1 .\" Copyright (c) 1991, 1992 Paul Kranenburg <pk@cs.few.eur.nl>
2 .\" Copyright (c) 1993 Branko Lankester <branko@hacktic.nl>
3 .\" Copyright (c) 1993, 1994, 1995, 1996 Rick Sladkey <jrs@world.std.com>
4 .\" All rights reserved.
6 .\" Redistribution and use in source and binary forms, with or without
7 .\" modification, are permitted provided that the following conditions
9 .\" 1. Redistributions of source code must retain the above copyright
10 .\" notice, this list of conditions and the following disclaimer.
11 .\" 2. Redistributions in binary form must reproduce the above copyright
12 .\" notice, this list of conditions and the following disclaimer in the
13 .\" documentation and/or other materials provided with the distribution.
14 .\" 3. The name of the author may not be used to endorse or promote products
15 .\" derived from this software without specific prior written permission.
17 .\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 .\" IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 .\" OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 .\" IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 .\" INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 .\" NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 .\" DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 .\" THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 .\" (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 .\" THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
37 .\" Macro IX is not defined in the groff macros
42 .TH STRACE 1 "2010-03-30"
44 strace \- trace system calls and signals
47 [\fB-CdffhikqrtttTvVxxy\fR]
54 [\fB-P\fIpath\fR]... \fB-p\fIpid\fR... /
56 [\fB-E\fIvar\fR[=\fIval\fR]]... [\fB-u\fIusername\fR]
57 \fIcommand\fR [\fIargs\fR]
65 [\fB-S\fIsortby\fR] \fB-p\fIpid\fR... /
67 [\fB-E\fIvar\fR[=\fIval\fR]]... [\fB-u\fIusername\fR]
68 \fIcommand\fR [\fIargs\fR]
70 .IX "strace command" "" "\fLstrace\fR command"
77 It intercepts and records the system calls which are called
78 by a process and the signals which are received by a process.
79 The name of each system call, its arguments and its return value
80 are printed on standard error or to the file specified with the
85 is a useful diagnostic, instructional, and debugging tool.
86 System administrators, diagnosticians and trouble-shooters will find
87 it invaluable for solving problems with
88 programs for which the source is not readily available since
89 they do not need to be recompiled in order to trace them.
90 Students, hackers and the overly-curious will find that
91 a great deal can be learned about a system and its system calls by
92 tracing even ordinary programs. And programmers will find that
93 since system calls and signals are events that happen at the user/kernel
94 interface, a close examination of this boundary is very
95 useful for bug isolation, sanity checking and
96 attempting to capture race conditions.
98 Each line in the trace contains the system call name, followed
99 by its arguments in parentheses and its return value.
100 An example from stracing the command "cat /dev/null" is:
102 open("/dev/null", O_RDONLY) = 3
104 Errors (typically a return value of \-1) have the errno symbol
105 and error string appended.
107 open("/foo/bar", O_RDONLY) = \-1 ENOENT (No such file or directory)
109 Signals are printed as signal symbol and decoded siginfo structure.
110 An excerpt from stracing and interrupting the command "sleep 666" is:
112 sigsuspend([] <unfinished ...>
113 --- SIGINT {si_signo=SIGINT, si_code=SI_USER, si_pid=...} ---
114 +++ killed by SIGINT +++
116 If a system call is being executed and meanwhile another one is being called
117 from a different thread/process then
119 will try to preserve the order of those events and mark the ongoing call as
122 When the call returns it will be marked as
125 [pid 28772] select(4, [3], NULL, NULL, NULL <unfinished ...>
126 [pid 28779] clock_gettime(CLOCK_REALTIME, {1130322148, 939977000}) = 0
127 [pid 28772] <... select resumed> ) = 1 (in [3])
129 Interruption of a (restartable) system call by a signal delivery is processed
130 differently as kernel terminates the system call and also arranges its
131 immediate reexecution after the signal handler completes.
