On Fri, Jun 24, 2005 at 04:54:48PM +0200, David Groep wrote:
> Hi all,
>
> Jeff Templon wrote:
> >A couple of mysterious things, see included message.
> >===========
> >Davide Salomoni wrote:
> >
> >And while I was looking on bosheks, I found these in /var/log/messages:
> >
> >Jun 24 10:07:37 bosheks kernel: application bug: edg-wl-in.ftpd(8787)
> >has SIGCHLD set to SIG_IGN but calls wait().
> >Jun 24 10:07:37 bosheks kernel: (see the NOTES section of 'man 2 wait').
> >Workaround activated.
> >
> >is this something you've seen before?
>
> This is a false-positive error message generated by a bug in the kernel
> 2.4.21-27.0.2 shipped with RHEL U4. See also
> https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=140552
>
> It has been resolved in U5 (according to RH bugzilla). Nothing to worry
> about and no reason to "fix" edg-wl-in.ftpd.
The kernel bug when your application has SIGCHLD set to ignore and causes
a module to be loaded is fixed in U5 but it's not the same bug....
Here is a comment from the bug:
Assuming that the application is not performing wait()/wait4()/waitpid()
system calls while SIGCHLD signals are being ignored, yes.
For more details have a look at:
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=98740
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=127849
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=128640
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=117333
edg-wl-in.ftpd is using wait*() while SIGCHLD is ignored so it *does* need
a fix. For an example look at how rsync was fixed to correct the bug
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=98740
Kostas
|