Copyright (c) 2004, Sun Microsystems, Inc. All Rights Reserved.
The contents of this file are subject to the terms of the Common Development and Distribution License (the "License"). You may not use this file except in compliance with the License.
You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE or http://www.opensolaris.org/os/licensing. See the License for the specific language governing permissions and limitations under the License.
When distributing Covered Code, include this CDDL HEADER in each file and include the License file at usr/src/OPENSOLARIS.LICENSE. If applicable, add the following below this CDDL HEADER, with the fields enclosed by brackets "[]" replaced with your own identifying information: Portions Copyright [yyyy] [name of copyright owner]
apptrace [-f] [-F [!] tracefromlist] [-T [!] tracetolist] [-o outputfile] [ [-tv] [!] call ,...] command [command arguments]
The apptrace utility runs the executable program specified by command and traces all function calls that the programcommand makes to the Solaris shared libraries. For each function call that is traceable, apptrace reports the name of the library interface called, the values of the arguments passed, and the return value.
By default, apptrace traces calls directly from the executable object to any of the shared objects it depends on. Indirect calls (that is, calls made between shared objects that the executable depends upon) are not reported by default.
Calls from or to additional shared objects may be traced using the -F or -T options (see below).
The default reporting format is a single line per call, with no formatted printing of arguments passed by reference or of data structures.
Formatted printing providing additional argument details is obtained using the -v option (see below).
By default, every interface provided by a shared object is traced if called. However, the set of interfaces to be traced can be restricted, using the -t and/or -v options.
Since it is generally possible to trace calls between any of the dynamic objects linked at runtime (the executable object and any of the shared objects depended upon), the report of each traced call gives the name of the object from which the call was made.
apptrace traces all of the procedure calls that occur between dynamic objects via the procedure linkage table, so only those procedure calls which are bound via the table will be traced. See the Linker and Libraries Guide.
The following options are supported:
-f
Follows all children created by fork(2). This option will also cause the process id to be printed at the beginning of each line.
-F [!]tracefromlist
Traces calls from a comma-separated list of shared objects. Only calls from these shared objects will be traced. The default is to trace calls from the main executable only. Only the basename of the shared object is required. For example, libc will match /usr/lib/libc.so.1. Additionally, shell style wildcard characters are supported as described in fnmatch(5). A list preceded by a ``!'' defines a list of objects from which calls should not be traced. If the tracing of calls from command is required, then command must be a member of tracefromlist.
-o outputfile
apptrace output will be directed to the outputfile. By default, apptrace output is placed on the stderr stream of the process being traced.
-t [!]call,\|.\|.\|.
Traces or excludes function calls. Those calls specified in the comma-separated list call are traced. If the list begins with a !, the specified function calls are excluded from the trace output. The default is -t *. The use of shell style wildcards is allowed.
-T [!]tracetolist
Traces calls to a comma-separated list of shared objects. The default is to trace calls to all shared objects. As above, the basename is all that is required and wildcarding is allowed. A list preceded by a ``!'' denotes a list of objects to which calls should not be traced.
-v [!]call,\|.\|.\|.
Provides verbose, formatted output of the arguments and return values of the function calls specified (as above in the -t option). Unlike truss(1), calls named by the -v option do not have to be named by the -t option. For example, apptrace -v open is equivalent to truss -t open -v open.