133 read(0, 0x7ffff72cf5cf, 1) = ? ERESTARTSYS (To be restarted)
135 rt_sigreturn(0xe) = 0
138 Arguments are printed in symbolic form with a passion.
139 This example shows the shell performing ">>xyzzy" output redirection:
141 open("xyzzy", O_WRONLY|O_APPEND|O_CREAT, 0666) = 3
143 Here the third argument of open is decoded by breaking down the
144 flag argument into its three bitwise-OR constituents and printing the
145 mode value in octal by tradition. Where traditional or native
146 usage differs from ANSI or POSIX, the latter forms are preferred.
149 output has proven to be more readable than the source.
151 Structure pointers are dereferenced and the members are displayed
152 as appropriate. In all cases arguments are formatted in the most C-like
154 For example, the essence of the command "ls \-l /dev/null" is captured as:
156 lstat("/dev/null", {st_mode=S_IFCHR|0666, st_rdev=makedev(1, 3), ...}) = 0
158 Notice how the 'struct stat' argument is dereferenced and how each member is
159 displayed symbolically. In particular, observe how the st_mode member
160 is carefully decoded into a bitwise-OR of symbolic and numeric values.
161 Also notice in this example that the first argument to lstat is an input
162 to the system call and the second argument is an output. Since output
163 arguments are not modified if the system call fails, arguments may not
164 always be dereferenced. For example, retrying the "ls \-l" example
165 with a non-existent file produces the following line:
167 lstat("/foo/bar", 0xb004) = \-1 ENOENT (No such file or directory)
169 In this case the porch light is on but nobody is home.
171 Character pointers are dereferenced and printed as C strings.
172 Non-printing characters in strings are normally represented by
173 ordinary C escape codes.
176 (32 by default) bytes of strings are printed;
177 longer strings have an ellipsis appended following the closing quote.
178 Here is a line from "ls \-l" where the
180 library routine is reading the password file:
182 read(3, "root::0:0:System Administrator:/"..., 1024) = 422
184 While structures are annotated using curly braces, simple pointers
185 and arrays are printed using square brackets with commas separating
186 elements. Here is an example from the command "id" on a system with
187 supplementary group ids:
189 getgroups(32, [100, 0]) = 2
191 On the other hand, bit-sets are also shown using square brackets
192 but set elements are separated only by a space. Here is the shell
193 preparing to execute an external command:
195 sigprocmask(SIG_BLOCK, [CHLD TTOU], []) = 0
197 Here the second argument is a bit-set of two signals, SIGCHLD and SIGTTOU.
198 In some cases the bit-set is so full that printing out the unset
199 elements is more valuable. In that case, the bit-set is prefixed by
202 sigprocmask(SIG_UNBLOCK, ~[], NULL) = 0
204 Here the second argument represents the full set of all signals.
209 Count time, calls, and errors for each system call and report a summary on
210 program exit. On Linux, this attempts to show system time (CPU time spent
211 running in the kernel) independent of wall clock time. If
217 (below), only aggregate totals for all traced processes are kept.
222 but also print regular output while processes are running.
225 Run tracer process as a detached grandchild, not as parent of the
226 tracee. This reduces the visible effect of
228 by keeping the tracee a direct child of the calling process.
231 Show some debugging output of
233 itself on the standard error.
236 Trace child processes as they are created by currently traced
237 processes as a result of the
242 system calls. Note that
246 will attach all threads of process PID if it is multi-threaded,
247 not only thread with thread_id = PID.
253 option is in effect, each processes trace is written to
255 where pid is the numeric process id of each process.
256 This is incompatible with
258 since no per-process counts are kept.
261 This option is now obsolete and it has the same functionality as
265 Print the help summary.
268 Print the instruction pointer at the time of the system call.
271 Print the execution stack trace of the traced processes after each system call (experimental).
272 This option is available only if
274 is built with libunwind.
277 Suppress messages about attaching, detaching etc. This happens
278 automatically when output is redirected to a file and the command
279 is run directly instead of attaching.
