gpg: failed to create temporary file '/var/lib/lurker/.#lk0x58582100.hantslug.org.uk.31174': Permission denied
gpg: keyblock resource '/var/lib/lurker/pubring.gpg': Permission denied
gpg: Signature made Sat Jul 14 19:51:38 2007 BST
gpg: using DSA key B2C27BC21C335860
gpg: Can't check signature: No public key
On Sat, Jul 14, 2007 at 07:34:11PM +0100, Victor Churchill wrote:
> On 14/07/07, Jonathan Hudson <jh+hlug@???> wrote:
> >On Sat, 14 Jul 2007 18:18:37 +0100
> >"Victor Churchill" <victorchurchill@???> wrote:
> >> But ...
> >> memtest says it is testing 768 MB.
> >> 'top' says 769500 K.
> >> System Monitor 751.5 MB.
> >> /proc/meminfo 769500 .
> >> dmesg : 762904K/786368K total.
> >>
> >> All I can think of is that one of my mem sticks is faulty and thinks
> >> it is 512 MB, and reports this, but in fact has 'lost' half of it
> >> (address line stuck?)
> >>
>
> >You have a kernel build with the (old) <1G memory model. In this mode
> >Linux does only see 768K.
> >
> >You need to build the kernel with 1+3 or 2+2 (mapping of VM between
> >kernel and user space) or whatever its called.
> >
> >I expect the hardware is fine.
>
> OK ... but rather surprising. This box is running feisty 7.04, and
> uname -a gives
> 2.6.20-16-generic #2 SMP Thu Jun 7 19:00:28 UTC 2007 x86-64 Gnu/Linux
>
> So not an 'old' kernel ... but maybe the default build uses the 'old'
> memory model.
Since it's an amd64 (x86_64) kernel, the memory model is always
flat, and will cope with up to... umm... some very large quantity of
RAM (2^44 bits worth, IIRC). There's no other memory models available.
There's also no appreciable differences in the kernel-space/user-space
split, since there's a huge address space to work in.
On x86, there's three memory models, giving approximately 1G, 4G
and 64G of RAM respectively, plus (for the 4G model) several different
ways of splitting the available address space up between kernel (which
includes the PCI memory map space) and user-space. However, that
doesn't apply in this instance.
It could conceivably be a problem with the PCI address space being
mapped in below the 1G boundary, which can potentially remove a large
chunk of RAM from use if the motherboard isn't all that clever. I
would check for BIOS settings with words like "memory hole", and see
what changing those do.
Hugo.
--
=== Hugo Mills: hugo@... carfax.org.uk | darksatanic.net | lug.org.uk ===
PGP key: 1C335860 from wwwkeys.eu.pgp.net or http://www.carfax.org.uk
--- Python is executable pseudocode; perl ---
is executable line-noise.
