Help with sigaction() and sigcontext

[Miro Jurisic]

In article <tobiaspatton-55FDA8.10451930082003@shawnews.vc.shawcable.net>,
Tobias Patton <tobiaspatton@NOSPAM.shaw.ca> wrote:

> I'm trying to install a signal handler that will dump a stack crawl.

Why?

meeroh

--
If this message helped you, consider buying an item
from my wish list: <http://web.meeroh.org/wishlist>

[Eric Albert]

In article
<tobiaspatton-55FDA8.10451930082003@shawnews.vc.shawcable.net>,
Tobias Patton <tobiaspatton@NOSPAM.shaw.ca> wrote:

> From reading the manpage for sigaction() and the comments in signal.h, I
> thought I could use the sigcontext pointer that's passed to the signal
> handler to get a pointer to my process's stack.

Yep.

> One of the fields in
> this struct is `sc_sp' which I thought was the stack pointer for the
> user-space process that caused the signal. So I wrote some test code to
> examine this value. A typical value for `sc_sp' is 0xfb4b7f0, which
> doesn't look like its in my process's memory space. Sure enough,
> attempting to dererefence this pointer to walk up the stack results in a
> segmentation fault.

Yeah, that doesn't look like a valid stack address.

> I've included the test code I wrote, and some sample output. The code
> was compiled using ProjectBuilder 2.1. Any help is appreciated.

Which version of the OS is this? SA_SIGINFO signal handlers didn't work
correctly until 10.2.

> #include <iomanip>
> #include <iostream>
> #include <signal.h>
>
> void SigActionHandler( int sigNum, siginfo_t* sigInfo, void* sigCtx )
> {
> ( void ) sigInfo;
>
> struct sigcontext* ctx = ( sigcontext* ) sigCtx;
>
> std::cout << "Signum : " << sigNum << std::endl

It's not safe to use STL from a signal handler. See
<http://www.opengroup.org/onlinepubs/007908799/xsh/sigaction.html> for a
list of functions that are async-signal safe.

-Eric

--
Eric Albert [email]ejalbert@stanford.edu[/email]
[url]http://rescomp.stanford.edu/~ejalbert/[/url]