using --prologue-searching=never turns prologue
314 searching deliberately off working around issue of int_arg() returning wrong
315 value when a 32-bit userspace binary having debug info is being probed with
316 active prologue searching (PR18649).
318 - The powerpc variant of nd_syscall.compat_sysctl got deprecated on favor of
319 nd_syscall.sysctl32. This aligns the nd_syscall to its respective syscall and
320 to ia64/s390/x86_64 variants too.
322 - New tapset function assert(expression, msg) has been added.
324 - Embedded-C functions may now use the new STAP_PRINTF(fmt, ...)
327 - New tapset functions fullname_struct_path and fullname_struct_nameidata
328 resolve full path names from internal kernel struct pointers.
330 - New tapset functions arch_bytes() and uarch_bytes() to obtain address size
331 for kernel and user space respectively.
333 - New tapset function switch_file() allows control over rotation
336 - The [nd_]syscall tapset got autodocumented. Related paragraph got added to PDF
337 and HTML tapset reference. Also a new tapset::syscall 3stap man page got added.
339 - Embedded-C functions with parameter arity-0 can now be marked with
340 the /* stable */ /* pure */ pragmas, if (roughly speaking) the
341 function is side-effect-free and idempotent. The translator may
342 execute these speculatively and have their results memoized. This
343 lets probes with multiple calls to such functions run faster.
345 Context variable ($foo) getter functions (in non-guru mode), and
346 numerous tapset functions are now marked as /* stable */ /* pure */.
347 Several example scripts have been modified to eschew explicit
350 - Callee probe points now support '.return' and '.call' suffix.
353 process("proc").function("foo").callee("bar").return
355 will fire upon returning from bar when called by foo.
357 process("proc").function("foo").callee("bar").call
359 will only fire for non-inlined callees.
361 - The following tapset variables and functions are deprecated in
363 - The '__int32_compat' library macro got deprecated in favor of
364 new '__compat_long' library macro.
365 - The 'uargs' convenience variable of the 'seccomp' syscall probe
366 got deprecated in favor of new 'uargs_uaddr' variable.
368 - SystemTap has reduced its memory consumption by using interned_strings (a
369 wrapper for boost::string_ref) in place of std::string instances. The change
370 is to reduce the number of duplicate strings created by replacing them with
371 interned_strings which act like pointers to existing strings.
373 For the implementation of interned_string, see stringtable.h
375 * What's new in version 2.8, 2015-06-17
377 - SystemTap has improved support for probing golang programs. Work has been
378 done to be able to handle DWARF information, reporting file names, line
379 numbers, and column numbers, and tolerance of odd characters in symbol names.
381 - The function::*, probe::* and new macro::* man pages cross-references the
382 enclosing tapset::* man page. For example:
384 function::pn(3stap) mentions tapset::pn(3stap) in the SEE ALSO section
386 - New stapref(1) man page provides a reference for the scripting language. The
387 stapref page contains an overview of the features available in the language,
388 such as keywords, data types, operators and more.
390 - The @task macro performs the very common @cast to a task_struct.
392 The embedded-C bodies of task_current() and pid2task() are now wrapped
393 by @task, which gives them a debuginfo type on the return value. With
394 autocast type propagation, this removes the need for any explicit @cast
397 Other places which take untyped task pointers as parameters, for
398 instance, now use @task as well to simplify their code.
400 - New namespace-aware tapset functions [task_]ns_*() and ia new option
401 --target-namespaces=PID to denote a target set of namespaces corresponding to
402 the PID's namespaces. The namespace-aware tapsets will return values
403 relative to the target namespaces if specified, or the stap process' namespaces.
405 - Netfilter probes now attempt to decode Spanning Tree Protocol packets
406 into local variables: probe netfilter.bridge.*, br_* variables,
407 stp_dump.stp sample script.
409 - Colorization of error string tokens is made more robust, especially
410 in presence of $N/@N substitution.
412 - The following tapset variables and functions are deprecated in
414 - The 'hostname_uaddr' variable in the syscall.setdomainname and
415 nd_syscall.setdomainname probe aliases have been deprecated in
416 favor of the new 'domainname_uaddr' variable.
417 - The 'fd' and 'fd_str' variables in the syscall.execveat and
418 nd_syscall.execveat probe aliases have been deprecated in favor of
419 the new 'dirfd' and 'dirfd_str' variables.
421 * What's new in version 2.7, 2015-02-18
423 - Some systemtap sample scripts are now identified with the "_best" keyword,
424 because they are generally useful or educational. They are now promoted
425 within the generated index files.
427 - Passing strings to and from functions has become faster due to optimization
428 (passing some strings by reference instead of by value/copy). It may
429 be disabled by using the unoptimize flag (-u).
431 To make embedded-C functions eligible for the same optimization, use the pragma
432 /* unmodified-fnargs */ to indicate that the function body will not modify
433 the function arguments. Remember to use MAXSTRINGLEN for string length,
434 rather than sizeof(string_arg) (which might now be a pointer).
436 - SystemTap now allows .function probes to be specified by their full function
437 name, file, and declaration line number. Use the .statement probe to probe a
438 specific line number.
440 - Tracepoint probes can now also be specified by the target subsystem. For
441 example, the following are all supported:
443 probe kernel.trace("sched:sched_switch") --> probe sched_switch found in the
445 probe kernel.trace("sched:*") --> probe all tracepoints in sched subsystem
447 As a result, tapset functions such as pn() will now return a different string
448 than before. To retain the previous behaviour, use '--compatible=2.6'.
450 - The following functions are deprecated in release 2.7:
451 - _adjtx_mode_str(), _statfs_f_type_str(), _waitid_opt_str(),
452 _internal_wait_opt_str(), and _epoll_events_str().
454 - New tapset functions [u]symfileline(), [u]symfile() and [u]symline() will
455 return a string containing the specified portion of the filename:linenumber
456 match from a given address.
458 Using these functions may result in large generated modules from stored
459 address->file:line information.
461 * What's new in version 2.6, 2014-09-05
463 - SystemTap now supports on-the-fly arming/disarming of certain probe types:
464 kprobes, uprobes, and timer.*s(NUM) probes. For example, this probe
466 probe kernel.function("vfs_read") if (i > 4) { ... }
468 will automatically register/unregister the associated kprobe on vfs_read
469 whenever the value of the condition changes (as some probe handler
470 modifies 'i'). This allows us to avoid probe overhead when we're not
471 interested. If the arming capability is not relevant/useful, nest the
472 condition in the normal probe handler:
474 probe kernel.function("vfs_read") { if (i > 4) { ... } }
476 - statement("*@file:NNN").nearest probes now available to let systemtap
477 translate probe to nearest probe-able line to one given if necessary
479 - process("PATH").library("PATH").plt("NAME").return probes are now supported.
481 - SystemTap now supports SDT probes with operands that refer to symbols.
483 - While in listing mode (-l/-L), probes printed are now more consistent
486 - Statement probes now support enumerated linenos to probe discontiguous
487 linenos using the form:
489 process.statement("foo@file.c:3,5-7,9")
491 - Statement counting is now suppressed in the generated c code for probes that
492 are non-recursive and loop-free. Statement counting can be turned back on in
493 unoptimize mode (-u).
495 - SystemTap now asserts that the PID provided for a process probe corresponds
496 to a running process.
498 - DWARF process probes can be bound to a specific process using the form:
500 process(PID).function("*")
502 - SystemTap now accepts additional scripts through the new -E SCRIPT option.
503 There still needs to be a main script specified through -e or file in order
504 to provide an additional script. This makes it feasible to have scripts in
505 the $HOME/.systemtap/rc file. For example:
507 -E 'probe begin, end, error { log("systemtap script " . pn()) }'
508 -E 'probe timer.s(30) { error ("timeout") }
510 The -E SCRIPT option can also be used in listing mode (-l/-L), such that
511 probe points for the additional scripts will not listed, but other parts of
512 the script are still available, such as macros or aliases.
514 - SystemTap now supports array slicing within foreach loop conditions, delete
515 statements and membership tests. Wildcards are represented by "*". Examples
516 of the expressions are:
518 foreach ([a,b,c] in val[*,2,*])
522 - Integer expressions which are derived from DWARF values, like context $vars,
523 @cast, and @var, will now carry that type information into subsequent reads.
524 Such expressions can now use "->" and "[]" operators, as can local variables
525 which were assigned such values.
527 foo = $param->foo; printf("x:%d y:%d\n", foo->x, foo->y)
528 printf("my value is %d\n", ($type == 42 ? $foo : $bar)->value)
529 printf("my parent pid is %d\n", task_parent(task_current())->tgid)
531 * What's new in version 2.5, 2014-04-30
533 - Systemtap now supports backtracing through its own, invoking module.
535 - Java probes now support backtracing using the print_java_backtrace()
536 and sprint_java_backtrace() functions.
538 - Statement probes (e.g. process.statement) are now faster to resolve,
539 more precise, and work better with inlined functions.
541 - New switches have been added to help inspect the contents of installed
544 stap --dump-functions --> list all library functions and their args
545 stap --dump-probe-aliases --> list all library probe aliases
547 - The heuristic algorithms used to search for function-prologue
548 endings were improved, to cover more optimization (or
549 lack-of-optimization, or incorrect-debuginfo) cases. These
550 heuristics are necessary to find $context parameters for some
551 function-call/entry probes. We recommend programs be built with
552 CFLAGS+=-grecord-gcc-switches to feed information to the heuristics.
554 - The stap --use-server option now more correctly supports address:port
555 type parametrization, for manual use in the absence of avahi.
557 - A new probe alias "oneshot" allows a single quick script fragment to run,
560 - The argv tapset now merges translate-time and run-time positional
561 arguments, so all of these work:
563 stap -e 'probe oneshot {println(argv[1]," ",argv[2])}' hello world
565 stap -e 'probe oneshot {println(argv[1]," ",argv[2])}' \
566 -G argv_1=hello -G argv_2=world
568 staprun hello.ko argv_1=hello argv_2=world
570 - SystemTap now falls back on the symbol table for probing
571 functions in processes if the debuginfo is not available.
573 - SystemTap now supports a %( guru_mode == 0 /* or 1 */ %)
574 conditional for making dual-use scripts.
576 - SystemTap now supports UEFI/SecureBoot systems, via
577 machine-owner-keys maintained by a trusted stap-server on the
578 network. (Key enrollment requires a one-time reboot and BIOS
580 https://sourceware.org/systemtap/wiki/SecureBoot
582 - SystemTap now reports more accurate and succinct errors on type
585 - Embedded-C functions may use STAP_RETURN(value) instead of the
586 more wordy STAP_RETVALUE assignment followed by a "goto out".
587 The macro supports numeric or string values as appropriate.
588 STAP_ERROR(...) is available to return with a (catchable) error.
590 - Some struct-sockaddr fields are now individually decoded for
591 socket-related syscalls:
592 probe syscall.connect { println (uaddr_af, ":", uaddr_ip) }
594 - The documentation for the SystemTap initscript service and the
595 SystemTap compile-server service have been completely converted from
596 README files to man pages (see systemtap(8) and stap-server(8)).
598 - SystemTap is now capable of inserting modules early during the boot
599 process on dracut-based systems. See the 'onboot' command in
600 systemtap(8) for more information.
602 - DWARF probes can now use the '.callee[s]' variants, which allow more
603 precise function probing. For example, the probe point
605 process("myproc").function("foo").callee("bar")
607 will fire upon entering bar() from foo(). A '.callees' probe will
608 instead place probes on all callees of foo().
609 Note that this also means that probe point wildcards should be used
610 with more care. For example, use signal.*.return rather than
611 signal.*.*, which would also match '.callees'. See stapprobes(3stap)
612 for more info. This feature requires at least GCC 4.7.
614 - A few new functions in the task_time tapsets, as well as a new tapset
615 function task_ancestry(), which prints out the parentage of a process.
