Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 23 May 2003 01:10:25 -0700
From:      Terry Lambert <tlambert2@mindspring.com>
To:        "Cagle, John (ISS-Houston)" <john.cagle@hp.com>
Cc:        current@freebsd.org
Subject:   Re: 5.1 PAE testing results: 2 panics
Message-ID:  <3ECDD771.A36B6832@mindspring.com>
References:  <C50AB9511EE59B49B2A503CB7AE1ABD10440E52F@cceexc19.americas.cpqcorp.net>

next in thread | previous in thread | raw e-mail | index | archive | help
"Cagle, John (ISS-Houston)" wrote:
> 
> As requested, here is the dmesg from this system with the PAE kernel.  I
> tried maxvnodes set to 100000, but had the same "kmem_map too small"
> panic.  I will try "clamping" the memory in machdep.c next...

Yes.  This is going to be necessary; we see here:

> real memory  = 9126805504 (8704 MB)
> avail memory = 8613085184 (8214 MB)

that it's seeing all the memory.  It has to base it's calculations
on the amount of real memory available for use in wired pages for
KVA.

Jake could better tell you how much KVA/UVA is lost to the bank
selection overhead; a good guesstimate (IMO) would be to use the
value of "3G" for the calculations for all the kernel allocations
that the "automatic" of "MAXUSERS=0" selects.

You may also want to just try (as a quick-and-dirty hack) setting
"MAXUSERS=256" or something similar, and see if it gets you to an
initial (badly tuned) boot.

-- Terry



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3ECDD771.A36B6832>