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.
40 .TH STRACE 1 "2010-03-30"
42 strace \- trace system calls and signals
46 .B \-CdDffhiqrtttTvxxy
111 .IX "strace command" "" "\fLstrace\fR command"
118 It intercepts and records the system calls which are called
119 by a process and the signals which are received by a process.
120 The name of each system call, its arguments and its return value
121 are printed on standard error or to the file specified with the
126 is a useful diagnostic, instructional, and debugging tool.
127 System administrators, diagnosticians and trouble-shooters will find
128 it invaluable for solving problems with
129 programs for which the source is not readily available since
130 they do not need to be recompiled in order to trace them.
131 Students, hackers and the overly-curious will find that
132 a great deal can be learned about a system and its system calls by
133 tracing even ordinary programs. And programmers will find that
134 since system calls and signals are events that happen at the user/kernel
135 interface, a close examination of this boundary is very
136 useful for bug isolation, sanity checking and
137 attempting to capture race conditions.
139 Each line in the trace contains the system call name, followed
140 by its arguments in parentheses and its return value.
141 An example from stracing the command ``cat /dev/null'' is:
143 open("/dev/null", O_RDONLY) = 3
145 Errors (typically a return value of \-1) have the errno symbol
146 and error string appended.
148 open("/foo/bar", O_RDONLY) = -1 ENOENT (No such file or directory)
150 Signals are printed as a signal symbol and a signal string.
151 An excerpt from stracing and interrupting the command ``sleep 666'' is:
153 sigsuspend([] <unfinished ...>
154 --- SIGINT (Interrupt) ---
155 +++ killed by SIGINT +++
157 If a system call is being executed and meanwhile another one is being called
158 from a different thread/process then
160 will try to preserve the order of those events and mark the ongoing call as
163 When the call returns it will be marked as
166 [pid 28772] select(4, [3], NULL, NULL, NULL <unfinished ...>
167 [pid 28779] clock_gettime(CLOCK_REALTIME, {1130322148, 939977000}) = 0
168 [pid 28772] <... select resumed> ) = 1 (in [3])
170 Interruption of a (restartable) system call by a signal delivery is processed
171 differently as kernel terminates the system call and also arranges its
172 immediate reexecution after the signal handler completes.
174 read(0, 0x7ffff72cf5cf, 1) = ? ERESTARTSYS (To be restarted)
175 --- SIGALRM (Alarm clock) @ 0 (0) ---
176 rt_sigreturn(0xe) = 0
177 read(0, ""..., 1) = 0
179 Arguments are printed in symbolic form with a passion.
180 This example shows the shell performing ``>>xyzzy'' output redirection:
182 open("xyzzy", O_WRONLY|O_APPEND|O_CREAT, 0666) = 3
184 Here the third argument of open is decoded by breaking down the
185 flag argument into its three bitwise-OR constituents and printing the
186 mode value in octal by tradition. Where traditional or native
187 usage differs from ANSI or POSIX, the latter forms are preferred.
190 output has proven to be more readable than the source.
192 Structure pointers are dereferenced and the members are displayed
193 as appropriate. In all cases arguments are formatted in the most C-like
195 For example, the essence of the command ``ls \-l /dev/null'' is captured as:
197 lstat("/dev/null", {st_mode=S_IFCHR|0666, st_rdev=makedev(1, 3), ...}) = 0
199 Notice how the `struct stat' argument is dereferenced and how each member is
200 displayed symbolically. In particular, observe how the st_mode member
201 is carefully decoded into a bitwise-OR of symbolic and numeric values.
202 Also notice in this example that the first argument to lstat is an input
203 to the system call and the second argument is an output. Since output
204 arguments are not modified if the system call fails, arguments may not
205 always be dereferenced. For example, retrying the ``ls \-l'' example
206 with a non-existent file produces the following line:
208 lstat("/foo/bar", 0xb004) = -1 ENOENT (No such file or directory)
210 In this case the porch light is on but nobody is home.
212 Character pointers are dereferenced and printed as C strings.
213 Non-printing characters in strings are normally represented by
214 ordinary C escape codes.
