-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi,

I'm running the 1.6B snapshot from about a month ago on a Sparc 20
with 223 megs of physical RAM (??? maybe 224... 223 is what dmesg says)
and no swap. I've been meaning to hack cfsd so it wires itself in
memory, and then I can have encrypted swap, but until then, no swap.

I'm writing a perl script that calls dump to write to a file, and then
burns this file to CD with cdrecord. What happens is that dump writes
out 650 megs worth of stuff, and it all goes in the buffer cache, and
then cdrecord can't allocate what it needs.

At the time that cdrecord is called, top reports:

Memory: 130M Act, 65M Inact, 584K Wired, 876K Exec, 119M File, 300K Free

I have a script that totals up the VSZ column for all processes,
which tells me how much memory has been promised to processes. It
reports only 66.60 megabytes promised.

The messages from cdrecord (may be in the wrong order; my script
tends to get stdout and stderr interleaved in strange ways) are:

/usr/pkg/bin/cdrecord: Cannot allocate memory. WARNING: Cannot do mlockall(2).
/usr/pkg/bin/cdrecord: WARNING: This causes a high risk for buffer underruns.
/usr/pkg/bin/cdrecord: Cannot allocate memory. Cannot get mmap for 4198400 Bytes on /dev/zero.

Shouldn't some of that big fat 119 meg buffer cache be getting pushed
out for a process that wants memory?

Or is this a known problem that's been fixed since the snapshot I
have?

Thanks,

	~Ben
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (NetBSD)
Comment: For info see http://www.gnupg.org

iD8DBQE9QwNxP3EgUnITKiERAtuUAJ4rtmSNdKXwPaqbgPwOloqPcPGkXQCfXOtE
qFPpxspG/bOHRhOotaxvlVw=
=Vq9u
-----END PGP SIGNATURE-----