Date: Wed, 24 Dec 1997 03:18:06 -0600 (CST) From: Joel Ray Holveck <joelh@gnu.org> To: freebsd-current@freebsd.org Subject: Random crash update Message-ID: <199712240918.DAA17870@detlev.UUCP>
next in thread | raw e-mail | index | archive | help
Well, as I said earlier, I was having random core dumps from Emacs,
bash, sendmail, and others when I rebuilt the OS. I rebuilt it again
(from the same sources, ctm 3176, w/o secure) and thought my woes were
over. It turns out that they were apparently just less frequent.
Instead of seeing one program crash every five minutes, it's more like
every 20 or so.
During the build process, everything worked fine. I had shut down all
non-essential processes (X, inetd, sendmail, etc) for the duration.
No core dumps. I rebuilt the new sources one at a time, by first
rebuilding include, then lib, then executing in csh:
foreach foo in ( bin gnu include lib libexec lkm sbin usr.bin usr.sbin )
cd $foo
make all install clean || break
cd ..
end
(Yes, I realize this makes include and lib build again. No harm, eh?)
A couple of times sources wouldn't complete (slattach and startslip
had problems with not including certain #includes, but I had noticed a
patch in the pipe so just didn't rebuild them... I'll submit a pr if
that's not fixed by the time this fiasco ends; also, something else
(I've got it written down somewhere around here) wanted to install a
file to the previously nonexistant /usr/share/locale/sl_SI.ISO_8859-2
directory), at which point I used the same csh command with the SUBDIR
variable, and omitted the sections I had problems with. (This is what
happens when you have a low-paying job and can't afford enough drive
space to make world.)
Rebooted. No worries, it appeared. Ran Emacs's M-1 2 M-x hanoi
command (towers of hanoi, n=12). It naturally gc'd several times, but
never gave me a problem. (Before the second rebuild, it would crash
before completing a n=6 run.) Shut down the system completely.
Next boot, start to get my mail. In about two hours, I've had the
following errors:
-bash in free(): warning: pointer to wrong page.
-bash in free(): warning: modified (chunk-) pointer.
[bash exits.]
-bash in free(): warning: junk pointer, too high to make sense.
-bash in free(): warning: modified (chunk-) pointer.
[bash hangs at very high CPU usage.]
Dec 24 02:30:37 detlev /kernel: pid 1275 (sendmail), uid 0: exited on
signal 11
[sendmail exits.]
Dec 24 03:00:38 detlev /kernel: pid 2211 (sendmail), uid 0: exited on
signal 11
[sendmail exits.]
(Does anybody know how to convince sendmail to leave core dumps?)
Recall that bash and sendmail each were giving me problems after my
first rebuild, among others. Additionally, bash is only dumping when
a child process dies. The first sendmail shouldn't have been forking
(IIRC, the mail queue was empty), but I could be wrong. The second
sendmail had an empty queue and no network connection, so it shouldn't
have forked at all. Nevertheless, I'll mention that
kern.fast_vfork=1; I haven't tried changing it yet. Note also that
although sendmail was recompiled, its sources have not changed since I
first installed FreeBSD on this box (28 Nov) save hub.mc, which of
course wasn't used. Note yet also that bash has not been recompiled.
The following timeline be of interest:
- The first time I connected, ijjppp dropped my connection during the
middle of the session. I had timeout set to 300 seconds. The
fetchmail session was very much active.
- Upon reconnecting and setting timeout to 0, I ran fetchmail and
ssh. Both uneventful.
- After retrieving some 200+ messages, and running them all through
sendmail, I verified that mailq was empty, set the timeout back to
300, and started a ping to test ppp's timeout. (ppp stayed up this
time.)
- During this ping, I noticed inetd consuming some 93% of the system
resources according to top. (Am486DX/4-100, 16MB RAM.) I killed
inetd; I didn't need it at the moment. (Restarted inetd later and
ping'd 127.0.0.1; no recongizable effect on inetd; besides, inetd
shouldn't even see ICMPs if I understand right.)
- Sometime around here (after checking mailq, before sending message)
was when sendmail died the first time.
- Killed ping, dropped ppp connection.
- Started writing this message. During the composition, sendmail died
again.
If anybody has any suggestions, I'd love to hear them.
Thanks,
joelh
--
Joel Ray Holveck - joelh@gnu.org - http://www.wp.com/piquan
Fourth law of programming:
Anything that can go wrong wi
sendmail: segmentation violation - core dumped
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199712240918.DAA17870>