617 - The kprocess.exec probe has been updated to use syscall.execve, which
618 allows access to the new process' arguments (through the new 'argstr'
619 or 'args' variables) as well as giving better support across kernel
620 versions. Note also that the 'filename' variable now holds the
621 filename (quoted), or the address (unquoted) if it couldn't be
624 - The [s]println() function can now be called without any arguments to
625 simply print a newline.
627 - Suggestions are now provided when markers could not be resolved. For
628 example, process("stap").mark("benchmart") will suggest 'benchmark'.
630 - SystemTap colors can now be turned off by simply setting
631 SYSTEMTAP_COLORS to be empty, rather than having to make it invalid.
633 - There is a new context tapset function, pnlabel(), which returns the
634 name of the label which fired.
636 - The following tapset variables and functions are deprecated in
638 - The 'clone_flags', 'stack_start', 'stack_size',
639 'parent_tid_uaddr', and 'child_tid_uaddr' variables in the
640 'syscall.fork' and 'nd_syscall.fork' probe aliases.
641 - The '_sendflags_str()' and '_recvflags_str()' functions have been
642 deprecated in favor of the new '_msg_flags_str()' function.
643 - The 'flags' and 'flags_str' variables in the 'syscall.accept' and
644 'nd_syscall.accept' probe alias.
645 - The 'first', 'second', and 'uptr_uaddr' variables in the
646 'syscall.compat_sys_shmctl', and 'nd_syscall.compat_sys_shmctl'
647 probe aliases have been deprecated in favor of the new 'shmid',
648 'cmd', and 'buf_uaddr' variables.
650 * What's new in version 2.4, 2013-11-06
652 - Better suggestions are given in many of the semantic errors in which
653 alternatives are provided. Additionally, suggestions are now provided
654 when plt and trace probes could not be resolved. For example,
655 kernel.trace("sched_siwtch") will suggest 'sched_switch'.
657 - SystemTap is now smarter about error reporting. Errors from the same
658 source are considered duplicates and suppressed. A message is
659 displayed on exit if any errors/warnings were suppressed.
661 - Statistics aggregate typed objects are now implemented locklessly,
662 if the translator finds that they are only ever read (using the
663 foreach / @count / etc. constructs) in a probe-begin/end/error.
665 - SystemTap now supports probing inside virtual machines using the
666 libvirt and unix schemes, e.g.
668 stap -ve 'probe timer.s(1) { printf("hello!\n") }' \
669 --remote=libvirt://MyVirtualMachine
671 Virtual machines managed by libvirt can be prepared using stapvirt.
672 See stapvirt(1) and the --remote option in stap(1) for more details.
674 - Systemtap now checks for and uses (when available) the .gnu_debugdata
675 section which contains a subset of debuginfo, useful for backtraces
678 - SystemTap map variables are now allocated with vmalloc() instead of
679 with kmalloc(), which should cause memory to be less fragmented.
681 - Although SystemTap itself requires elfutils 0.148+, staprun only
682 requires elfutils 0.142+, which could be useful with the
683 '--disable-translator' configure switch.
685 - Under FIPS mode (/proc/sys/crypto/fips_enabled=1), staprun will
686 refuse to load systemtap modules (since these are not normally
687 signed with the kernel's build-time keys). This protection may
688 be suppressed with the $STAP_FIPS_OVERRIDE environment variable.
690 - The stap-server client & server code now enable all SSL/TLS
691 ciphers rather than just the "export" subset.
693 - For systems with in-kernel utrace, 'process.end' and 'thread.end'
694 probes will hit before the target's parent process is notified of
695 the target's death. This matches the behavior of newer kernels
696 without in-kernel utrace.
698 * What's new in version 2.3, 2013-07-25
700 - More context-accessing functions throw systemtap exceptions upon a
701 failure, whereas in previous versions they might return non-error
702 sentinel values like "" or "<unknown>". Use try { } / catch { }
703 around these, or new wrapper functions such as user_string_{n_}quoted()
704 that internally absorb exceptions.
706 - java("org.my.MyApp") probes are now restricted to pre-existing jvm pid's with
707 a listing in jps -l output to avoid recursive calls
709 - The tapset [nd_]syscall.semop parameter tsops_uaddr is renamed sops_uaddr for
710 consistency with [nd_]syscall.semtimedop.
712 - The udp.stp tapset adds some ip-address/port variables.
714 - A new guru-mode-only tapset function raise() is available to send signals
717 - Support for the standard Posix ERE named character classes has been
718 added to the regexp engine, e.g. [:digit:], [:alpha:], ...
720 - A substantial internal overhaul of the regexp engine has resulted in
721 correct behaviour on further obscure edge cases. The regexp engine
722 now implements the ERE standard and correctly passes the testsuite
723 for the glibc regexp engine (minus portions corresponding to
724 unimplemented features -- i.e. subexpression capture and reuse).
726 - Alternative functions are now suggested when function probes could not be
727 resolved. For example, kernel.function("vfs_reads") will suggest vfs_read.
728 Other probes for which suggestions are made are module.function,
729 process.function, and process.library.function.
731 - Has life been a bit bland lately? Want to spice things up? Why not write a
732 few faulty probes and feast your eyes upon the myriad of colours adorning
733 your terminal as SystemTap softly whispers in your ear... 'parse error'.
734 Search for '--color' in 'man stap' for more info.
736 - The following tapset functions are deprecated in release 2.3:
737 'stap_NFS_CLIENT', '__getfh_inode', '_success_check',
738 '_sock_prot_num', '_sock_fam_num', '_sock_state_num',
739 '_sock_type_num', and '_sock_flags_num'.
741 * What's new in version 2.2.1, 2013-05-16
742 * What's new in version 2.2, 2013-05-14
744 - Experimental support has been added for probing Java methods using
745 Byteman 2.0 as a backend. Java method probes can target method entries,
746 returns, or specific statements in the method as specified by line number.
748 probe java("org.my.MyApp").class("^java.lang.Object").method("foo(int)")
751 See java/README for information on how to set up Java/Byteman
752 functionality. Set env STAPBM_VERBOSE=yes for more tracing.
754 - The stap -l output and pn() tapset function's return value may be slightly
755 different for complicated web of wildcarded/aliased probes.
757 - The dyninst backend has improved in several aspects:
759 - Setting custom values for global variables is now supported, both
760 with -G when compiling a script, and from the stapdyn command line
761 when loading a precompiled module.
763 - A high-performance shared-memory-based transport is used for
766 - A systemd service file and tmpfile have been added to allow
767 systemtap-server to be managed natively by systemd.
769 - Due to the removal of register_timer_hook in recent kernels, the
770 behaviour of timer.profile has been changed slightly. This probe is
771 now an alias which uses the old mechanism where possible, but falls
772 back to perf.sw.cpu_clock or another mechanism when the kernel timer
773 hook is not available.
775 To require the kernel timer hook mechanism in your script, use
776 timer.profile.tick instead of timer.profile.
778 - The following tapset variables are deprecated in release 2.2:
779 - The 'origin' variables in the 'generic.fop.llseek',
780 'generic.fop.llseek.return', and 'nfs.fop.llseek' probes. The
781 'origin' variable has been replaced by the 'whence' variable.
782 - The 'page_index' variable in the 'vfs.block_sync_page' and
783 'vfs.buffer_migrate_page' probe aliases.
784 - The 'write_from' and 'write_upto' variables in the
785 '_vfs.block_prepare_write' and '_vfs.block_prepare_write.return'
787 - The 'regs' variable in the 'syscall.sigaltstack',
788 'nd_syscall.sigaltstack', 'syscall.fork', and 'nd_syscall.fork'
790 - The 'first', 'second', 'third', and 'uptr_uaddr' variables in the
791 'syscall.compat_sys_shmat' and 'nd_syscall.compat_sys_shmat' probe
794 - The following tapset functions are deprecated in release 2.2:
795 'ppos_pos', '_dev_minor', and '_dev_major'
797 - The folowing tapset functions used to return error strings instead
798 of raising an error. The original behavior is deprecated in release
801 'ctime', 'probemod', 'modname'
803 * What's new in version 2.1, 2013-02-13
805 - EMACS and VIM editor modes for systemtap source files are included / updated.
807 - The translator now eliminates duplicate tapset files between its
808 preferred directory (as configured during the build with --prefix=/
809 or specified with the -I /path option), and files it may find under
810 $XDG_DATA_DIRS. This should eliminate a class of conflicts between
811 parallel system- and hand-built systemtap installations.
813 - The translator accepts a --suppress-time-limits option, which defeats
814 time-related constraints, to allows probe handlers to run for indefinite
815 periods. It requires the guru mode (-g) flag to work. Add the earlier
816 --suppress-handler-errors flag for a gung-ho "just-keep-going" attitude.
818 - Perf event probes may now be read on demand. The counter probe is
819 defined using the counter-name part:
820 probe perf.type(0).config(0).counter("NAME"). The counter is
821 read in a user space probe using @perf("NAME"), e.g.
822 process("PROCESS").statement("func@file") {stat <<< @perf("NAME")}
824 - Perf event probes may now be bound to a specific task using the
825 process-name part: probe perf.type(0).config(0).process("NAME") { }
826 If the probed process name is not specified, then it is inferred
827 from the -c CMD argument.
829 - Some error messages and warnings now refer to additional information
830 that is found in man pages. These are generally named
831 error::FOO or warning::BAR (in the 7stap man page section)
836 - The dyninst backend has improved in several aspects:
837 - The runtime now allows much more concurrency when probing multithreaded
838 processes, and will also follow probes across forks.
839 - Several new probe types are now supported, including timers, function
840 return, and process.begin/end and process.thread.begin/end.
841 - Semaphores for SDT probes are now set properly.
842 - Attaching to existing processes with -x PID now works.
844 - The foreach looping construct can now sort aggregate arrays by the user's
845 choice of aggregating function. Previously, @count was implied. e.g.:
846 foreach ([x,y] in array @sum +) { println(@sum(array[x,y])) }
848 - Proof of concept support for regular expression matching has been added:
849 if ("aqqqqqb" =~ "q*b") { ... }
850 if ("abc" !~ "q*b") { ... }
852 The eventual aim is to support roughly the same functionality as
853 the POSIX Extended Regular Expressions implemented by glibc.
854 Currently missing features include extraction of the matched string
855 and subexpressions, and named character classes ([:alpha:], [:digit:], &c).
857 Special thanks go to the re2c project, whose public domain code this
858 functionality has been based on. For more info on re2c, see:
859 http://sourceforge.net/projects/re2c/
861 - The folowing tapset variables are deprecated in release 2.1 and will
862 be removed in release 2.2:
863 - The 'send2queue' variable in the 'signal.send' probe.
864 - The 'oldset_addr' and 'regs' variables in the 'signal.handle' probe.
866 - The following tapset probes are deprecated in release 2.1 and will
867 be removed in release 2.2:
869 - signal.handle.return
871 * What's new in version 2.0, 2012-10-09
873 - Systemtap includes a new prototype backend, which uses Dyninst to instrument
874 a user's own processes at runtime. This backend does not use kernel modules,
875 and does not require root privileges, but is restricted with respect to the
876 kinds of probes and other constructs that a script may use.
878 Users from source should configure --with-dyninst and install a
879 fresh dyninst snapshot such as that in Fedora rawhide. It may be
880 necessary to disable conflicting selinux checks; systemtap will advise.
882 Select this new backend with the new stap option --runtime=dyninst
883 and a -c target process, along with normal options. (-x target
884 processes are not supported in this prototype version.) For example:
886 stap --runtime=dyninst -c 'stap -l begin' \
887 -e 'probe process.function("main") { println("hi from dyninst!") }'
889 - To aid diagnosis, when a kernel panic occurs systemtap now uses
890 the panic_notifier_list facility to dump a summary of its trace
891 buffers to the serial console.