Example 1 Tracing the date command
% apptrace date -> date -> libc.so.1:atexit(0xff3bf9ac, 0x22000, 0x0) ** NR -> date -> libc.so.1:atexit(0x11550, 0xfefeef80, 0xab268) ** NR -> date -> libc.so.1:setlocale(0x6, 0x11560, 0x0) ** NR -> date -> libc.so.1:textdomain(0x11564, 0xfefce156, 0xff160200) ** NR -> date -> libc.so.1:int getopt(int = 0x1, const char * * = 0xffbffa5c, const char * = 0x11574 "a:u") <- date -> libc.so.1:getopt() = 0xffffffff -> date -> libc.so.1:time_t time(time_t * = 0x225c0) <- date -> libc.so.1:time() = 0x41ab6e82 -> date -> libc.so.1:char * nl_langinfo(nl_item = 0x3a) <- date -> libc.so.1:nl_langinfo() = 0xfefd3e10 -> date -> libc.so.1:struct tm * localtime(const time_t * = 0x225c0) <- date -> libc.so.1:localtime() = 0xff160240 -> date -> libc_psr.so.1:memcpy(0xffbff9cc, 0xff160240, 0x24) ** NR -> date -> libc.so.1:size_t strftime(char * = 0x225c4 "", size_t = 0x400, const char * = 0xfefd3e10 "%a %b %e %T %Z %Y", const struct tm * = 0xffbff9cc) <- date -> libc.so.1:strftime() = 0x1c -> date -> libc.so.1:int puts(const char * = 0x225c4 "Mon Nov 29 10:46:26 PST 2004") Mon Nov 29 10:46:26 PST 2004 <- date -> libc.so.1:puts() = 0x1d -> date -> libc.so.1:exit(0x0, 0x22400, 0x0) ** NR
Example 2 Tracing a specific set of interfaces with verbosity set
% apptrace -v localtime,strftime,puts date -> date -> libc.so.1:struct tm * localtime(const time_t * = 0x225c0) arg0 = (const time_t *) 0x225c0 return = (struct tm *) 0xff160280 (struct tm) { tm_sec: (int) 0x4 tm_min: (int) 0x34 tm_hour: (int) 0xa tm_mday: (int) 0x1d tm_mon: (int) 0xa tm_year: (int) 0x68 tm_wday: (int) 0x1 tm_yday: (int) 0x14d tm_isdst: (int) 0 } <- date -> libc.so.1:localtime() = 0xff160280 -> date -> libc.so.1:size_t strftime(char * = 0x225c4 "", size_t = 0x400, const char * = 0xfefd3e10 "%a %b %e %T %Z %Y", const struct tm * = 0xffbff99c) arg0 = (char *) 0x225c4 "" arg1 = (size_t) 0x400 arg2 = (const char *) 0xfefd3e10 "%a %b %e %T %Z %Y" arg3 = (const struct tm *) 0xffbff99c (struct tm) { tm_sec: (int) 0x4 tm_min: (int) 0x34 tm_hour: (int) 0xa tm_mday: (int) 0x1d tm_mon: (int) 0xa tm_year: (int) 0x68 tm_wday: (int) 0x1 tm_yday: (int) 0x14d tm_isdst: (int) 0 } return = (size_t) 0x1c <- date -> libc.so.1:strftime() = 0x1c -> date -> libc.so.1:int puts(const char * = 0x225c4 "Mon Nov 29 10:52:04 PST 2004") arg0 = (const char *) 0x225c4 "Mon Nov 29 10:52:04 PST 2004" Mon Nov 29 10:52:04 PST 2004 return = (int) 0x1d <- date -> libc.so.1:puts() = 0x1d
** NR - The return value of a function call will not be traced.
Basic runtime support for apptrace is provided by the link auditing feature of the Solaris runtime linker (ld.so.1(1)) and the apptrace command's use of this facility relies on an auditing object (apptrace.so.1) kept in /usr/lib/abi.
In general, apptrace cannot trace calls to functions accepting variable argument lists. There has been some clever coding in several specific cases to work around this limitation, most notably in the printf and scanf families.
The apptrace utility can not trace the return value of a function call whose return type is a struct or union.
Functions that attempt to probe the stack or otherwise extract information about the caller cannot be traced. Some examples are [gs]etcontext(), [sig]longjmp(), [sig]setjmp(), and vfork().
Functions such as exit(2) that do not return will not be traced for their return values.
For security reasons, only those processes with appropriate privileges can use apptrace to trace setuid/setgid programs.
Tracing functions whose usage requires the inclusion of <varargs.h>, such as vwprintw(3XCURSES) and vwscanw(3XCURSES), will not provide formatted printing of arguments.
See attributes(5) for descriptions of the following attributes:
| ATTRIBUTE TYPE | ATTRIBUTE VALUE |
| Interface Stability | Unstable |
ld.so.1(1), truss(1), vwprintw(3XCURSES), vwscanw(3XCURSES), attributes(5), fnmatch(5)
Linker and Libraries Guide