Date: Fri, 11 Jun 2004 15:22:06 +0000 From: Bosko Milekic <bmilekic@FreeBSD.org> To: othermark <atkin901@yahoo.com> Cc: freebsd-current@freebsd.org Subject: Re: Today's -current panics Message-ID: <20040611152206.GA11917@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
othermark wrote: >I get a very similar stack track traversing through sossend(), under heavy >NFS load on a 1GB machine. Note the panic message here, and the >peculiarity that previous incarnations of -current did not panic under >similar load. It is highly reproduceable via a 'make installworld' via >NFS with /usr/src and /usr/obj mounted. The NFS serving machine will >always panic using vanilla GENERIC: > >[root@pippin root]$ panic: kmem_malloc(4096): kmem_map too small: 40894464 >total > allocated Do you have the kern.ipc.nmbclusters boot-time tunable set to 0? I just noticed that if this is set to zero then kmem_map will not be scaled larger to accomodate clusters and mbufs. In this scenario, what I recommend that you do is increase the vm.kmem_size boot-time tunable to ~300,000,000 or ~400,000,000. Currently, your kmem_map is way too small (looks like only ~40M). Be careful not to overdo it, though, because you might also then have to increase the available KVA (KVA_PAGES). -Bosko
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040611152206.GA11917>