893 - The systemtap preprocessor now has a simple macro facility as follows:
895 @define add(a,b) %( ((@a)+(@b)) %)
896 @define probegin(x) %(
902 @probegin( foo = @add(40, 2); print(foo) )
904 Macros defined in the user script and regular tapset .stp files are
905 local to the file. To get around this, the tapset library can define
906 globally visible 'library macros' inside .stpm files. (A .stpm file
907 must contain a series of @define directives and nothing else.)
909 The status of the feature is experimental; semantics of macroexpansion
910 may change (unlikely) or expand in the future.
912 - Systemtap probe aliases may be used with additional suffixes
913 attached. The suffixes are passed on to the underlying probe
914 point(s) as shown below:
916 probe foo = bar, baz { }
917 probe foo.subfoo.option("gronk") { }
918 // expands to: bar.subfoo.option("gronk"), baz.subfoo.option("gronk")
920 In practical terms, this allows us to specify additional options to
921 certain tapset probe aliases, by writing e.g.
922 probe syscall.open.return.maxactive(5) { ... }
924 - To support the possibility of separate kernel and dyninst backends,
925 the tapsets have been reorganized into separate folders according to
926 backend. Thus kernel-specific tapsets are located under linux/, the
927 dyninst-specific ones under dyninst/
929 - The backtrace/unwind tapsets have been expanded to allow random
930 access to individual elements of the backtrace. (A caching mechanism
931 ensures that the backtrace computation run at most once for each
932 time a probe fires, regardless of how many times or what order the
933 query functions are called in.) New tapset functions are:
934 stack/ustack - return n'th element of backtrace
935 callers/ucallers - return first n elements of backtrace
936 print_syms/print_usyms - print full information on a list of symbols
937 sprint_syms/sprint_usyms - as above, but return info as a string
939 The following existing functions have been superseded by print_syms()
940 et al.; new scripts are recommended to avoid using them:
946 - The probefunc() tapset function is now myproc-unprivileged, and can
947 now be used in unprivileged scripts for such things as profiling in
948 userspace programs. For instance, try running
949 systemtap.examples/general/para-callgraph.stp in unprivileged mode
950 with a stapusr-permitted probe. The previous implementation of
951 probefunc() is available with "stap --compatible=1.8".
953 - Preprocessor conditional to vary code based on script privilege level:
954 unprivileged -- %( systemtap_privilege == "stapusr" %? ... %)
955 privileged -- %( systemtap_privilege != "stapusr" %? ... %)
956 or, alternately %( systemtap_privilege == "stapsys"
957 || systemtap_privilege == "stapdev" %? ... %)
959 - To ease migration to the embedded-C locals syntax introduced in 1.8
960 (namely, STAP_ARG_* and STAP_RETVALUE), the old syntax can now be
961 re-enabled on a per-function basis using the /* unmangled */ pragma:
963 function add_foo:long(a:long, b:long) %{ /* unmangled */
964 THIS->__retvalue = THIS->a + STAP_ARG_b;
967 Note that both the old and the new syntax may be used in an
968 /* unmangled */ function. Functions not marked /* unmangled */
969 can only use the new syntax.
971 - Adjacent string literals are now glued together irrespective of
972 intervening whitespace or comments:
973 "foo " "bar" --> "foo bar"
974 "foo " /* comment */ "bar" --> "foo bar"
975 Previously, the first pair of literals would be glued correctly,
976 while the second would cause a syntax error.
978 * What's new in version 1.8, 2012-06-17
980 - staprun accepts a -T timeout option to allow less frequent wake-ups
981 to poll for low-throughput output from scripts.
983 - When invoked by systemtap, the kbuild $PATH environment is sanitized
984 (prefixed with /usr/bin:/bin:) in an attempt to exclude compilers
985 other than the one the kernel was presumed built with.
987 - Printf formats can now use "%#c" to escape non-printing characters.
989 - Pretty-printed bitfields use integers and chars use escaped formatting
992 - The systemtap compile-server and client now support IPv6 networks.
993 - IPv6 addresses may now be specified on the --use-server option and will
994 be displayed by --list-servers, if the avahi-daemon service is running and
996 - Automatic server selection will automatically choose IPv4 or IPv6 servers
997 according to the normal server selection criteria when avahi-daemon is
998 running. One is not preferred over the other.
999 - The compile-server will automatically listen on IPv6 addresses, if
1001 - To enable IPv6 in avahi-daemon, ensure that /etc/avahi/avahi-daemon.conf
1002 contains an active "use-ipv6=yes" line. After adding this line run
1003 "service avahi-daemon restart" to activate IPv6 support.
1004 - See man stap(1) for details on how to use IPv6 addresses with the
1005 --use-server option.
1007 - Support for DWARF4 .debug_types sections (for executables and shared
1008 libraries compiled with recent GCC's -gdwarf-4 / -fdebug-types-section).
1009 PR12997. SystemTap now requires elfutils 0.148+, full .debug_types support
1010 depends on elfutils 0.154+.
1012 - Systemtap modules are somewhat smaller & faster to compile. Their
1013 debuginfo is now suppressed by default; use -B CONFIG_DEBUG_INFO=y to
1016 - @var now an alternative language syntax for accessing DWARF variables
1017 in uprobe and kprobe handlers (process, kernel, module). @var("somevar")
1018 can be used where $somevar can be used. The @var syntax also makes it
1019 possible to access non-local, global compile unit (CU) variables by
1020 specifying the CU source file as follows @var("somevar@some/src/file.c").
1021 This will provide the target variable value of global "somevar" as defined
1022 in the source file "some/src/file.c". The @var syntax combines with all
1023 normal features of DWARF target variables like @defined(), @entry(),
1024 [N] array indexing, field access through ->, taking the address with
1025 the & prefix and shallow or deep pretty printing with a $ or $$ suffix.
1027 - Stap now has resource limit options:
1033 All resource limiting has been moved from the compile server to stap
1034 itself. When running the server as "stap-server", default resource
1035 limit values are specified in ~stap-server/.systemtap/rc.
1037 - Bug CVE-2012-0875 (kernel panic when processing malformed DWARF unwind data)
1040 - The systemtap compile-server now supports multiple concurrent connections.
1041 Specify the desired maximum number of concurrent connections with
1042 the new stap-server/stap-serverd --max-threads option. Specify a
1043 value of '0' to tell the server not to spawn any new threads (handle
1044 all connections serially in the main thread). The default value is
1045 the number of processor cores on the host.
1047 - The following tapset functions are deprecated in release 1.8 and will be
1048 removed in release 1.9:
1051 - SystemTap now mangles local variables to avoid collisions with C
1052 headers included by tapsets. This required a change in how
1053 embedded-C functions access local parameters and the return value slot.
1055 Instead of THIS->foo in an embedded-C function, please use the newly
1056 defined macro STAP_ARG_foo (substitute the actual name of the
1057 argument for 'foo'); instead of THIS->__retvalue, use the newly
1058 defined STAP_RETVALUE. All of the tapsets and test cases have been
1059 adapted to use this new notation.
1061 If you need to run code which uses the old THIS-> notation, run stap
1062 with the --compatible=1.7 option.
1064 - There is updated support for user-space probing against kernels >=
1065 3.5, which have no utrace but do have the newer inode-uprobes work
1066 by Srikar Dronamraju and colleagues. For kernels < 3.5, the
1067 following 3 sets of kernel patches would need to be backported to
1068 your kernel to use this preliminary user-space probing support:
1070 - inode-uprobes patches:
1071 - 2b144498350860b6ee9dc57ff27a93ad488de5dc
1072 - 7b2d81d48a2d8e37efb6ce7b4d5ef58822b30d89
1073 - a5f4374a9610fd7286c2164d4e680436727eff71
1074 - 04a3d984d32e47983770d314cdb4e4d8f38fccb7
1075 - 96379f60075c75b261328aa7830ef8aa158247ac
1076 - 3ff54efdfaace9e9b2b7c1959a865be6b91de96c
1077 - 35aa621b5ab9d08767f7bc8d209b696df281d715
1078 - 900771a483ef28915a48066d7895d8252315607a
1079 - e3343e6a2819ff5d0dfc4bb5c9fb7f9a4d04da73
1080 - exec tracepoint kernel patch:
1081 - 4ff16c25e2cc48cbe6956e356c38a25ac063a64d
1082 - task_work_add kernel patches:
1083 - e73f8959af0439d114847eab5a8a5ce48f1217c4
1084 - 4d1d61a6b203d957777d73fcebf19d90b038b5b2
1085 - 413cd3d9abeaef590e5ce00564f7a443165db238
1086 - dea649b8ac1861107c5d91e1a71121434fc64193
1087 - f23ca335462e3c84f13270b9e65f83936068ec2c
1089 * What's new in version 1.7, 2012-02-01
1091 - Map inserting and deleting is now significantly faster due to
1092 improved hashing and larger hash tables. The hashes are also
1093 now randomized to provide better protection against deliberate
1096 - Formatted printing is faster by compiling the formatting directives
1097 to C code rather than interpreting at run time.
1099 - Systemtap loads extra command line options from $SYSTEMTAP_DIR/rc
1100 ($HOME/.systemtap/rc by default) before the normal argc/argv. This
1101 may be useful to activate site options such as --use-server or
1102 --download-debuginfo or --modinfo.
1104 - The stap-server has seen many improvements, and is no longer considered
1107 - The stap-server service (initscript) now supports four new options:
1112 These allow the specification of macro definitions to be passed to stap
1113 by the server, the location of the log file, network port number and
1114 NSS certificate database location respectively. These options are also
1115 supported within individual server configuration files. See stap-server
1116 and initscript/README.stap-server for details. The stap-server is no
1117 longer activated by default.
1119 - process("PATH").[library("PATH")].function("NAME").exported probes are now
1120 supported to filter function() to only exported instances.
1122 - The translator supports a new --suppress-handler-errors option, which
1123 causes most runtime errors to be turned into quiet skipped probes. This
1124 also disables the MAXERRORS and MAXSKIPPED limits.
1126 - Translator warnings have been standardized and controlled by the -w / -W
1129 - The translator supports a new --modinfo NAME=VALUE option to emit additional
1130 MODULE_INFO(n,v) macros into the generated code.
1132 - There is no more fixed maximum number of VMA pages that will be tracked
1133 at runtime. This reduces memory use for those scripts that don't need any,
1134 or only limited target process VMA tracking and allows easier system
1135 wide probes inspecting shared library variables and/or user backtraces.
1136 stap will now silently ignore -DTASK_FINDER_VMA_ENTRY_ITEMS.
1138 - The tapset functions remote_id() and remote_uri() identify the member of a
1139 swarm of "stap --remote FOO --remote BAR baz.stp" concurrent executions.
1141 - Systemtap now supports a new privilege level and group, "stapsys", which
1142 is equivalent to the privilege afforded by membership in the group "stapdev",
1143 except that guru mode (-g) functionality may not be used. To support this, a
1144 new option, --privilege=[stapusr|stapsys|stapdev] has been added.
1145 --privilege=stapusr is equivalent to specifying the existing --unprivileged
1146 option. --privilege=stapdev is the default. See man stap(1) for details.
1148 - Scripts that use kernel.trace("...") probes compile much faster.
1150 - The systemtap module cache is cleaned less frequently, governed by the
1151 number of seconds in the $SYSTEMTAP_DIR/cache/cache_clean_interval_s file.
1153 - SDT can now define up to 12 arguments in a probe point.
1155 - Parse errors no longer generate a cascade of false errors. Instead, a
1156 parse error skips the rest of the current probe or function, and resumes
1157 at the next one. This should generate fewer and better messages.
1159 - Global array wrapping is now supported for both associative and statistics typed
1160 arrays using the '%' character to signify a wrapped array. For example,
1161 'global foo%[100]' would allow the array 'foo' to be wrapped if more than 100
1162 elements are inserted.
1164 - process("PATH").library("PATH").plt("NAME") probes are now supported.
