Date: Tue, 5 Jan 1999 03:59:48 +0000 (GMT) From: Terry Lambert <tlambert@primenet.com> To: lem@cantv.net (Luis =?iso-8859-1?Q?Mu=F1oz?=) Cc: spork@super-g.com, kaleb@ics.com, hackers@FreeBSD.ORG Subject: Re: inetd in realloc(): warning: junk pointer, too low to make Message-ID: <199901050359.UAA04503@usr05.primenet.com> In-Reply-To: <3.0.6.32.19981221171019.008451f0@pop.cantv.net> from "Luis =?iso-8859-1?Q?Mu=F1oz?=" at Dec 21, 98 05:10:19 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> >Well, here's what sendmail does here, it's pretty scary, with those > >truncated messages about spwd.db: > > > >Dec 17 15:42:27 super-g Dec 17 15:42:27sendmail[: NOQUEUE > >Dec 17 15:51:11 super-g Dec 17 15:51:11sendmail[: /etc/spwd.db > >Dec 17 15:51:11 super-g sendmail[16594]: NOQUEUE: SYSERR(root): Out of > >memory!!: Cannot allocate memory > >Dec 17 16:30:01 super-g Dec 17 16:30:00sendmail[: /etc/spwd.db > >Dec 17 16:30:16 super-g /kernel: pid 20618 (sendmail), uid 0: exited on > >signal 11 > >Dec 17 16:30:57 super-g Dec 17 16:30:57sendmail[: /etc/spwd.db > >Dec 17 16:30:57 super-g Dec 17 16:30:57sendmail[: NOQUEUE > >Dec 17 16:30:57 super-g Dec 17 16:30:57sendmail[: /etc/spwd.db > > > >Charles > > Agreed. Are your files consistent/sane? I've never seen this > before! This is related to running out of memory. The problem occurs because clean mmap'ed pages are reaped for reuse in a swap low condition, but the "page present" bit is not correctly unset in the places where it is referenced. Effectively, this corrupts one page of the libc (which is mmap'ed) that the sendmail is linked against, and basically damages its ability to fork children. You can work around the problem by statically link sendmail and/or not running out of memory (e.g. run a newer version of Netscape, use a DISPLAY variable that disables Netscape using the MIT SHM extension, run Netscape from a remote machine using the local machine as a display, which will also prevent it from using the MIT SHM extension, or just disable the MIT SHM extension entirely). You can also workaround the problem by not allowing the reaping of pages in a mapping that references a vnode with a greater than one reference count (not an ideal workaround, admittedly, but it works). Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199901050359.UAA04503>