[lkml]   [2017]   [Feb]   [17]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Subjectsystemtap 3.1 release
The SystemTap team announces release 3.1!

Highlights include syscall probes default to non-dwarf fallback,
python function probes, @variance and statistics optimizations,
java probe argument generalization, and user-space value-setting

= Where to get it - our project page
git tag release-3.1 (commit b8ea350dc13adb)

There have been 950ish commits since the last release.
There have been 84 "features" removed/features added.

= How to build it

See the README and NEWS files at;a=tree

Further information at

- Systemtap now needs C++11 to build

= SystemTap frontend (stap) changes

- Systemtap now warns if script arguments given on the command line are unused,
instead of mentioned by the script with $n/@n.

- New -T option allows the script to be terminated after a specified number
of seconds. This is a shortcut for adding the probe, timer.s(N) {exit()}.

= SystemTap script language changes

- Support has been added for probing python 2 and 3 functions using a
custom python helper module. Python function probes can target
function entry, returns, or specific line numbers.

probe python2.module("myscript").function("foo")
{ println($$parms) }

To run with the custom python helper module, you'd use python's '-m'
option like the following:

stap myscript.stp -c "python -m HelperSDT"

- Java method probes now convert all types of java parameters to
strings using the java toString() method before passing them to
systemtap probes; new argN variables copy them into string
variables. Previously, only numeric types were passed, and only by
casting to integers. The previous behaviour is available with
--compatible=3.0 .

3.1: probe java(...).class(...).method(...) { printf("%s", arg1) }
3.0: probe java(...).class(...).method(...) { printf("%d", $arg1) }

- Context variables in .return probes should be accessed with @entry($var)
rather than $var, to make it clear that entry-time snapshots are being
used. The latter construct now generates a warning. Availability testing
with either @defined(@entry($var)) or @defined($var) works.

- New statistics @variance() operator using the Welford's online algorithm
for per-cpu computation, and the Total Variance formula authored by
Niranjan Kamat and Arnab Nandi from the Ohio State University for the
cross-cpu aggregation.

- The implementation of "var <<< X" for each aggregate variable is now
specially compiled to compute only the script-requested @op(var) values,
not all potential ones. This speeds up the <<< operations.

- Translator now accepts new @const() operator for convenient expressing
constants in tapset code, or guru-mode scripts. See stap(1) for details.

= SystemTap runtime changes

- An older defensive measure to suppress kernel kprobes optimizations
since the 3.x era has been disabled for recent kernels. This improves
the performance of kernel function probes. In case of related problems,
please report and work around with:
# echo 0 > /proc/sys/debug/kprobes-optimization

- New installcheck-parallel testsuite feature allows running the tests in
parallel in order to save time. See testsuite/README for details.

= SystemTap tapset changes

- Syscall and nd_syscall tapsets have been merged in a way that either
dwarf-based, or non-dwarf probe gets automatically used based on
debuginfo availability (e.g. probe

To force use the dwarf based probe, a dw_syscall has been introduced
(e.g. probe and the non-dwarf syscall probes were
left untouched (e.g.

- The syscall tapset files have been reorganized in a way that original
big tapset files carrying many syscall probes were split into smaller
'sysc_' prefixed tapset files. This should reduce the syscall tapset
maintenance burden.

- The powerpc variant of syscall.compat_sysctl got deprecated on favor of
syscall.sysctl32. This aligns the syscall to its respective nd_syscall and
to ia64/s390/x86_64 variants too.

- The syscall.compat_pselect7a (this was actually a typo, but still available
for compatibility purposes with --compatible 1.3) has beed deprecated.

- The 'description_auddr' convenience variable of syscall.add_key has been

- Tapsets containing process probes may now be placed in the special
$prefix/share/systemtap/tapset/PATH/ directory to have their process parameter
prefixed with the location of the tapset. For example,

process("foo").function("NAME") expands to
when placed in $prefix/share/systemtap/tapset/PATH/usr/bin/

This is intended to help write more reusable tapsets for userspace binaries.

- Netfilter tapsets now provide variables data_hex and data_str to
display packet
contents in hexadecimal and ASCII respectively.

- New tapset functions set_user_string(), set_user_string_n(), set_user_long()
set_user_int(), set_user_short(), set_user_char() and set_user_pointer() to
write a value of specified type directly to a user space address.