217 (32 by default) bytes of strings are printed;
218 longer strings have an ellipsis appended following the closing quote.
219 Here is a line from ``ls \-l'' where the
221 library routine is reading the password file:
223 read(3, "root::0:0:System Administrator:/"..., 1024) = 422
225 While structures are annotated using curly braces, simple pointers
226 and arrays are printed using square brackets with commas separating
227 elements. Here is an example from the command ``id'' on a system with
228 supplementary group ids:
230 getgroups(32, [100, 0]) = 2
232 On the other hand, bit-sets are also shown using square brackets
233 but set elements are separated only by a space. Here is the shell
234 preparing to execute an external command:
236 sigprocmask(SIG_BLOCK, [CHLD TTOU], []) = 0
238 Here the second argument is a bit-set of two signals, SIGCHLD and SIGTTOU.
239 In some cases the bit-set is so full that printing out the unset
240 elements is more valuable. In that case, the bit-set is prefixed by
243 sigprocmask(SIG_UNBLOCK, ~[], NULL) = 0
245 Here the second argument represents the full set of all signals.
250 Count time, calls, and errors for each system call and report a summary on
251 program exit. On Linux, this attempts to show system time (CPU time spent
252 running in the kernel) independent of wall clock time. If
258 (below), only aggregate totals for all traced processes are kept.
263 but also print regular output while processes are running.
266 Run tracer process as a detached grandchild, not as parent of the
267 tracee. This reduces the visible effect of
269 by keeping the tracee a direct child of the calling process.
272 Show some debugging output of
274 itself on the standard error.
277 Trace child processes as they are created by currently traced
278 processes as a result of the
282 On non-Linux platforms the new process is
283 attached to as soon as its pid is known (through the return value of
285 in the parent process). This means that such children may run
286 uncontrolled for a while (especially in the case of a
288 until the parent is scheduled again to complete its
290 call. On Linux the child is traced from its first instruction with no delay.
291 If the parent process decides to
293 for a child that is currently
294 being traced, it is suspended until an appropriate child process either
295 terminates or incurs a signal that would cause it to terminate (as
296 determined from the child's current signal disposition).
302 option is in effect, each processes trace is written to
304 where pid is the numeric process id of each process.
305 This is incompatible with
307 since no per-process counts are kept.
310 This option is now obsolete and it has the same functionality as
314 Print the help summary.
317 Print the instruction pointer at the time of the system call.
320 Suppress messages about attaching, detaching etc. This happens
321 automatically when output is redirected to a file and the command
322 is run directly instead of attaching.
325 Print a relative timestamp upon entry to each system call. This
326 records the time difference between the beginning of successive
330 Prefix each line of the trace with the time of day.
333 If given twice, the time printed will include the microseconds.
336 If given thrice, the time printed will include the microseconds
337 and the leading portion will be printed as the number
338 of seconds since the epoch.
341 Show the time spent in system calls. This records the time
342 difference between the beginning and the end of each system call.
345 Print unabbreviated versions of environment, stat, termios, etc.
346 calls. These structures are very common in calls and so the default
347 behavior displays a reasonable subset of structure members. Use
348 this option to get all of the gory details.
351 Print the version number of
355 Print all non-ASCII strings in hexadecimal string format.
358 Print all strings in hexadecimal string format.
361 Print paths associated with file descriptor arguments.
364 Align return values in a specific column (default column 40).
367 A qualifying expression which modifies which events to trace
368 or how to trace them. The format of the expression is:
371 [\fIqualifier\fB=\fR][\fB!\fR]\fIvalue1\fR[\fB,\fIvalue2\fR]...