282 If given twice, suppress messages about process exit status.
285 Print a relative timestamp upon entry to each system call. This
286 records the time difference between the beginning of successive
290 Prefix each line of the trace with the time of day.
293 If given twice, the time printed will include the microseconds.
296 If given thrice, the time printed will include the microseconds
297 and the leading portion will be printed as the number
298 of seconds since the epoch.
301 Show the time spent in system calls. This records the time
302 difference between the beginning and the end of each system call.
305 Summarise the time difference between the beginning and end of
306 each system call. The default is to summarise the system time.
309 Print unabbreviated versions of environment, stat, termios, etc.
310 calls. These structures are very common in calls and so the default
311 behavior displays a reasonable subset of structure members. Use
312 this option to get all of the gory details.
315 Print the version number of
319 Print all non-ASCII strings in hexadecimal string format.
322 Print all strings in hexadecimal string format.
325 Print paths associated with file descriptor arguments.
328 Print protocol specific information associated with socket file descriptors.
331 Align return values in a specific column (default column 40).
334 If specified syscall is reached, detach from traced process.
337 syscall is supported. This option is useful if you want to trace
338 multi-threaded process and therefore require -f, but don't want
339 to trace its (potentially very complex) children.
342 A qualifying expression which modifies which events to trace
343 or how to trace them. The format of the expression is:
346 [\,\fIqualifier\/\fB=\fR][\fB!\fR]\,\fIvalue1\/\fR[\fB,\,\fIvalue2\/\fR]...
363 is a qualifier-dependent symbol or number. The default
366 Using an exclamation mark negates the set of values. For example,
369 .BR \-e "\ " trace = open
370 which in turn means trace only the
372 system call. By contrast,
373 .BR \-e "\ " trace "=!" open
374 means to trace every system call except
376 In addition, the special values
380 have the obvious meanings.
382 Note that some shells use the exclamation point for history
383 expansion even inside quoted arguments. If so, you must escape
384 the exclamation point with a backslash.
386 \fB\-e\ trace\fR=\,\fIset\fR
387 Trace only the specified set of system calls. The
389 option is useful for determining which system calls might be useful
390 to trace. For example,
391 .BR trace = open,close,read,write
393 trace those four system calls. Be careful when making inferences
394 about the user/kernel boundary if only a subset of system calls
395 are being monitored. The default is
398 .BR "\-e\ trace" = file
399 Trace all system calls which take a file name as an argument. You
400 can think of this as an abbreviation for
401 .BR "\-e\ trace" = open , stat , chmod , unlink ,...
402 which is useful to seeing what files the process is referencing.
403 Furthermore, using the abbreviation will ensure that you don't
404 accidentally forget to include a call like
406 in the list. Betchya woulda forgot that one.
408 .BR "\-e\ trace" = process
409 Trace all system calls which involve process management. This
410 is useful for watching the fork, wait, and exec steps of a process.
412 .BR "\-e\ trace" = network
413 Trace all the network related system calls.
415 .BR "\-e\ trace" = signal
416 Trace all signal related system calls.
418 .BR "\-e\ trace" = ipc
419 Trace all IPC related system calls.
421 .BR "\-e\ trace" = desc
422 Trace all file descriptor related system calls.
424 .BR "\-e\ trace" = memory
425 Trace all memory mapping related system calls.
427 \fB\-e\ abbrev\fR=\,\fIset\fR
428 Abbreviate the output from printing each member of large structures.
433 option has the effect of
436 \fB\-e\ verbose\fR=\,\fIset\fR
437 Dereference structures for the specified set of system calls. The
441 \fB\-e\ raw\fR=\,\fIset\fR
442 Print raw, undecoded arguments for the specified set of system calls.
443 This option has the effect of causing all arguments to be printed
444 in hexadecimal. This is mostly useful if you don't trust the
445 decoding or you need to know the actual numeric value of an
448 \fB\-e\ signal\fR=\,\fIset\fR
449 Trace only the specified subset of signals. The default is
455 causes SIGIO signals not to be traced.