- New tapset functions user_buffer_quoted(), user_buffer_quoted_error(),
kernel_buffer_quoted(), and kernel_buffer_quoted_error() to print a
buffer of an exact length. These functions can handle '\0' characters
as well.

= SystemTap sample scripts

All 163 examples can be found at

- New Samples:

socket-events.stp Prints the life cycle of all sockets
associated with a process.
This includes bytes and timing. The timing
information that is
tracked includes event completion relative
to the start of said
event and the end of the previous event.
Currently tracks read,
write, recv, send, connect and close.

nfsd-trace.stp This script traces all nfsd server
operations by client_ip address,
operation, and complete file name (if possible).

packet_contents.stp The packet_contents.stp script displays the
length of each
network packet and its contents in both
hexadecimal and ASCII.
Systemtap strings are MAXSTRINGLEN in
length by default which
may not be enough for larger packets. In
order to print larger
packets, this limit can be increased by
passing in the
"-DMAXSTRINGLEN=65536" command line option.

tcp_retransmission.stp The tcp_retransmission.stp prints out a
line for each tcp
retransmission packet.

sched-latency.stp This script periodically reports a
histogram of the latency
between a task (thread) being woken up and
it actually being
dispatched to a CPU: the amount of time
it's spent in the
runnable queue.

container_check.stp The container_check.stp script monitors the
use of linux
capablities and optionally forbidden
syscalls by a process
and its children. On exit the script
prints out lists
showing the capabilies used by each
executable, which syscall
used specific capabilites for each
executable, a list of
forbidden syscalls used, and details on any
syscalls that
failed during monitoring. This script is
designed to help
diagnose issues caused by restricted
capabilies and syscalls
when running an application in a container.
If the script
warns about skipped probes, the number of
active kretprobes
may need to be increased with
"-DKRETACTIVE=100" option on the
command line.

cve-2016-5195.stp historical emergency security band-aid, for
reference/education only.

- New command within interactive mode, sample. Allows you to search through
all included example scripts to load for further editing or running. Sample
and example scripts have been moved to /usr/share/systemtap/examples. A
symlink in the former location under $docdir links to it.

= Examples of tested kernel versions

2.6.18 (RHEL 5 x86 and x86_64)
2.6.32 (RHEL 6 x86 and x86_64)
3.10.0 (RHEL 7 x86_64)
4.1.6 (Fedora 22 x86_64)
4.3.4 (Fedora 22 x86_64)
4.6.0-rc0 (Fedora rawhide x86_64)
4.6.0-rc6 (Fedora rawhide x86_64)
4.8.10-200 (Fedora 24 x86_64)
4.10.0-rc0 (Fedora rawhide x86_64)
4.10.0-rc6 (Fedora rawhide x86_64)
4.10.0-rc8 (Fedora rawhide x86_64)

= Known issues with this release

- Some kernel crashes continue to be reported when a script probes
broad kernel function wildcards. (PR2725)

- An upstream kernel commit #2062afb4f804a put "-fno-var-tracking-assignments"
into KCFLAGS, reducing debuginfo quality which can cause debuginfo failures.
A proposed workaround to this issue exists in: . Fedora kernels are not affected by
this issue.

= Contributors for this release

Abegail Jakop, Alexander Lochmann, Benjamin Coddington*, Bingwu
Yang*, Cody Santing*, David Smith, Felix Lu, Francis Giraldeau*,
Frank Ch. Eigler, Hemant Kumar, Igor Zhbanov*, Joe Gorse*, Josh
Stone, Kyle Walker*, Lukas Berk, Marcin Nowakowski*, Mark Wielaard,
Martin Cermak, Masanari Iida, Mateusz Guzik*, Michal Toman*, Nikolay
Borisov, Petr Matousek, Ravi Bangoria*, Ross Burton*, Tetsuo Handa,
Torsten Polle, William Cohen

Special thanks to new contributors, marked with '*' above.
Special thanks to Cody Santing for drafting these notes.