387 is a qualifier-dependent symbol or number. The default
390 Using an exclamation mark negates the set of values. For example,
393 .BR \-e "\ " trace = open
394 which in turn means trace only the
396 system call. By contrast,
397 .BR \-e "\ " trace "=!" open
398 means to trace every system call except
400 In addition, the special values
404 have the obvious meanings.
406 Note that some shells use the exclamation point for history
407 expansion even inside quoted arguments. If so, you must escape
408 the exclamation point with a backslash.
410 \fB\-e\ trace\fR=\fIset\fR
411 Trace only the specified set of system calls. The
413 option is useful for determining which system calls might be useful
414 to trace. For example,
415 .BR trace = open,close,read,write
417 trace those four system calls. Be careful when making inferences
418 about the user/kernel boundary if only a subset of system calls
419 are being monitored. The default is
422 .BR "\-e\ trace" = file
423 Trace all system calls which take a file name as an argument. You
424 can think of this as an abbreviation for
425 .BR "\-e\ trace" = open , stat , chmod , unlink ,...
426 which is useful to seeing what files the process is referencing.
427 Furthermore, using the abbreviation will ensure that you don't
428 accidentally forget to include a call like
430 in the list. Betchya woulda forgot that one.
432 .BR "\-e\ trace" = process
433 Trace all system calls which involve process management. This
434 is useful for watching the fork, wait, and exec steps of a process.
436 .BR "\-e\ trace" = network
437 Trace all the network related system calls.
439 .BR "\-e\ trace" = signal
440 Trace all signal related system calls.
442 .BR "\-e\ trace" = ipc
443 Trace all IPC related system calls.
445 .BR "\-e\ trace" = desc
446 Trace all file descriptor related system calls.
448 \fB\-e\ abbrev\fR=\fIset\fR
449 Abbreviate the output from printing each member of large structures.
454 option has the effect of
457 \fB\-e\ verbose\fR=\fIset\fR
458 Dereference structures for the specified set of system calls. The
462 \fB\-e\ raw\fR=\fIset\fR
463 Print raw, undecoded arguments for the specified set of system calls.
464 This option has the effect of causing all arguments to be printed
465 in hexadecimal. This is mostly useful if you don't trust the
466 decoding or you need to know the actual numeric value of an
469 \fB\-e\ signal\fR=\fIset\fR
470 Trace only the specified subset of signals. The default is
476 causes SIGIO signals not to be traced.
478 \fB\-e\ read\fR=\fIset\fR
479 Perform a full hexadecimal and ASCII dump of all the data read from
480 file descriptors listed in the specified set. For example, to see
481 all input activity on file descriptors
486 \fB\-e\ read\fR=\fI3\fR,\fI5\fR.
487 Note that this is independent from the normal tracing of the
489 system call which is controlled by the option
490 .BR -e "\ " trace = read .
492 \fB\-e\ write\fR=\fIset\fR
493 Perform a full hexadecimal and ASCII dump of all the data written to
494 file descriptors listed in the specified set. For example, to see
495 all output activity on file descriptors
500 \fB\-e\ write\fR=\fI3\fR,\fI5\fR.
501 Note that this is independent from the normal tracing of the
503 system call which is controlled by the option
504 .BR -e "\ " trace = write .
507 Write the trace output to the file
509 rather than to stderr.
515 If the argument begins with `|' or with `!' then the rest of the
516 argument is treated as a command and all output is piped to it.
517 This is convenient for piping the debugging output to a program
518 without affecting the redirections of executed programs.
521 Set the overhead for tracing system calls to
524 This is useful for overriding the default heuristic for guessing
525 how much time is spent in mere measuring when timing system calls using
528 option. The accuracy of the heuristic can be gauged by timing a given
529 program run without tracing (using
531 and comparing the accumulated
532 system call time to the total produced using
536 Attach to the process with the process
540 The trace may be terminated
541 at any time by a keyboard interrupt signal (\c
544 will respond by detaching itself from the traced process(es)
545 leaving it (them) to continue running.
