Organization: Netcom - Online Communication Services (408 241-9760 guest)
References: <1992Nov9.193850.14331@lgc.com>
Date: Mon, 9 Nov 1992 23:36:04 GMT
Lines: 77
In article <1992Nov9.193850.14331@lgc.com> danson@lgc.com (Doug Anson) writes:
>Hi:
>
>I have a small two node network where one machine is running 386bsd and the
>other is running SVR4.0.3.6. I have been NFS mounting a partition from the
>386bsd machine to the SVR4.0.3.6 machine. With this mount, I have tried compiling a large application (i.e. gcc.2.3.1) for use in the SVR4 environment.
>However, I have run into what looks like a memory leak in 386bsd within the
>NFS (or wd8003) subsystem.
>
>Basically, what happens is that the NFS mount works fine when the NFS traffic
>is fairly light. However, during heavy NFS I/O, I'll sometimes panic the
>386bsd kernel. The panic complains:
> kmem_malloc: kmem_map too small
>
This may help... (another posting to 4bsd.bugs included a patch which
doesn't apply to 386BSD).
-AM
From: julian@tfs.com (Julian Elischer)
Subject: IMPORTANT PATCH!
Message-ID: <1992Nov3.071934.25466@tfs.com>
Organization: Trw Financial Systems
Date: Tue, 3 Nov 1992 07:19:34 GMT
Lines: 49
The following patch fixes a randomly initialised variable in the icmp
and raw socket code that can cause 'ping' and 'traceroute'
to never fully close their socket's internal structures
leading to a geometrically increasing number of lost mbufs until
the system runs out of appropriate virtual space, at which
time the system panics with a 'kmem_alloc: map too small'
and halts.
EVERY time you run ping and traceroute you make the problem
worse
The 'rights' entry should not be there.
julian
+----------------------------------+ ______ _ __
| __--_|\ Julian Elischer | \ U \/ / On assignment
| / \ julian@tfs.com +------>x USA \ in a very strange
| ( OZ ) 2118 Milvia st. Berkeley CA. \___ ___ | country !