Subject: Re: Process hang on LFS
To: None <current-users@netbsd.org>
From: Christopher Tribo <t1345@hopi.dtcc.edu>
List: current-users
Date: 01/29/2003 17:17:56
During auto-mount, the rc.subr script does mounts piping the
output to /dev/null. I find that if I mount an LFS file system from a
shell, or unmount and remount an LFS file system that LFS complains with
lfs_mountfs: please consider increasing BUFPAGES to at least 5451.
If I do systat bufcache it tells me I have 6211 pages for cached
file data using 24844 kBytes of memory.
I'm kind of confused, after reading the options man page for
BUFPAGES, you can either dynamically control UBC related options with
sysctl and maxvnodes or compile a kernel with a BUFPAGES option, but that
locks things in and doesn't allow for any sort of dynamically adjustable
buffer cache. Am I misreading things? It seems as though I have more than
the requested number of bufpages. This is after bumping kern.maxvnodes up
to 32768.