1165 Wildcards are supported in the plt-name part, to refer to any function in the
1166 program linkage table which matches the glob pattern and the rest of the
1169 - A new option, --dump-probe-types, will dump a list of supported probe types.
1170 If --unprivileged is also specified, the list will be limited to probe types
1171 which are available to unprivileged users.
1173 - Systemtap can now automatically download the required debuginfo
1174 using abrt. The --download-debuginfo[=OPTION] can be used to
1175 control this feature. Possible values are: 'yes', 'no', 'ask',
1176 and a positive number representing the timeout desired. The
1177 default behavior is to not automatically download the debuginfo.
1179 - The translator has better support for probing C++ applications by
1180 better undertanding of compilation units, nested types, templates,
1181 as used in probe point and @cast constructs.
1183 - On 2.6.29+ kernels, systemtap can now probe kernel modules that
1184 arrive and/or depart during the run-time of a session. This allows
1185 probing of device driver initialization functions, which had formerly been
1188 - New tapset functions for cpu_clock and local_clock access were added.
1190 - There is some limited preliminary support for user-space probing
1191 against kernels such as linux-next, which have no utrace but do have
1192 the newer inode-uprobes work by Srikar Dronamraju and colleagues.
1194 - The following probe types are deprecated in release 1.7 and will be
1195 removed in release 1.8:
1196 kernel.function(number).inline
1197 module(string).function(number).inline
1198 process.function(number).inline
1199 process.library(string).function(number).inline
1200 process(string).function(number).inline
1201 process(string).library(string).function(number).inline
1203 - The systemtap-grapher is deprecated in release 1.7 and will be removed in
1206 - The task_backtrace() tapset function was deprecated in 1.6 and has been
1209 - MAXBACKTRACE did work in earlier releases, but has now been documented
1210 in the stap 1 manual page.
1212 - New tapset function probe_type(). Returns a short string describing
1213 the low level probe handler type for the current probe point.
1215 - Both unwind and symbol data is now only collected and emitted for
1216 scripts actually using backtracing or function/data symbols.
1217 Tapset functions are marked with /* pragma:symbols */ or
1218 /* pragma:unwind */ to indicate they need the specific data.
1220 - Kernel backtraces can now be generated for non-pt_regs probe context
1221 if the kernel support dump_trace(). This enables backtraces from
1222 certain timer probes and tracepoints.
1224 - ubacktrace() should now also work for some kernel probes on x86 which can
1225 use the dwarf unwinder to recover the user registers to provide
1226 more accurate user backtraces.
1228 - For s390x the systemtap runtime now properly splits kernel and user
1229 addresses (which are in separate address spaces on that architecture)
1230 which enable user space introspection.
1232 - ppc and s390x now supports user backtraces through the DWARF unwinder.
1234 - ppc now handles function descriptors as symbol names correctly.
1236 - arm support kernel backtraces through the DWARF unwinder.
1238 - arm now have a uprobes port which enables user probes. This still
1239 requires some kernel patches (user_regsets and tracehook support for
1242 - Starting in release 1.7, these old variables will be deprecated:
1243 - The 'pid' variable in the 'kprocess.release' probe has been
1244 deprecated in favor of the new 'released_pid' variable.
1245 - The 'args' variable in the
1246 '_sunrpc.clnt.create_client.rpc_new_client_inline' probe has been
1247 deprecated in favor of the new internal-only '__args' variable.
1249 - Experimental support for recent kernels without utrace has been
1250 added for the following probe types:
1253 process("PATH").begin
1255 process(PID).thread.begin
1256 process("PATH").thread.begin
1257 process.thread.begin
1261 process(PID).thread.end
1262 process("PATH").thread.end
1264 process(PID).syscall
1265 process("PATH").syscall
1267 process(PID).syscall.return
1268 process("PATH").syscall.return
1269 process.syscall.return
1271 - staprun disables kprobe-optimizations in recent kernels, as problems
1272 were found. (PR13193)
1274 * What's new in version 1.6, 2011-07-25
1276 - Security fixes for CVE-2011-2503: read instead of mmap to load modules,
1277 CVE-2011-2502: Don't allow path-based auth for uprobes
1279 - The systemtap compile-server no longer uses the -k option when calling the
1280 translator (stap). As a result, the server will now take advantage of the
1281 module cache when compiling the same script more than once. You may observe
1282 an improvement in the performance of the server in this situation.
1284 - The systemtap compile-server and client now each check the version of the
1285 other, allowing both to adapt when communicating with a down-level
1286 counterpart. As a result, all version of the client can communicate
1287 with all versions of the server and vice-versa. Client will prefer newer
1288 servers when selecting a server automatically.
1290 - SystemTap has improved support for the ARM architecture. The
1291 kread() and kwrite() operations for ARM were corrected allowing many
1292 of the tapsets probes and function to work properly on the ARM
1295 - Staprun can now rename the module to a unique name with the '-R' option before
1296 inserting it. Systemtap itself will also call staprun with '-R' by default.
1297 This allows the same module to be inserted more than once, without conflicting
1300 - Systemtap error messages now provide feedback when staprun or any other
1301 process fails to launch. This also specifically covers when the user
1302 doesn't have the proper permissions to launch staprun.
1304 - Systemtap will now map - to _ in module names. Previously,
1305 stap -L 'module("i2c-core").function("*")' would be empty. It now returns
1306 a list had stap -L 'module("i2c_core").function("*") been specified.
1308 - Systemtap now fills in missing process names to probe points, to
1309 avoid having to name them twice twice:
1310 % stap -e 'probe process("a.out").function("*") {}' -c 'a.out ...'
1311 Now the probed process name is inferred from the -c CMD argument.
1312 % stap -e 'probe process.function("*") {}' -c 'a.out ...'
1314 - stap -L 'process("PATH").syscall' will now list context variables
1316 - Depends on elfutils 0.142+.
1318 - Deprecated task_backtrace:string (task:long). This function will go
1319 away after 1.6. Please run your scripts with stap --check-version.
1321 * What's new in version 1.5, 2011-05-23
1323 - Security fixes for CVE-2011-1781, CVE-2011-1769: correct DW_OP_{mod,div}
1324 division-by-zero bug
1326 - The compile server and its related tools (stap-gen-ert, stap-authorize-cert,
1327 stap-sign-module) have been re-implemented in C++. Previously, these
1328 components were a mix of bash scripts and C code. These changes should be
1329 transparent to the end user with the exception of NSS certificate database
1330 password prompting (see below). The old implementation would prompt more
1331 than once for the same password in some situations.
1333 - eventcount.stp now allows for event counting in the format of
1334 'stap eventcount.stp process.end syscall.* ...', and also reports
1335 corresponding event tid's.
1337 - Systemtap checks that the build-id of the module being probed matches the
1338 build-id saved in the systemtap module. Invoking systemtap with
1339 -DSTP_NO_BUILDID_CHECK will bypass this build-id runtime verification. See
1340 man ld(1) for info on --build-id.
1342 - stapio will now report if a child process has an abnormal exit along with
1343 the associated status or signal.
1345 - Compiler optimization may sometimes result in systemtap not being able to
1346 access a user-space probe argument. Compiling the application with
1347 -DSTAP_SDT_ARG_CONSTRAINT=nr will force the argument to be an immediate or
1348 register value which should enable systemtap to access the argument.
1350 - GNU Gettext has now been intergrated with systemtap. Our translation
1351 page can be found at http://www.transifex.net/projects/p/systemtap/ .
1352 "make update-po" will generate the necessary files to use translated
1353 messages. Please refer to the po/README file for more info and
1354 please consider contributing to this I18N effort!
1356 - The new addr() function returns the probe's instruction pointer.
1358 - process("...").library("...") probes are now supported. Wildcards
1359 are supported in the library-name part, to refer to any shared
1360 library that is required by process-name, which matches the glob
1361 pattern and the rest of the probe point.
1363 - The "--remote USER@HOST" functionality can now be specified multiple times
1364 to fan out on multiple targets. If the targets have distinct kernel and
1365 architecture configurations, stap will automatically build the script
1366 appropriately for each one. This option is also no longer considered
1369 - The NSS certificate database generated for use by the compile server is now
1370 generated with no password. Previously, a random password was generated and
1371 used to access the database. This change should be transparent to most users.
1372 However, if you are prompted for a password when using systemtap, then
1373 running $libexecdir/stap-gen-cert should correct the problem.
1375 - The timestamp tapset includes jiffies() and HZ() for lightweight approximate
1378 - A powerful new command line option --version has been added.
1380 - process.mark now supports $$parms for reading probe parameters.
1382 - A new command line option, --use-server-on-error[=yes|no] is available
1383 for stap. It instructs stap to retry compilation of a script using a
1384 compile server if it fails on the local host. The default setting
1387 - The following deprecated tools have been removed:
1389 stap-authorize-server-cert
1390 stap-authorize-signing-cert
1391 stap-find-or-start-server
1393 Use the --use-server, --trust-server and --list-servers options of stap
1396 * What's new in version 1.4, 2011-01-17
1398 - Security fixes for CVE-2010-4170, CVE-2010-4171: staprun module
1401 - A new /* myproc-unprivileged */ marker is now available for embedded C
1402 code and and expressions. Like the /* unprivileged */ marker, it makes
1403 the code or expression available for use in unprivileged mode (see
1404 --unprivileged). However, it also automatically adds a call to
1405 assert_is_myproc() to the code or expression, thus, making it available
1406 to the unprivileged user only if the target of the current probe is within
1407 the user's own process.
1409 - The experimental "--remote USER@HOST" option will run pass 5 on a given
1410 ssh host, after building locally (or with --use-server) for that target.
1412 - Warning messages from the script may now be suppressed with the stap
1413 and/or staprun -w option. By default, duplicate warning messages are
1414 suppressed (up to a certain limit). With stap --vp 00002 and above,
1415 the duplicate elimination is defeated.
1417 - The print_ubacktrace and usym* functions attempt to print the full
1418 path of the user-space binaries' paths, instead of just the basename.
1419 The maximum saved path length is set by -DTASK_FINDER_VMA_ENTRY_PATHLEN,
1420 default 64. Warning messages are produced if unwinding fails due to
1421 a missing 'stap -d MODULE' option, providing preloaded unwind data.
1423 - The new tz_ctime() tapset function prints times in the local time zone.
1425 - More kernel tracepoints are accessible to the kernel.trace("...") mechanism,
1426 if kernel source trees or debuginfo are available. These formerly "hidden"
1427 tracepoints are those that are declared somewhere other than the usual
1428 include/linux/trace/ headers, such as xfs and kvm.
1430 - debuginfo-based process("...").function/.statement/.mark probes support
1431 wildcards in the process-name part, to refer to any executable files that
1432 match the glob pattern and the rest of the probe point.
1434 - The -t option now displays information per probe-point rather than a summary
1435 for each probe. It also now shows the derivation chain for each probe-point.
1437 - A rewrite of the sys/sdt.h header file provides zero-cost startup (few or
1438 no ELF relocations) for the debuginfo-less near-zero-cost runtime probes.
1439 Binaries compiled with earlier sdt.h versions remain supported. The
1440 stap -L (listing) option now lists parameters for sys/sdt.h markers.
1442 - The implementation of the integrated compile-server client has been
1444 o --use-server now accepts an argument representing a particular server and
1445 may be specified more than once.
1446 o --list-servers now accepts an expanded range of arguments.
1447 o a new --trust-servers option has been added to stap to replace several
1448 old certificate-management scripts.
1449 o The following tools are now deprecated and will be removed in release 1.5:
1451 stap-authorize-server-cert
1452 stap-authorize-signing-cert
1453 stap-find-or-start-server
1455 See man stap(1) for complete details.
1457 - The compile-server now returns the uprobes.ko to the client when it is
1458 required by the script being compiled. The integrated compile-server client
1459 now makes it available to be loaded by staprun. The old (deprecated)
1460 stap-client does not do this.