548 options can be used to attach to up to 32 processes in addition to
550 (which is optional if at least one
555 Trace only system calls accessing
559 options can be used to specify up to 256 paths.
562 Specify the maximum string size to print (the default is 32). Note
563 that filenames are not considered strings and are always printed in
567 Sort the output of the histogram printed by the
569 option by the specified criterion. Legal values are
579 Run command with the user \s-1ID\s0, group \s-2ID\s0, and
580 supplementary groups of
582 This option is only useful when running as root and enables the
583 correct execution of setuid and/or setgid binaries.
584 Unless this option is used setuid and setgid programs are executed
585 without effective privileges.
587 \fB\-E\ \fIvar\fR=\fIval\fR
590 in its list of environment variables.
595 from the inherited list of environment variables before passing it on to
602 exits with the same exit status.
605 is terminated by a signal,
607 terminates itself with the same signal, so that
609 can be used as a wrapper process transparent to the invoking parent process.
615 is zero unless there was an unexpected error in doing the tracing.
616 .SH "SETUID INSTALLATION"
619 is installed setuid to root then the invoking user will be able to
620 attach to and trace processes owned by any user.
621 In addition setuid and setgid programs will be executed and traced
622 with the correct effective privileges.
623 Since only users trusted with full root privileges should be allowed
625 it only makes sense to install
627 as setuid to root when the users who can execute it are restricted
628 to those users who have this trust.
629 For example, it makes sense to install a special version of
631 with mode `rwsr-xr--', user
637 group are trusted users.
638 If you do use this feature, please remember to install
639 a non-setuid version of
641 for ordinary lusers to use.
648 It is a pity that so much tracing clutter is produced by systems
649 employing shared libraries.
651 It is instructive to think about system call inputs and outputs
652 as data-flow across the user/kernel boundary. Because user-space
653 and kernel-space are separate and address-protected, it is
654 sometimes possible to make deductive inferences about process
655 behavior using inputs and outputs as propositions.
657 In some cases, a system call will differ from the documented behavior
658 or have a different name. For example, on System V-derived systems
661 system call does not take an argument and the
665 and takes an extra leading argument. These
666 discrepancies are normal but idiosyncratic characteristics of the
667 system call interface and are accounted for by C library wrapper
670 On some platforms a process that has a system call trace applied
673 option will receive a
675 This signal may interrupt a system call that is not restartable.
676 This may have an unpredictable effect on the process
677 if the process takes no action to restart the system call.
679 Programs that use the
684 privileges while being traced.
686 A traced process ignores
688 except on SVR4 platforms.
690 A traced process which tries to block SIGTRAP will be sent a SIGSTOP
691 in an attempt to force continuation of tracing.
693 A traced process runs slowly.
695 Traced processes which are descended from
697 may be left running after an interrupt signal (\c
700 On Linux, exciting as it would be, tracing the init process is forbidden.
704 option is weakly supported.
709 was written by Paul Kranenburg
710 for SunOS and was inspired by its trace utility.
713 was ported to Linux and enhanced
714 by Branko Lankester, who also wrote the Linux kernel support.
715 Even though Paul released
718 Branko's work was based on Paul's
720 1.5 release from 1991.
721 In 1993, Rick Sladkey merged
723 2.5 for SunOS and the second release of
725 for Linux, added many of the features of
727 from SVR4, and produced an
729 that worked on both platforms. In 1994 Rick ported
731 to SVR4 and Solaris and wrote the
732 automatic configuration support. In 1995 he ported
735 and tired of writing about himself in the third person.
737 The SIGTRAP signal is used internally by the kernel implementation of
738 system call tracing. When a traced process receives a SIGTRAP signal not
739 associated with tracing, strace will not report that signal correctly.
740 This signal is not normally used by programs, but could be via a hard-coded
741 break instruction or via
746 should be reported via the Debian Bug Tracking System,
749 mailing list at <strace\-devel@lists.sourceforge.net>.