Date: Sun, 5 Jul 1998 12:56:08 -0400 (EDT) From: Bill Paul <wpaul@skynet.ctr.columbia.edu> To: kkennawa@physics.adelaide.edu.au (Kris Kennaway) Cc: current@FreeBSD.ORG Subject: Re: inetd problems Message-ID: <199807051656.MAA14927@skynet.ctr.columbia.edu> In-Reply-To: <Pine.OSF.3.90.980706000351.17086A-100000@mercury> from "Kris Kennaway" at Jul 6, 98 00:09:27 am
next in thread | previous in thread | raw e-mail | index | archive | help
Of all the gin joints in all the towns in all the world, Kris Kennaway had to walk into mine and say: > I'm getting the following on my smtp port when I connect to it: > > [morden|kkenn] 0:05 ~ telnet localhost smtp > Trying 127.0.0.1... > Connected to localhost.rebel.net.au. > Escape character is '^]'. > inetd in realloc(): warning: junk pointer, too low to make sense. > 220 morden.net.au ESMTP > > I recall noticing this some time ago as well, but didnt look into it any > further - but I'm currently having intermittent problems delivering messages > via smtp, and this is possibly related. > > [morden|kkenn] 0:08 ~ ls -l /usr/sbin/inetd > -r-xr-xr-x 1 bin bin 24576 Jun 21 22:18 /usr/sbin/inetd* > > (A diff between this and a copy compiled from current sources shows no > differences) > > Any ideas? Hm. You know, I've been watching messages like this go by (people getting noise from inetd about malloc()/realloc()/free() problems) and I can't help but wonder why nobody has really sat down and tried to debug things thoroughly. I encounter messages like this occasionally with my own code (mostly NIS+ stuff) and without exception, the cause has always been some kind of bug on my part (double free(), free()ing static buffers by mistake, corrupting the heap by wandering past the end of a malloc()ed buffer, etc...). All the theories I've heard so far point to some kind of VM problem; nobody as yet has been willing to admit that the problem is a bug in inetd. I've found that the best thing for nailing these kinds of bugs is Electric Fence and gdb. Recompile inetd with -g, link it with Electric Fence, and run it, then see if you can get it to explode. Once it dumps core, gdb will tell you exactly where the real problem is (alternatively, you can attach to the running instance of inetd with gdb and catch the crash immediately). I should point out however that I'm uncertain that the particular problem you're seeing is actually being caused by inetd. It seems unusual that inetd would give you such a message for a simple thing like spawing an smtp process for you, so I'm wondering if maybe the error message about realloc() isn't coming from the smtp server instead, and it only _says_ inetd because inetd was the parent process that launched smtpd. (It shouldn't do that of course, but anything's possible.) More details would help here: under what conditions does this problem occur? Do you see the 'realloc' error all the time, or just when the system is heavily loaded? Exactly what smtp server is this? How many users does the server handle mail for? Can you debug this by recompiling inetd with the Electric Fence library and analyzing the crash dump? If not, can you at least provide a recipe for duplicating the bug so that someone else can try debugging it? -Bill -- ============================================================================= -Bill Paul (212) 854-6020 | System Manager, Master of Unix-Fu Work: wpaul@ctr.columbia.edu | Center for Telecommunications Research Home: wpaul@skynet.ctr.columbia.edu | Columbia University, New York City ============================================================================= "It is not I who am crazy; it is I who am mad!" - Ren Hoek, "Space Madness" ============================================================================= To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199807051656.MAA14927>