= Bugs fixed for this release <>

6978 process.syscall extensions: abort, $$parms
10234 clean up aggregate hard-coded logic
10485 auto-path tapset support for process.* probes
10655 SDT semaphores should be prepared for multiple tasks per probe
10791 parallelize systemtap testsuite
11308 aggregate operations for @variance, @skew, @kurtosis
11637 set_user_* functions
12596 blacklist is too broad (raw_.*)
12748 need syscall-number database in tapset
14787 consider making stap -L output prettier/more structured
14924 warn on complex $ptr->foo expressions in .return probes
15076 Merge MIPS patches from Cisco
15671 systemtap (rpm version) can't find debuginfo for @var() use
15932 %m/%M should have a variant that reads user memory (instead of
kernel memory)
17055 _stp_perf_read needs a sleepable context
17231 sysroot is too often prepended
17962 dtrace.exp --no-parsing fallback test fails on rhel6
18079 autocast doesn't work with @defined
19489 printing array from memory
19624 Duplicate function parameter names are not detected
19802 bad hash value distribution and horrible performance for large
arrays with multiple small-integer indices
19873 staprun -o /NO/SUCH/FILE -c CMD imperfect cleanup
19874 stap -c CMD run-time limited to 60s due to uncleared alarm()
19875 membarrier missing from syscall tapset
19876 userfaultfd missing from syscall tapset
19882 copy_file_range missing from syscall tapset
19905 preadv2/pwritev2 missing from syscall tapset
19906 file name lookups in vfs etc. tapsets
19915 flight recorder's "logrorate" feature broken
19926 we need a better way to express constants in tapset code
19940 page_cache_release() missing from the latest rawhide kernel
19953 netfilter tapsets should provide variables to assist printing
of packet contents
19954 "suspicious RCU usage" message on rawhide
19990 on rawhide, the get_user_pages() function has changed
19992 polymorphic operation
20013 stap --dump-functions broken
20040 the task_exe_file function getting "BUG: sleeping function
called from invalid context"
20042 on rawhide, tracepoint handlers have a changed function signature
20056 improve parse error message involving expect_op("...")
20064 Linking stapio failed because of misplaced libraries flags
20065 Configure script is not in sync with
20122 use base os toolchain consistently in the developer toolset environment
20131 listing_mode.exp wildcard library path failures
20132 on rawhide, struct inode has changed
20136 Use the @const() operator across the tapset scripts.
20149 a function probe with a line number acts like a statement probe
20158 on kernel 4.6, print_backtrace() gets a compile error
20161 VM_FAULT_MINOR has been removed from rawhide kernels
20187 on rawhide, the 'size' convience variable of socket.recvmsg doesn't work
20189 on rawhide, PAGE_CACHE_SIZE is no longer defined (which
breaks the vfs tapset)
20192 "suspicious RCU usage." warning from kernel when running testsuite
20211 testsuite resume feature
20217 warn for degenerate case overloaded functions
20236 code cleanup: simplify user/kernel memory access routines
20281 probe process("") kills stap with SIGABRT
20282 implicit declaration of function ‘__get_user_bad’ on recent
aarch64 kernel
20286 probe handlers using hrtimers taking too long
20298 the unprivileged_embedded_C.exp testcase needs updating
20307 'private' on tapset global arrays causes errors
20333 merge syscall and nd_syscall tapsets
20416 @entry(@perf("foo")) not translated correctly
20423 improve error message for dwarf $var 'struct ... being
accessed instead of member'
20433 "NULL pointer dereference" crash on fedora
20504 trouble finding some tracepoints on kernel 4.7+
20510 stap -L colorizes non-tty stdout
20589 kernel warning from calling kernel_buffer_quoted()
20594 Compile error on GCC 6.1.1: misleading indentation
20597 broken @avg() calculations
20599 histogram breaks @variance
20601 __get_skb_iphdr() failing on 32-bit rawhide
20672 @defined(@cast()) regression
20735 "soft lockup" bug on RHEL7 ppc64
20820 another "soft lockup" BUG on RHEL7 ppc64
20821 @defined(@entry($var)) does not nest correctly
20850 The systemtap boot time probing feature doesn't work on rhel6
20879 For stap -t, print out global variable contention report
20889 metadatabase.db location
20982 function::stack doesn't descend if _stack_raw() fails
21020 reorganize argument passing from java probes
21063 dtrace script causes mysterious build failures due to improper
forming of gcc command line
21065 dtrace script reports syntax error for valid .d files
21101 errors when compiling a systemtap module with gcc 7
21102 the ioblock.stp tapset needs to be updated
21105 syscall testsuite failures on rawhide

 \ /
  Last update: 2017-02-17 19:33    [W:0.042 / U:2.504 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site