1462 - process probes with scripts as the target are recognized by stap and the
1463 interpreter would be selected for probing.
1465 - Starting in release 1.5, these old variables/functions will be deprecated
1466 and will only be available when the '--compatible=1.4' flag is used:
1468 - In the 'syscall.add_key' probe, the 'description_auddr' variable
1469 has been deprecated in favor of the new 'description_uaddr'
1471 - In the 'syscall.fgetxattr', 'syscall.fsetxattr',
1472 'syscall.getxattr', 'syscall.lgetxattr', and
1473 'syscall.lremovexattr' probes, the 'name2' variable has been
1474 deprecated in favor of the new 'name_str' variable.
1475 - In the 'nd_syscall.accept' probe the 'flag_str' variable
1476 has been deprecated in favor of the new 'flags_str' variable.
1477 - In the 'nd_syscall.dup' probe the 'old_fd' variable has been
1478 deprecated in favor of the new 'oldfd' variable.
1479 - In the 'nd_syscall.fgetxattr', 'nd_syscall.fremovexattr',
1480 'nd_syscall.fsetxattr', 'nd_syscall.getxattr', and
1481 'nd_syscall.lremovexattr' probes, the 'name2' variable has been
1482 deprecated in favor of the new 'name_str' variable.
1483 - The tapset alias 'nd_syscall.compat_pselect7a' was misnamed. It should
1484 have been 'nd_syscall.compat_pselect7' (without the trailing 'a').
1485 - The tapset function 'cpuid' is deprecated in favor of the better known
1487 - In the i386 'syscall.sigaltstack' probe, the 'ussp' variable has
1488 been deprecated in favor of the new 'uss_uaddr' variable.
1489 - In the ia64 'syscall.sigaltstack' probe, the 'ss_uaddr' and
1490 'oss_uaddr' variables have been deprecated in favor of the new
1491 'uss_uaddr' and 'uoss_uaddr' variables.
1492 - The powerpc tapset alias 'syscall.compat_sysctl' was deprecated
1493 and renamed 'syscall.sysctl32'.
1494 - In the x86_64 'syscall.sigaltstack' probe, the 'regs_uaddr'
1495 variable has been deprecated in favor of the new 'regs' variable.
1497 * What's new in version 1.3, 2010-07-21
1499 - The uprobes kernel module now has about half the overhead when probing
1500 NOPs, which is particularly relevant for sdt.h markers.
1502 - New stap option -G VAR=VALUE allows overriding global variables
1503 by passing the settings to staprun as module options.
1505 - The tapset alias 'syscall.compat_pselect7a' was misnamed. It should
1506 have been 'syscall.compat_pselect7' (without the trailing 'a').
1507 Starting in release 1.4, the old name will be deprecated and
1508 will only be available when the '--compatible=1.3' flag is used.
1510 - A new procfs parameter .umask(UMASK) which provides modification of
1511 file permissions using the proper umask value. Default file
1512 permissions for a read probe are 0400, 0200 for a write probe, and
1513 0600 for a file with a read and write probe.
1515 - It is now possible in some situations to use print_ubacktrace() to
1516 get a user space stack trace from a kernel probe point. e.g. for
1517 user backtraces when there is a pagefault:
1518 $ stap -d /bin/sort --ldd -e 'probe vm.pagefault {
1519 if (pid() == target()) {
1520 printf("pagefault @0x%x\n", address); print_ubacktrace();
1523 pagefault @0x7fea0595fa70
1524 0x000000384f07f958 : __GI_strcmp+0x12b8/0x1440 [libc-2.12.so]
1525 0x000000384f02824e : __gconv_lookup_cache+0xee/0x5a0 [libc-2.12.so]
1526 0x000000384f021092 : __gconv_find_transform+0x92/0x2cf [libc-2.12.so]
1527 0x000000384f094896 : __wcsmbs_load_conv+0x106/0x2b0 [libc-2.12.so]
1528 0x000000384f08bd90 : mbrtowc+0x1b0/0x1c0 [libc-2.12.so]
1529 0x0000000000404199 : ismbblank+0x39/0x90 [sort]
1530 0x0000000000404a4f : inittables_mb+0xef/0x290 [sort]
1531 0x0000000000406934 : main+0x174/0x2510 [sort]
1532 0x000000384f01ec5d : __libc_start_main+0xfd/0x1d0 [libc-2.12.so]
1533 0x0000000000402509 : _start+0x29/0x2c [sort]
1536 - New tapset functions to get a string representation of a stack trace:
1537 sprint_[u]backtrace() and sprint_[u]stack().
1539 - New tapset function to get the module (shared library) name for a
1540 user space address umodname:string(long). The module name will now
1541 also be in the output of usymdata() and in backtrace addresses even
1542 when they were not given with -d at the command line.
1544 - Kernel backtraces are now much faster (replaced a linear search
1545 with a binary search).
1547 - A new integrated compile-server client is now available as part of stap.
1549 o 'stap --use-server ...' is equivalent to 'stap-client ...'
1550 o 'stap --list-servers' is equivalent to 'stap-find-servers'
1551 o 'stap --list-servers=online' is equivalent to 'stap-find-servers --all'
1552 o stap-client and its related tools will soon be deprecated.
1553 o the nss-devel and avahi-devel packages are required for building stap with
1554 the integrated client (checked during configuration).
1555 o nss and avahi are required to run the integrated client.
1557 - A new operator @entry is available for automatically saving an expression
1558 at entry time for use in a .return probe.
1559 probe foo.return { println(get_cycles() - @entry(get_cycles())) }
1561 - Probe $target variables and @cast() can now use a suffix to print complex
1562 data types as strings. Use a single '$' for a shallow view, or '$$' for a
1563 deeper view that includes nested types. For example, with fs_struct:
1564 $fs$ : "{.users=%i, .lock={...}, .umask=%i,
1565 .in_exec=%i, .root={...}, .pwd={...}}"
1566 $fs$$ : "{.users=%i, .lock={.raw_lock={.lock=%u}}, .umask=%i, .in_exec=%i,
1567 .root={.mnt=%p, .dentry=%p}, .pwd={.mnt=%p, .dentry=%p}}"
1569 - The <sys/sdt.h> user-space markers no longer default to an implicit
1570 MARKER_NAME_ENABLED() semaphore check for each marker. To check for
1571 enabled markers use a .d declaration file, then:
1572 if (MARKER_NAME_ENABLED()) MARKER_NAME()
1574 - Hyphenated <sys/sdt.h> marker names such as process(...).mark("foo-bar")
1575 are now accepted in scripts. They are mapped to the double-underscore
1578 - More robust <sys/sdt.h> user-space markers support is included. For
1579 some platforms (x86*, ppc*), this can let systemtap probe the markers
1580 without debuginfo. This implementation also supports preserving
1581 the "provider" name associated with a marker:
1582 probe process("foo").provider("bar").mark("baz") to match
1583 STAP_PROBE<n>(bar, baz <...>)
1584 (Compile with -DSTAP_SDT_V1 to revert to the previous implementation.
1585 Systemtap supports pre-existing or new binaries using them.)
1587 - Embedded-C may be used within expressions as values, when in guru mode:
1588 num = %{ LINUX_VERSION_CODE %} // int64_t
1589 name = %{ /* string */ THIS_MODULE->name %} // const char*
1590 printf ("%s %x\n", name, num)
1591 The usual /* pure */, /* unprivileged */, and /* guru */ markers may be used
1592 as with embedded-C functions.
1594 - By default the systemtap-runtime RPM builds now include a shared
1595 library, staplog.so, that allows crash to extract systemtap data from
1598 - Iterating with "foreach" can now explicitly save the value for the loop.
1599 foreach(v = [i,j] in array)
1600 printf("array[%d,%s] = %d\n", i, j, v /* array[i,j] */)
1602 - The new "--ldd" option automatically adds any additional shared
1603 libraries needed by probed or -d-listed userspace binaries to the -d
1604 list, to enable symbolic backtracing through them. Similarly, the
1605 new "--all-modules" option automatically adds any currently loaded
1606 kernel modules (listed in /proc/modules) to the -d list.
1608 - A new family of set_kernel_* functions make it easier for gurus to write
1609 new values at arbitrary memory addresses.
1611 - Probe wildcards can now use '**' to cross the '.' separator.
1612 $ stap -l 'sys**open'
1616 - Backward compatibility flags (--compatible=VERSION, and matching
1617 script preprocessing predicate %( systemtap_v CMP "version" %)
1618 and a deprecation policy are being introduced, in case future
1619 tapset/language changes break valid scripts.
1621 * What's new in version 1.2, 2010-03-22
1623 - Prototype support for "perf events", where the kernel supports the
1624 2.6.33 in-kernel API. Probe points may refer to low-level
1625 perf_event_attr type/config numbers, or to a number of aliases
1626 defined in the new perf.stp tapset:
1627 probe perf.sw.cpu_clock, perf.type(0).config(4) { }
1629 - Type-casting can now use multiple headers to resolve codependencies.
1630 @cast(task, "task_struct",
1631 "kernel<linux/sched.h><linux/fs_struct.h>")->fs->umask
1633 - Tapset-related man pages have been renamed. 'man -k 3stap' should show
1634 the installed list, which due to prefixing should no longer collide over
1635 ordinary system functions.
1637 - User space marker arguments no longer use volatile if the version of gcc,
1638 which must be at least 4.5.0, supports richer DWARF debuginfo. Use cflags
1639 -DSTAP_SDT_VOLATILE=volatile or -DSTAP_SDT_VOLATILE= when building
1640 the sys/sdt.h application to override this one way or another.
1642 - A new construct for error handling is available. It is similar to c++
1643 exception catching, using try and catch as new keywords. Within a handler
1644 or function, the following is valid and may be nested:
1645 try { /* arbitrary statements */ }
1646 catch (er) { /* e.g. println("caught error ", er) */ }
1648 - A new command line flag '-W' forces systemtap to abort translation of
1649 a script if any warnings are produced. It is similar to gcc's -Werror.
1650 (If '-w' is also supplied to suppress warnings, it wins.)
1652 - A new predicate @defined is available for testing whether a
1653 particular $variable/expression is resolvable at translate time:
1654 probe foo { if (@defined($bar)) log ("$bar is available here") }
1656 - Adjacent string literals are glued together, making this
1658 probe process("/usr" @1 "/bin").function("*") { ... }
1660 - In order to limit potential impact from future security problems,
1661 the stap-server process does not permit its being launched as root.
1663 - On recent kernels, for some architectures/configurations, hardware
1664 breakpoint probes are supported. The probe point syntax is:
1666 probe kernel.data(ADDRESS).write
1667 probe kernel.data(ADDRESS).length(LEN).write
1668 probe kernel.data("SYMBOL_NAME").write
1670 * What's new in version 1.1, 2010-01-15
1672 - New tracepoint based tapset for memory subsystem.
1674 - The loading of signed modules by staprun is no longer allowed for
1675 ordinary, unprivileged users. This means that only root, members of
1676 the group 'stapdev' and members of the group 'stapusr' can load
1677 systemtap modules using staprun, stap or stap-client. The minimum
1678 privilege required to run arbitrary --unprivileged scripts is now
1679 'stapusr' membership.
1681 - The stap-server initscript is available. This initscript allows you
1682 to start systemtap compile servers as a system service and to manage
1683 these servers as a group or individually. The stap-server initscript
1684 is installed by the systemtap-server rpm. The build directory for
1685 the uprobes module (/usr/share/systemtap/runtime/uprobes) is made
1686 writable by the 'stap-server' group. All of the files generated when
1687 building the uprobes module, including the digital signature, are
1688 also writable by members of stap-server.
1690 See initscript/README.stap-server for details.
1692 - Some of the compile server client, server and certificate management
1693 tools have been moved from $bindir to $libexecdir/systemtap.