457 \fB\-e\ read\fR=\,\fIset\fR
458 Perform a full hexadecimal and ASCII dump of all the data read from
459 file descriptors listed in the specified set. For example, to see
460 all input activity on file descriptors
465 \fB\-e\ read\fR=\,\fI3\fR,\fI5\fR.
466 Note that this is independent from the normal tracing of the
468 system call which is controlled by the option
469 .BR -e "\ " trace = read .
471 \fB\-e\ write\fR=\,\fIset\fR
472 Perform a full hexadecimal and ASCII dump of all the data written to
473 file descriptors listed in the specified set. For example, to see
474 all output activity on file descriptors
479 \fB\-e\ write\fR=\,\fI3\fR,\,\fI5\fR.
480 Note that this is independent from the normal tracing of the
482 system call which is controlled by the option
483 .BR -e "\ " trace = write .
485 \fB\-e\ fault\fR=\,\fIset\/\fR[:\fBerror\fR=\,\fIerrno\/\fR][:\fBwhen\fR=\,\fIexpr\/\fR]
486 Perform a syscall fault injection for the specified set of syscalls.
487 When a fault is injected into a syscall invocation, the syscall number
488 is replaced by -1 which corresponds to an invalid syscall.
490 If an error code is specified using a symbolic
494 or a numeric value within 1..4095 range, this error code overrides
495 the default error code returned by the kernel, which is traditionally
497 for invalid syscall numbers on most architectures.
499 Unless a :\fBwhen\fR=\,\fIexpr\fR subexpression is specified,
500 a fault is injected into every invocation of each syscall from the
503 The format of the subexpression is one of the following:
508 For every syscall from the
510 perform a syscall fault injection for the syscall invocation number
517 For every syscall from the
519 perform syscall fault injections for the syscall invocation number
521 and all subsequent invocations.
524 \fIfirst\/\fB+\fIstep\fR
526 For every syscall from the
528 perform syscall fault injections for syscall invocations number
531 .IR first + step + step ,
536 For example, to fail each third and subsequent chdir syscalls with
539 \fB\-e\ fault\fR=\,\fIchdir\/\fR:\fBerror\fR=\,\fIENOENT\/\fR:\fBwhen\fR=\,\fI3\/\fB+\fR.
541 The valid range for numbers
547 If a fault expression contains multiple
549 specifications, the last one takes precedence.
550 Likewise, if a fault expression contains multiple
552 specifications, the last one takes precedence.
554 Accounting of syscalls that are subject to fault injection
555 is done per syscall and per tracee.
557 Specification of syscall fault injection can be combined
558 with other syscall filtering options, for example,
559 \fB\-P \fI/dev/urandom \fB\-e fault\fR=\,\fIall\/\fR:\fBerror\fR=\,\fIENOENT\fR.
562 .BI "\-I " interruptible
563 When strace can be interrupted by signals (such as pressing ^C).
564 1: no signals are blocked; 2: fatal signals are blocked while decoding syscall
565 (default); 3: fatal signals are always blocked (default if '-o FILE PROG');
566 4: fatal signals and SIGTSTP (^Z) are always blocked (useful to make
567 strace -o FILE PROG not stop on ^Z).
570 Write the trace output to the file
572 rather than to stderr.
578 If the argument begins with '|' or with '!' then the rest of the
579 argument is treated as a command and all output is piped to it.
580 This is convenient for piping the debugging output to a program
581 without affecting the redirections of executed programs.
584 Set the overhead for tracing system calls to
587 This is useful for overriding the default heuristic for guessing
588 how much time is spent in mere measuring when timing system calls using
591 option. The accuracy of the heuristic can be gauged by timing a given
592 program run without tracing (using
594 and comparing the accumulated
595 system call time to the total produced using
599 Attach to the process with the process
603 The trace may be terminated
604 at any time by a keyboard interrupt signal (\c
607 will respond by detaching itself from the traced process(es)
608 leaving it (them) to continue running.
611 options can be used to attach to many processes in addition to
613 (which is optional if at least one
617 "`pidof PROG`" syntax is supported.
