Subject: Re: kern/4099: Gather coredumps in one place
To: None <current-users@NetBSD.ORG>
From: Ty Sarna <firstname.lastname@example.org>
Date: 09/08/1997 20:46:57
In article <Pine.GSO.3.96.970908171236.19276Hemail@example.com> you write:
> QUick -- what's the PID of the mail client you're running right now?
> And if it crashes while you're reading this, how do you intend to get
> that information?
A feature that I think would be useful, with or without the other
changes being discussed, is logging of coredumps. When a core dump is
generated, the kernel should log at kern.notice (by default, with a
sysctl-changeable severity level) the pid, complete path of the dump,
and perhaps real/effective uid/gid info:
pid 42 dumped core in /some/dir/prog.core, uid=12 euid=13 gid=14 egid=15
One could then grep the log (for "dumped core") to find core files
instead of searching the filesystem, have a permanent log of what
crashed (might be useful for security purposes), and find out more
immediately when daemons die. It might also help in the case above
(just look in the logs for a dump matching the program name and your uid
in the appropriate time frame).
>From a quick glance at kern_sig.c, it seems like this should be really
easy to implement (with the possible exception of getting the full path
of the dump... can I get that from the nameidata?).