1694 You should use the new stap-server script or the stap-server initscript
1695 for server management where possible. The stap-server script provides the same
1696 functionality as the stap-server initscript except that the servers are
1697 run by the invoking user by default as opposed to servers started by the
1698 stap-server initscript which are run by the user stap-server
1699 by default. See stap-server(8) for more information.
1701 You may continue to use these tools by adding $libexecdir/systemtap to
1702 your path. You would need to do this, for example, if you are not root,
1703 you want to start a compile server and you are not running systemtap from a
1704 private installation. In this case you still need to use stap-start-server.
1706 - Any diagnostic output line that starts with "ERROR", as in
1707 error("foo"), will promote a "Pass 5: run failed", and the return
1710 - Systemtap now warns about global variables being referenced from other
1711 script files. This aims to protect against unintended local-vs-global
1712 namespace collisions such as:
1714 % cat some_tapset.stp
1715 probe baz.one = bar { foo = $foo; bar = $bar }
1716 % cat end_user_script.stp
1717 global foo # intended to be private variable
1718 probe timer.s(1) { foo ++ }
1719 probe baz.* { println(foo, pp()) }
1720 % stap end_user_script.stp
1721 WARNING: cross-file global variable reference to foo from some_tapset.stp
1723 - Preprocessor conditional for kernel configuration testing:
1724 %( CONFIG_foo == "y" %? ... %)
1726 - ftrace(msg:string) tapset function to send strings to the system-wide
1727 ftrace ring-buffer (if any).
1729 - Better support for richer DWARF debuginfo output from GCC 4.5
1730 (variable tracking assignments). Kernel modules are now always resolved
1731 against all their dependencies to find any info referring to missing
1732 symbols. DW_AT_const_value is now supported when no DW_AT_location
1735 * What's new in verson 1.0, 2009-09-22
1737 - process().mark() probes now use an enabling semaphore to reduce the
1738 computation overhead of dormant probes.
1740 - The function spec for dwarf probes now supports C++ scopes, so you can
1741 limit the probes to specific namespaces or classes. Multiple scopes
1742 can be specified, and they will be matched progressively outward.
1743 probe process("foo").function("std::vector<*>::*") { }
1744 probe process("foo").function("::global_function") { }
1746 - It is now possible to cross-compile systemtap scripts for foreign
1747 architectures, using the new '-a ARCH' and '-B OPT=VALUE' flags.
1748 For example, put arm-linux-gcc etc. into your $PATH, and point
1749 systemtap at the target kernel build tree with:
1750 stap -a arm -B CROSS_COMPILE=arm-linux- -r /build/tree [...]
1751 The -B option is passed to kbuild make. -r identifies the already
1752 configured/built kernel tree and -a its architecture (kbuild ARCH=...).
1753 Systemtap will infer -p4.
1755 - Cross compilation using the systemtap client and server
1756 - stap-start-server now accepts the -r, -R, -I, -B and -a options in
1757 order to start a cross compiling server. The server will correctly
1758 advertise itself with respect to the kernel release and architecture
1759 that it compiles for.
1760 - When specified on stap-client, the -r and -a options will be
1761 considered when searching for a suitable server.
1763 - When using the systemtap client and server udp port 5353 must be open
1764 in your firewall in order for the client to find servers using
1765 avahi-browse. Also the systemtap server will choose a random port in
1766 the range 1024-63999 for accepting ssl connections.
1768 - Support for unprivileged users:
1769 ***********************************************************************
1770 * WARNING!!!!!!!!!! *
1771 * This feature is EXPERIMENTAL at this time and should be used with *
1772 * care. This feature allows systemtap kernel modules to be loaded by *
1773 * unprivileged users. The user interface and restrictions will change *
1774 * as this feature evolves. *
1775 ***********************************************************************
1776 - Systemtap modules generated from scripts which use a restricted
1777 subset of the features available may be loaded by staprun for
1778 unprivileged users. Previously, staprun would load modules only for
1779 root or for members of the groups stapdev and stapusr.
1780 - Using the --unprivileged option on stap enables translation-time
1781 checking for use by unprivileged users (see restrictions below).
1782 - All modules deemed suitable for use by unprivileged users will be
1783 signed by the systemtap server when --unprivileged is specified on
1784 stap-client. See module signing in release 0.9.8 and stap-server in
1786 - Modules signed by trusted signers (servers) and verified by staprun
1787 will be loaded by staprun regardless of the user's privilege level.
1788 - The system administrator asserts the trustworthiness of a signer
1789 (server) by running stap-authorize-signing-cert <cert-file> as root,
1790 where the <cert-file> can be found in
1791 ~<user>/.systemtap/ssl/server/stap.cert for servers started by
1792 ordinary users and in $sysconfdir/systemtap/ssl/server/stap.cert for
1793 servers started by root.
1794 - Restrictions are intentionally strict at this time and may be
1795 relaxed in the future:
1796 - probe points are restricted to:
1797 begin, begin(n), end, end(n), error, error(n), never,
1798 timer.{jiffies,s,sec,ms,msec,us,usec,ns,nsec}(n)*, timer.hz(n),
1799 process.* (for processes owned by the user).
1800 - use of embedded C code is not allowed.
1801 - use of tapset functions is restricted.
1802 - some tapset functions may not be used at all. A message will be
1803 generated at module compilation time.
1804 - some actions by allowed tapset functions may only be performed
1805 in the context of the user's own process. A runtime fault will
1806 occur in these situations, for example, direct memory access.
1807 - The is_myproc() tapset function has been provided so that
1808 tapset writers for unprivileged users can check that the
1809 context is of the users own process before attempting these
1811 - accessing the kernel memory space is not allowed.
1812 - The following command line options may not be used by stap-client
1814 - The following environment variables are ignored by stap-client:
1815 SYSTEMTAP_RUNTIME, SYSTEMTAP_TAPSET, SYSTEMTAP_DEBUGINFO_PATH
1816 - nss and nss-tools are required to use this feature.
1818 - Support output file switching by SIGUSR2. Users can command running
1819 stapio to switch output file by sending SIGUSR2.
1821 - Memory consumption for scripts involving many uprobes has been
1822 dramatically reduced.
1824 - The preprocessor now supports || and && in the conditions.
1825 e.g. %( arch == "x86_64" || arch == "ia64" %: ... %)
1827 - The systemtap notion of "architecture" now matches the kernel's, rather
1828 than that of "uname -m". This means that 32-bit i386 family are all
1829 known as "i386" rather than "i386" or "i686"; "ppc64" as "powerpc";
1830 "s390x" as "s390", and so on. This is consistent between the new
1831 "-a ARCH" flag and the script-level %( arch ... %) conditional.
1833 - It is now possible to define multiple probe aliases with the same name.
1834 A probe will expand to all matching aliases.
1837 probe foo { } # expands twice, once to bar and once to baz
1839 - A new experimental transport mechanism, using ftrace's ring_buffer,
1840 has been added. This may become the default transport mechanism in
1841 future versions of systemtap. To test this new transport mechanism,
1842 define 'STP_USE_RING_BUFFER'.
1844 - Support for recognizing DW_OP_{stack,implicit}_value DWARF expressions
1845 as emitted by GCC 4.5.
1847 * What's new in version 0.9.9, 2009-08-04
1849 - Systemwide kernel .function.return (kretprobe) maxactive defaults may
1850 be overridden with the -DKRETACTIVE=nnn parameter.
1852 - Translation pass 2 is significantly faster by avoiding unnecessary
1853 searching through a kernel build/module directory tree.
1855 - When compiled against elfutils 0.142 systemtap now handles the new
1856 DW_OP_call_frame_CFA generated by by GCC.
1858 - uprobes and ustack() are more robust when used on applications that
1859 depend on prelinked/separate debuginfo shared libraries.
1861 - User space PROBE marks are not always found with or without separate
1862 debuginfo. The .probes section itself is now always put in the main
1863 elf file and marked as allocated. When building pic code the section
1864 is marked writable. The selinux memory check problems seen with
1865 programs using STAP_PROBES is fixed.
1867 - statement() probes can now override "address not at start of statement"
1868 errors in guru mode. They also provide alternative addresses to use
1871 - The stapgraph application can generate graphs of data and events
1872 emitted by systemtap scripts in real time. Run "stapgraph
1873 testsuite/systemtap.examples/general/grapher.stp" for an example of
1874 graphing the system load average and keyboard events.
1876 - Dwarf probes now show parameters and local variables in the verbose
1879 - Symbol aliases are now resolved to their canonical dwarf names. For
1880 example, probing "malloc" in libc resolves to "__libc_malloc".
1882 - The syntax for dereferencing $target variables and @cast() gained new
1884 - Array indexes can now be arbitrary numeric expressions.
1885 - Array subscripts are now supported on pointer types.
1886 - An '&' operator before a @cast or $target returns the address of the
1887 final component, especially useful for nested structures.
1889 - For reading all probe variables, kernel.mark now supports $$vars and
1890 $$parms, and process.syscall now supports $$vars.
1892 - The SNMP tapset provides probes and functions for many network
1893 statistics. See stapprobes.snmp(3stap) for more details.
1895 - The dentry tapset provides functions to map kernel VFS directory entries
1896 to file or full path names: d_path(), d_name() and reverse_path_walk().
1898 - SystemTap now has userspace markers in its own binaries, and the stap
1899 tapset provides the available probepoints and local variables.
1901 - Miscellaneous new tapset functions:
1902 - pgrp() returns the process group ID of the current process
1903 - str_replace() performs string replacement
1905 * What's new in version 0.9.8, 2009-06-11
1907 - Miscellaneous new tapset functions:
1908 - sid() returns the session ID of the current process
1909 - stringat() indexes a single character from a string.
1911 - Using %M in print formats for hex dumps can now print entire buffers,
1912 instead of just small numbers.
1914 - Dwarfless syscalls: The nd_syscalls tapset is now available to probe
1915 system calls without requiring kernel debugging information. All of
1916 the same probepoints in the normal syscalls tapset are available with
1917 an "nd_" prefix, e.g. syscall.open becomes nd_syscall.open. Most
1918 syscall arguments are also available by name in nd_syscalls.
1920 - Module signing: If the appropriate nss libraries are available on your
1921 system, stap-server will sign each compiled module using a self-generated
1922 certificate. This is the first step toward extending authority to
1923 load certain modules to unprivileged users. For now, if the system
1924 administrator adds a certificate to a database of trusted signers
1925 (stap-authorize-signing-cert), modules signed using that certificate
1926 will be verified by staprun against tampering. Otherwise, you should
1927 notice no difference in the operation of stap or staprun.
1929 * What's new in version 0.9.7, 2009-04-23
1931 - @cast can now determine its type information using an explicit header
1932 specification. For example:
1933 @cast(tv, "timeval", "<sys/time.h>")->tv_sec
1934 @cast(task, "task_struct", "kernel<linux/sched.h>")->tgid
1936 - The overlapping process.* tapsets are now separated. Those probe points
1937 documented in stapprobes(3stap) remain the same. Those that were formerly
1938 in stapprobes.process(3stap) have been renamed to kprocess, to reflect
1939 their kernel perspective on processes.
1941 - The --skip-badvars option now also suppresses run-time error
1942 messages that would otherwise result from erroneous memory accesses.
1943 Such accesses can originate from $context expressions fueled by
1944 erroneous debug data, or by kernel_{long,string,...}() tapset calls.
1946 - New probes kprobe.function(FUNCTION) and kprobe.function(FUNCTION).return
1947 for dwarfless probing. These postpone function address resolution to
1948 run-time and use the kprobe symbol-resolution mechanism.
1949 Probing of absolute statements can be done using the
1950 kprobe.statement(ADDRESS).absolute construct.
1952 - EXPERIMENTAL support for user process unwinding. A new collection of
1953 tapset functions have been added to handle user space backtraces from
1954 probe points that support them (currently process and timer probes -
1955 for timer probes test whether or not in user space first with the
1956 already existing user_mode() function). The new tapset functions are:
1957 uaddr - User space address of current running task.