620 Trace only system calls accessing
624 options can be used to specify several paths.
627 Specify the maximum string size to print (the default is 32). Note
628 that filenames are not considered strings and are always printed in
632 Sort the output of the histogram printed by the
634 option by the specified criterion. Legal values are
644 Run command with the user \s-1ID\s0, group \s-2ID\s0, and
645 supplementary groups of
647 This option is only useful when running as root and enables the
648 correct execution of setuid and/or setgid binaries.
649 Unless this option is used setuid and setgid programs are executed
650 without effective privileges.
652 \fB\-E\ \fIvar\fR=\,\fIval\fR
655 in its list of environment variables.
660 from the inherited list of environment variables before passing it on to
667 exits with the same exit status.
670 is terminated by a signal,
672 terminates itself with the same signal, so that
674 can be used as a wrapper process transparent to the invoking parent process.
675 Note that parent-child relationship (signal stop notifications,
676 getppid() value, etc) between traced process and its parent are not preserved
687 is zero unless no processes has been attached or there was an unexpected error
688 in doing the tracing.
689 .SH "SETUID INSTALLATION"
692 is installed setuid to root then the invoking user will be able to
693 attach to and trace processes owned by any user.
694 In addition setuid and setgid programs will be executed and traced
695 with the correct effective privileges.
696 Since only users trusted with full root privileges should be allowed
698 it only makes sense to install
700 as setuid to root when the users who can execute it are restricted
701 to those users who have this trust.
702 For example, it makes sense to install a special version of
704 with mode 'rwsr-xr--', user
710 group are trusted users.
711 If you do use this feature, please remember to install
712 a non-setuid version of
714 for ordinary lusers to use.
721 It is a pity that so much tracing clutter is produced by systems
722 employing shared libraries.
724 It is instructive to think about system call inputs and outputs
725 as data-flow across the user/kernel boundary. Because user-space
726 and kernel-space are separate and address-protected, it is
727 sometimes possible to make deductive inferences about process
728 behavior using inputs and outputs as propositions.
730 In some cases, a system call will differ from the documented behavior
731 or have a different name. For example, on System V-derived systems
734 system call does not take an argument and the
738 and takes an extra leading argument. These
739 discrepancies are normal but idiosyncratic characteristics of the
740 system call interface and are accounted for by C library wrapper
743 Some system calls have different names in different architectures and
744 personalities. In these cases, system call filtering and printing
745 uses the names that match corresponding
747 kernel macros of the tracee's architecture and personality.
748 There are two exceptions from this general rule:
749 .BR arm_fadvise64_64 (2)
751 .BR xtensa_fadvise64_64 (2)
752 Xtensa syscall are filtered and printed as
753 .BR fadvise64_64 (2).
755 On some platforms a process that is attached to with the
757 option may observe a spurious EINTR return from the current
758 system call that is not restartable. (Ideally, all system calls
759 should be restarted on strace attach, making the attach invisible
760 to the traced process, but a few system calls aren't.
761 Arguably, every instance of such behavior is a kernel bug.)
762 This may have an unpredictable effect on the process
763 if the process takes no action to restart the system call.
765 Programs that use the
770 privileges while being traced.
772 A traced process runs slowly.
774 Traced processes which are descended from
776 may be left running after an interrupt signal (\c
781 option is weakly supported.
785 was written by Paul Kranenburg
786 for SunOS and was inspired by its trace utility.
789 was ported to Linux and enhanced
790 by Branko Lankester, who also wrote the Linux kernel support.
791 Even though Paul released
794 Branko's work was based on Paul's
796 1.5 release from 1991.
797 In 1993, Rick Sladkey merged
799 2.5 for SunOS and the second release of
801 for Linux, added many of the features of
803 from SVR4, and produced an
805 that worked on both platforms. In 1994 Rick ported
807 to SVR4 and Solaris and wrote the
808 automatic configuration support. In 1995 he ported
811 and tired of writing about himself in the third person.
815 should be reported to the
817 mailing list at <strace\-devel@lists.sourceforge.net>.