Re: [SQU] Getting Squid to core-dump

Solaris by dafault refuses to dump core for processes using setuid() to
drop priviledges. This started with some solaris 7 patch, ( 106541-06
IIRC ) and is present (with man pages etc) in Solaris 8 too.

Quoting from the coreadm(1m) man page:

     A process that is or ever has been setuid or setgid since
     its last exec(2), including a process that began life with
     super-user privileges and gave up that privilege by way of
     setuid(2), presents security issues with respect to dumping
     core, as it may contain sensitive information in its address
     space to which the current non-privileged owner of the pro-
     cess should not have access.

If you're running a Solaris 7 kernel with patch >= 106541-06 ( use
'uname -v' to find out), or Solaris 8, you will have to execute
'coreadm -e proc-setid' before starting squid, for solaris to dump
a core file at the crash. Otherwise, solaris will not drop a squid
core file.

As Henrik pointed out though, you're better off starting squid from the
debugger.

And NO, I don't work for Sun, i found that out when trying to figure out
a (similar?) squid trouble... :)

On Thu, Feb 01, 2001 at 04:25:07PM +1030, David Gameau wrote:
> Hiya,
>
> When we run a 'squid -k shutdown', near the end of process, it catches a
> Segmentation Violation.
> The nature of Squid is to catch these signals and restart itself.
> We'd like to get a core file, to try and work out why it is breaking.
>
> Any suggestions about how to go about this?
>
> System:
> Squid-2.2STABLE5-hno
> Solaris 7 (SPARC)
>
> Thanks,
> David.
> __
>
> David Gameau
> I.T.S. - Unix Team
> University of South Australia
>
> email: David.Gameau@UniSA.edu.au
> phone: +61 8 302 3533
> fax: +61 8 302 5800
>
> Disclaimer: I didn't do it. Nobody saw me do it. You can't prove anything.
>

--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html