1958 usymname - Return the symbol of an address in the current task.
1959 usymdata - Return the symbol and module offset of an address.
1960 print_ustack - Print out stack for the current task from string.
1961 print_ubacktrace - Print stack back trace for current task.
1962 ubacktrace - Hex backtrace of current task stack.
1963 Please read http://sourceware.org/ml/systemtap/2009-q2/msg00364.html
1964 on the current restrictions and possible changes in the future and
1965 give feedback if you want to influence future developments.
1967 * What's new in version 0.9.5, 2009-03-27
1969 - New probes process().insn and process().insn.block that allows
1970 inspection of the process after each instruction or block of
1971 instructions executed. So to count the total number of instructions
1972 a process executes during a run do something like:
1973 $ stap -e 'global steps; probe process("/bin/ls").insn {steps++}
1974 probe end {printf("Total instructions: %d\n", steps);}' \
1976 This feature can slow down execution of a process somewhat.
1978 - Systemtap probes and function man pages extracted from the tapsets
1979 are now available under 3stap. To show the page for probe vm.pagefault
1980 or the stap function pexecname do:
1981 $ man 3stap vm.pagefault
1982 $ man 3stap pexecname
1984 - Kernel tracepoints are now supported for probing predefined kernel
1985 events without any debuginfo. Tracepoints incur less overhead than
1986 kprobes, and context parameters are available with full type
1987 information. Any kernel 2.6.28 and later should have defined
1988 tracepoints. Try the following to see what's available:
1989 $ stap -L 'kernel.trace("*")'
1991 - Typecasting with @cast now supports modules search paths, which is
1992 useful in case there are multiple places where the type definition
1993 may be found. For example:
1994 @cast(sdev, "scsi_device", "kernel:scsi_mod")->sdev_state
1996 - On-file flight recorder is supported. It allows stap to record huge
1997 trace log on the disk and to run in background.
1998 Passing -F option with -o option runs stap in background mode. In this
1999 mode, staprun is detached from console, and stap itself shows staprun's
2001 Specifying the max size and the max number of log files are also available
2002 by passing -S option. This option has one or two arguments seperated by
2003 a comma. The first argument is the max size of a log file in MB. If the
2004 size of a log file exceeds it, stap switches to the next log file
2005 automatically. The second is how many files are kept on the disk. If the
2006 number of log files exceeds it, the oldest log file is removed
2007 automatically. The second argument can be omitted.
2009 For example, this will record output on log files each of them is smaller
2010 than 1024MB and keep last 3 logs, in background.
2011 % stap -F -o /tmp/staplog -S 1024,3 script.stp
2013 - In guru mode (-g), the kernel probing blacklist is disabled, leaving
2014 only a subset - the kernel's own internal kprobe blacklist - to attempt
2015 to filter out areas unsafe to probe. The differences may be enough to
2016 probe more interrupt handlers.
2018 - Variables unavailable in current context may be skipped by setting a
2019 session level flag with command line option --skip-badvars now available.
2020 This replaces any dwarf $variable expressions that could not be resolved
2021 with literal numeric zeros, along with a warning message.
2023 - Both kernel markers and kernel tracepoint support argument listing
2024 through stap -L 'kernel.mark("*")' or stap -L 'kernel.trace("*")'
2026 - Users can use -DINTERRUPTIBLE=0 to prevent interrupt reentrancy in
2027 their script, at the cost of a bit more overhead to toggle the
2030 - Added reentrancy debugging. If stap is run with the arguments
2031 "-t -DDEBUG_REENTRANCY", additional warnings will be printed for
2032 every reentrancy event, including the probe points of the
2033 resident and interloper probes.
2035 - Default to --disable-pie for configure.
2036 Use --enable-pie to turn it back on.
2038 - Improved sdt.h compatibility and test suite for static dtrace
2039 compatible user space markers.
2041 - Some architectures now use syscall wrappers (HAVE_SYSCALL_WRAPPERS).
2042 The syscall tapset has been enhanced to take care of the syscall
2043 wrappers in this release.
2045 - Security fix for CVE-2009-0784: stapusr module-path checking race.
2047 * What's new in version 0.9, 2009-02-19
2049 - Typecasting is now supported using the @cast operator. A script can
2050 define a pointer type for a "long" value, and then access type members
2051 using the same syntax as with $target variables. For example, this will
2052 retrieve the parent pid from a kernel task_struct:
2053 @cast(pointer, "task_struct", "kernel")->parent->pid
2055 - process().mark() probes are now possible to trace static user space
2056 markers put in programs with the STAP_PROBE macro using the new
2057 sys/sdt.h include file. This also provides dtrace compatible markers
2058 through DTRACE_PROBE and an associated python 'dtrace' script that
2059 can be used in builds based on dtrace that need dtrace -h or -G
2062 - For those that really want to run stap from the build tree there is
2063 now the 'run-stap' script in the top-level build directory that sets
2064 up the SYSTEMTAP_TAPSET, SYSTEMTAP_RUNTIME, SYSTEMTAP_STAPRUN, and
2065 SYSTEMTAP_STAPIO environment variables (installing systemtap, in a
2066 local prefix, is still recommended for common use).
2068 - Systemtap now comes with a new Beginners Guide that walks the user
2069 through their first steps setting up stap, understanding how it all
2070 works, introduces some useful scripts and describes some common
2071 pitfalls. It isn't created by default since it needs a Publican
2072 setup, but full build instructions can be found in the wiki:
2073 http://sourceware.org/systemtap/wiki/PublicanQuikHowto
2074 An online version can be found at:
2075 http://sourceware.org/systemtap/SystemTap_Beginners_Guide.pdf
2077 - Standard tapsets included with Systemtap were modified to include
2078 extractable documentation information based on the kernel-doc
2079 infrastructure. When configured --enabled-docs a HTML and PDF
2080 version of the Tapset Reference Manual is produced explaining probes
2081 defined in each tapset.
2083 - The systemtap client and compile server are now available.
2084 These allow you to compile a systemtap module on a host other than
2085 the one which it will be run, providing the client and server
2086 are compatible. Other than using a server for passes 1 through
2087 4, the client behaves like the 'stap' front end itself. This
2088 means, among other things, that the client will automatically
2089 load the resulting module on the local host unless -p[1234]
2090 was specified. See stap-server(8) for more details.
2091 The client/server now use SSL for network connection security and
2094 The systemtap client and server are prototypes only. Interfaces, options
2095 and usage may change at any time.
2097 - function("func").label("label") probes are now supported to allow matching
2098 the label of a function.
2100 - Systemtap initscript is available. This initscript allows you to run
2101 systemtap scripts as system services (in flight recorder mode) and
2102 control those scripts individually.
2103 See README.systemtap for details.
2105 - The stap "-r DIR" option may be used to identify a hand-made kernel
2106 build directory. The tool determines the appropriate release string
2107 automatically from the directory.
2109 - Serious problems associated with user-space probing in shared libraries
2110 were corrected, making it now possible to experiment with probe shared
2111 libraries. Assuming dwarf debugging information is installed, use this
2112 twist on the normal syntax:
2114 probe process("/lib64/libc-2.8.so").function("....") { ... }
2116 This would probe all threads that call into that library. Running
2117 "stap -c CMD" or "stap -x PID" naturally restricts this to the target
2118 command+descendants only. $$vars etc. may be used.
2120 - For scripts that sometimes terminate with excessive "skipped" probes,
2121 rerunning the script with "-t" (timing) will print more details about
2122 the skippage reasons.
2124 - Symbol tables and unwind (backtracing) data support were formerly
2125 compiled in for all probed modules as identified by the script
2126 (kernel; module("name"); process("file")) plus those listed by the
2127 stap "-d BINARY" option. Now, this data is included only if the systemtap
2128 script uses tapset functions like probefunc() or backtrace() that require
2129 such information. This shrinks the probe modules considerably for the rest.
2131 - Per-pass verbosity control is available with the new "--vp {N}+" option.
2132 "stap --vp 040" adds 4 units of -v verbosity only to pass 2. This is useful
2133 for diagnosing errors from one pass without excessive verbosity from others.
2135 - Most probe handlers now run with interrupts enabled, for improved
2136 system responsiveness and less probing overhead. This may result
2137 in more skipped probes, for example if a reentrant probe handler
2138 is attempted from within an interrupt handler. It may also make the
2139 systemtap overload detection facility more likely to be triggered, as
2140 interrupt handlers' run time would be included in the self-assessed
2141 overhead of running probe handlers.
2143 * What's new in version 0.8, 2008-11-13
2145 - Cache limiting is now available. If the compiled module cache size is
2146 over a limit specified in the $SYSTEMTAP_DIR/cache/cache_mb_limit file,
2147 some old cache entries will be unlinked. See man stap(1) for more.
2149 - Error and warning messages are now followed by source context displaying
2150 the erroneous line/s and a handy '^' in the following line pointing to the
2153 - A bug reporting tool "stap-report" is now available which will quickly
2154 retrieve much of the information requested here:
2155 http://sourceware.org/systemtap/wiki/HowToReportBugs
2157 - The translator can resolve members of anonymous structs / unions:
2158 given struct { int foo; struct { int bar; }; } *p;
2159 this now works: $p->bar
2161 - The stap "-F" flag activates "flight recorder" mode, which consists of
2162 translating the given script as usual, but implicitly launching it into
2163 the background with staprun's existing "-L" (launch) option. A user
2164 can later reattach to the module with "staprun -A MODULENAME".
2166 - Additional context variables are available on user-space syscall probes.
2167 - $argN ($arg1, $arg2, ... $arg6) in process(PATH_OR_PID).syscall
2168 gives you the argument of the system call.
2169 - $return in process(PATH_OR_PID).syscall.return gives you the return
2170 value of the system call.
2172 - Target process mode (stap -c CMD or -x PID) now implicitly restricts all
2173 "process.*" probes to the given child process. (It does not affect
2174 kernel.* or other probe types.) The CMD string is normally run directly,
2175 rather than via a /bin/sh -c subshell, since then utrace/uprobe probes
2176 receive a fairly "clean" event stream. If metacharacters like
2177 redirection operators were present in CMD, then "sh -c CMD" is still
2178 used, and utrace/uprobe probes will receive events from the shell.
2180 % stap -e 'probe process.syscall, process.end {
2181 printf("%s %d %s\n", execname(), pid(), pp())}'\
2183 ls 2323 process.syscall
2184 ls 2323 process.syscall
2187 - Probe listing mode is improved: "-L" lists available script-level variables
2189 % stap -L 'syscall.*open*'
2190 syscall.mq_open name:string name_uaddr:long filename:string mode:long u_attr_uaddr:long oflag:long argstr:string
2191 syscall.open name:string filename:string flags:long mode:long argstr:string
2192 syscall.openat name:string filename:string flags:long mode:long argstr:string
2194 - All user-space-related probes support $PATH-resolved executable
2197 probe process("ls").syscall {}
2198 probe process("./a.out").syscall {}
2200 work now, instead of just
2202 probe process("/bin/ls").syscall {}
2203 probe process("/my/directory/a.out").syscall {}
2205 - Prototype symbolic user-space probing support:
2207 # stap -e 'probe process("ls").function("*").call {
2208 log (probefunc()." ".$$parms)
2213 - debugging information for the named program
2214 - a version of utrace in the kernel that is compatible with the "uprobes"
2215 kernel module prototype. This includes RHEL5 and older Fedora, but not
2216 yet current lkml-track utrace; a "pass 4a"-time build failure means
2217 your system cannot use this yet.
2219 - Global variables which are written to but never read are now
2220 automatically displayed when the session does a shutdown. For example:
2222 global running_tasks
2223 probe timer.profile {running_tasks[pid(),tid()] = execname()}
2224 probe timer.ms(8000) {exit()}
2226 - A formatted string representation of the variables, parameters, or local
2227 variables at a probe point is now supported via the special $$vars,
2228 $$parms, and $$locals context variables, which expand to a string
2229 containing a list "var1=0xdead var2=0xbeef var3=?". (Here, var3 exists
2230 but is for some reason unavailable.) In return probes only, $$return
2231 expands to an empty string for a void function, or "return=0xf00".
