1
0
Fork 0
mirror of https://github.com/tobast/libunwind-eh_elf.git synced 2024-11-30 10:37:37 +01:00
libunwind-eh_elf/doc/libunwind-ptrace.man
2003-01-28 07:32:15 +00:00

193 lines
4.3 KiB
Groff

'\" t
.\" Manual page created with latex2man on Mon Jan 27 23:19:42 PST 2003
.\" NOTE: This file is generated, DO NOT EDIT.
.de Vb
.ft CW
.nf
..
.de Ve
.ft R
.fi
..
.TH "LIBUNWIND\-PTRACE" "3" "27 January 2003" "Programming Library " "Programming Library "
.SH NAME
libunwind\-ptrace \-\- ptrace() support in libunwind
.PP
.SH SYNOPSIS
.PP
#include <libunwind.h>
.br
.PP
unw_accessors_t
_UPT_accessors;
.br
.PP
void *_UPT_create(pid_t);
.br
void
_UPT_destroy(void *);
.br
.PP
int
_UPT_find_proc_info(unw_addr_space_t,
unw_word_t,
unw_proc_info_t *,
int,
void *);
.br
void
_UPT_put_unwind_info(unw_addr_space_t,
unw_proc_info_t *,
void *);
.br
int
_UPT_get_dyn_info_list_addr(unw_addr_space_t,
unw_word_t *,
void *);
.br
int
_UPT_access_mem(unw_addr_space_t,
unw_word_t,
unw_word_t *,
int,
void *);
.br
int
_UPT_access_reg(unw_addr_space_t,
unw_regnum_t,
unw_word_t *,
int,
void *);
.br
int
_UPT_access_fpreg(unw_addr_space_t,
unw_regnum_t,
unw_fpreg_t *,
int,
void *);
.br
.PP
.SH DESCRIPTION
.PP
The ptrace(2)
system\-call makes it possible for a process to
gain access to the machine\-state and virtual memory of \fIanother\fP
process. With the right set of call\-back routines, it is therefore
possible to hook up libunwind
to another process via
ptrace(2).
While it\&'s not very difficult to do so directly,
libunwind
further facilitates this task by providing
ready\-to\-use callbacks for this purpose. The routines and variables
implementing this facility use a name\-prefix of _UPT,
which is
stands for ``unwind\-via\-ptrace\&''\&.
.PP
An application that wants to use the _UPT\-facility
first needs
to create a new libunwind
address\-space that represents the
target process. This is done by calling
unw_create_addr_space().
In many cases, the application
will simply want to pass the address of _UPT_accessors
as the
first argument to this routine. Doing so will ensure that
libunwind
will be able to properly unwind the target process.
However, in special circumstances, an application may prefer to use
only portions of the _UPT\-facility.
For this reason, the
individual callback routines (_UPT_find_proc_info(),
_UPT_put_unwind_info(),
etc.) are also available for direct
use. Of course, the addresses of these routines could also be picked
up from _UPT_accessors,
but doing so would prevent static
initialization. Also, when using _UPT_accessors,
\fIall\fP
the callback routines will be linked into the application, even if
they are never actually called.
.PP
Next, the application can turn on ptrace\-mode on the target process,
either by forking a new process, invoking PTRACE_TRACEME,
and
then starting the target program (via execve(2)),
or by
directly attaching to an already running process (via
PTRACE_ATTACH).
Either way, once the process\-ID (pid) of the
target process is known, a _UPT\-info\-structure
can be created
by calling _UPT_create(),
passing the pid of the target process
as the only argument. The returned void\-pointer then needs to be
passed as the ``argument\&'' pointer (third argument) to
unw_init_remote().
.PP
When the application is done using libunwind
on the target
process, _UPT_destroy()
needs to be called, passing it the
void\-pointer that was returned by the corresponding call to
_UPT_create().
This ensures that all memory and other
resources are freed up.
.PP
.SH AVAILABILITY
.PP
Since ptrace(2)
works within a single machine only, the
_UPT\-facility
by definition is not available in
libunwind\-versions
configured for cross\-unwinding.
.PP
.SH THREAD SAFETY
.PP
The _UPT\-facility
assumes that a single _UPT\-info
structure is never shared between threads. Because of this, no
explicit locking is used. As long as only one thread uses
a _UPT\-info
structure at any given time, this facility
is thread\-safe.
.PP
.SH RETURN VALUE
.PP
_UPT_create()
may return a NULL
pointer if it fails
to create the _UPT\-info\-structure
for any reason. For the
current implementation, the only reason this call may fail is when the
system is out of memory.
.PP
.SH SEE ALSO
.PP
execve(2),
libunwind(3),
ptrace(2)
.PP
.SH AUTHOR
.PP
David Mosberger\-Tang
.br
Hewlett\-Packard Labs
.br
Palo\-Alto, CA 94304
.br
Email: \fBdavidm@hpl.hp.com\fP
.br
WWW: \fBhttp://www.hpl.hp.com/research/linux/libunwind/\fP\&.
.\" NOTE: This file is generated, DO NOT EDIT.