2234 * What's new in version 0.7, 2008-07-15
2236 - .statement("func@file:*") and .statement("func@file:M-N") probes are now
2237 supported to allow matching a range of lines in a function. This allows
2238 tracing the execution of a function.
2240 - Scripts relying on probe point wildcards like "syscall.*" that expand
2241 to distinct kprobes are processed significantly faster than before.
2243 - The vector of script command line arguments is available in a
2244 tapset-provided global array argv[]. It is indexed 1 ... argc,
2245 another global. This can substitute for of preprocessor
2246 directives @NNN that fail at parse time if there are not
2249 printf("argv: %s %s %s", argv[1], argv[2], argv[3])
2251 - .statement("func@file+line") probes are now supported to allow a
2252 match relative to the entry of the function incremented by line
2253 number. This allows using the same systemtap script if the rest
2254 of the file.c source only changes slightly.
2256 - A probe listing mode is available.
2265 - More user-space probe types are added:
2267 probe process(PID).begin { }
2268 probe process("PATH").begin { }
2269 probe process(PID).thread.begin { }
2270 probe process("PATH").thread.begin { }
2271 probe process(PID).end { }
2272 probe process("PATH").end { }
2273 probe process(PID).thread.end { }
2274 probe process("PATH").thread.end { }
2275 probe process(PID).syscall { }
2276 probe process("PATH").syscall { }
2277 probe process(PID).syscall.return { }
2278 probe process("PATH").syscall.return { }
2280 - Globals now accept ; terminators
2283 global little[10], big[5];
2285 * What's new in version 0.6, 2007-12-15
2287 - A copy of the systemtap tutorial and language reference guide
2290 - There is a new format specifier, %m, for the printf family of
2291 functions. It functions like %s, except that it does not stop when
2292 a nul ('\0') byte is encountered. The number of bytes output is
2293 determined by the precision specifier. The default precision is 1.
2296 printf ("%m", "My String") // prints one character: M
2297 printf ("%.5", myString) // prints 5 bytes beginning at the start
2300 - The %b format specifier for the printf family of functions has been enhanced
2303 1) When the width and precision are both unspecified, the default is %8.8b.
2304 2) When only one of the width or precision is specified, the other defaults
2305 to the same value. For example, %4b == %.4b == %4.4b
2306 3) Nul ('\0') bytes are used for field width padding. For example,
2308 printf ("%b", 0x1111deadbeef2222) // prints all eight bytes
2309 printf ("%4.2b", 0xdeadbeef) // prints \0\0\xbe\xef
2311 - Dynamic width and precision are now supported for all printf family format
2312 specifiers. For example:
2316 printf ("%*.*b", four, two, 0xdeadbbeef) // prints \0\0\xbe\xef
2317 printf ("%*d", four, two) // prints <space><space><space>2
2319 - Preprocessor conditional expressions can now include wildcard style
2320 matches on kernel versions.
2321 %( kernel_vr != "*xen" %? foo %: bar %)
2323 - Prototype support for user-space probing is showing some progress.
2324 No symbolic notations are supported yet (so no probing by function names,
2325 file names, process names, and no access to $context variables), but at
2326 least it's something:
2328 probe process(PID).statement(ADDRESS).absolute { }
2330 This will set a uprobe on the given process-id and given virtual address.
2331 The proble handler runs in kernel-space as usual, and can generally use
2332 existing tapset functions.
2334 - Crash utility can retrieve systemtap's relay buffer from a kernel dump
2335 image by using staplog which is a crash extension module. To use this
2336 feature, type commands as below from crash(8)'s command line:
2338 crash> extend staplog.so
2339 crash> help systemtaplog
2341 Then, you can see more precise help message.
2343 - You can share a relay buffer amoung several scripts and merge outputs from
2344 several scripts by using "-DRELAY_HOST" and "-DRELAY_GUEST" options.
2348 % stap -ve 'probe begin{}' -o merged.out -DRELAY_HOST &
2349 # wait until starting the host.
2350 % stap -ve 'probe begin{print("hello ");exit()}' -DRELAY_GUEST
2351 % stap -ve 'probe begin{print("world\n");exit()}' -DRELAY_GUEST
2353 Then, you'll see "hello world" in merged.out.
2355 - You can add a conditional statement for each probe point or aliase, which
2356 is evaluated when the probe point is hit. If the condition is false, the
2357 whole probe body(including aliases) is skipped. For example:
2360 probe syscall.* if (switch) { ... }
2361 probe procfs.write {switch = strtol($value,10)} /* enable/disable ctrl */
2363 - Systemtap will warn you if your script contains unused variables or
2364 functions. This is helpful in case of misspelled variables. If it
2365 doth protest too much, turn it off with "stap -w ...".
2367 - You can add error-handling probes to a script, which are run if a
2368 script was stopped due to errors. In such a case, "end" probes are
2369 not run, but "error" ones are.
2371 probe error { println ("oops, errors encountered; here's a report anyway")
2372 foreach (coin in mint) { println (coin) } }
2374 - In a related twist, one may list probe points in order of preference,
2375 and mark any of them as "sufficient" beyond just "optional". Probe
2376 point sequence expansion stops if a sufficient-marked probe point has a hit.
2377 This is useful for probes on functions that may be in a module (CONFIG_FOO=m)
2378 or may have been compiled into the kernel (CONFIG_FOO=y), but we don't know
2381 probe module("sd").function("sd_init_command") ? ,
2382 kernel.function("sd_init_command") ? { ... }
2384 which might match neither, now one can write this:
2386 probe module("sd").function("sd_init_command") ! , /* <-- note excl. mark */
2387 kernel.function("sd_init_command") { ... }
2389 - New security model. To install a systemtap kernel module, a user
2390 must be one of the following: the root user; a member of the
2391 'stapdev' group; or a member of the 'stapusr' group. Members of the
2392 stapusr group can only use modules located in the
2393 /lib/modules/VERSION/systemtap directory (where VERSION is the
2394 output of "uname -r").
2396 - .statement("...@file:line") probes now apply heuristics to allow an
2397 approximate match for the line number. This works similarly to gdb,
2398 where a breakpoint placed on an empty source line is automatically
2399 moved to the next statement. A silly bug that made many $target
2400 variables inaccessible to .statement() probes was also fixed.
2402 - LKET has been retired. Please let us know on <systemtap@sourceware.org>
2403 if you have been a user of the tapset/tools, so we can help you find
2406 - New families of printing functions println() and printd() have been added.
2407 println() is like print() but adds a newline at the end;
2408 printd() is like a sequence of print()s, with a specified field delimiter.
2410 * What's new since version 0.5.14?, 2007-07-03
2412 - The way in which command line arguments for scripts are substituted has
2413 changed. Previously, $1 etc. would interpret the corresponding command
2414 line argument as an numeric literal, and @1 as a string literal. Now,
2415 the command line arguments are pasted uninterpreted wherever $1 etc.
2416 appears at the beginning of a token. @1 is similar, but is quoted as
2417 a string. This change does not modify old scripts, but has the effect
2418 of permitting substitution of arbitrary token sequences.
2420 # This worked before, and still does:
2421 % stap -e 'probe timer.s($1) {}' 5
2422 # Now this also works:
2423 % stap -e 'probe syscall.$1 {log(@1)}' open
2424 # This won't crash, just signal a recursion error:
2426 # As before, $1... is recognized only at the beginning of a token
2427 % stap -e 'probe begin {foo$1=5}'
2429 * What's new since version 0.5.13?, 2007-03-26
2431 - The way in which systemtap resolves function/inline probes has changed:
2432 .function(...) - now refers to all functions, inlined or not
2433 .inline(...) - is deprecated, use instead:
2434 .function(...).inline - filters function() to only inlined instances
2435 .function(...).call - filters function() to only non-inlined instances
2436 .function(...).return - as before, but now pairs best with .function().call
2437 .statement() is unchanged.
2439 * What's new since version 0.5.12?, 2007-01-01
2441 - When running in -p4 (compile-only) mode, the compiled .ko file name
2442 is printed on standard output.
2444 - An array element with a null value such as zero or an empty string
2445 is now preserved, and will show up in a "foreach" loop or "in" test.
2446 To delete such an element, the scripts needs to use an explicit
2447 "delete array[idx]" statement rather than something like "array[idx]=0".
2449 - The new "-P" option controls whether prologue searching heuristics
2450 will be activated for function probes. This was needed to get correct
2451 debugging information (dwarf location list) data for $target variables.
2452 Modern compilers (gcc 4.1+) tend not to need this heuristic, so it is
2453 no longer default. A new configure flag (--enable-prologues) restores
2454 it as a default setting, and is appropriate for older compilers (gcc 3.*).
2456 - Each systemtap module prints a one-line message to the kernel informational
2457 log when it starts. This line identifies the translator version, base
2458 address of the probe module, a broken-down memory consumption estimate, and
2459 the total number of probes. This is meant as a debugging / auditing aid.
2461 - Begin/end probes are run with interrupts enabled (but with
2462 preemption disabled). This will allow begin/end probes to be
2463 longer, to support generating longer reports.
2465 - The numeric forms of kernel.statement() and kernel.function() probe points
2466 are now interpreted as relocatable values - treated as relative to the
2467 _stext symbol in that kernel binary. Since some modern kernel images
2468 are relocated to a different virtual address at startup, such addresses
2469 may shift up or down when actually inserted into a running kernel.
2471 kernel.statement(0xdeadbeef): validated, interpreted relative to _stext,
2472 may map to 0xceadbeef at run time.
2474 In order to specify unrelocated addresses, use the new ".absolute"
2475 probe point suffix for such numeric addresses. These are only
2476 allowed in guru mode, and provide access to no $target variables.
2477 They don't use debugging information at all, actually.
2479 kernel.statement(0xfeedface).absolute: raw, unvalidated, guru mode only
2481 * What's new since version 0.5.10?, 2006-10-19
2483 - Offline processing of debugging information, enabling general
2484 cross-compilation of probe scripts to remote hosts, without
2485 requiring identical module/memory layout. This slows down
2486 compilation/translation somewhat.
2488 - Kernel symbol table data is loaded by staprun at startup time
2489 rather than compiled into the module.
2491 - Support the "limit" keyword for foreach iterations:
2492 foreach ([x,y] in ary limit 5) { ... }
2493 This implicitly exits after the fifth iteration. It also enables
2494 more efficient key/value sorting.
2496 - Support the "maxactive" keyword for return probes:
2497 probe kernel.function("sdfsdf").maxactive(848) { ... }
2498 This allows up to 848 concurrently outstanding entries to
2499 the sdfsdf function before one returns. The default maxactive
2500 number is smaller, and can result in missed return probes.
2502 - Support accessing of saved function arguments from within
2503 return probes. These values are saved by a synthesized
2504 function-entry probe.
2506 - Add substantial version/architecture checking in compiled probes to
2507 assert correct installation of debugging information and correct
2508 execution on a compatible kernel.
2510 - Add probe-time checking for sufficient free stack space when probe
2511 handlers are invoked, as a safety improvement.
2513 - Add an optional numeric parameter for begin/end probe specifications,
2514 to order their execution.
2515 probe begin(10) { } /* comes after */ probe begin(-10) {}
2517 - Add an optional array size declaration, which is handy for very small
2519 global little[5], big[20000]
2521 - Include some example scripts along with the documentation.
2523 - Change the start-time allocation of probe memory to avoid causing OOM
2524 situations, and to abort cleanly if free kernel memory is short.
2526 - Automatically use the kernel DWARF unwinder, if present, for stack
2529 - Many minor bug fixes, performance, tapset, and error message