Re: [PATCH] x86: 46 bit PAE support

From: Ingo Molnar
Date: Thu May 07 2009 - 10:49:53 EST



* H. Peter Anvin <hpa@xxxxxxxxx> wrote:

> Ingo Molnar wrote:
> >
> > Yes, struct page is ~64 bytes, and 64*64 == 4096.
> >
> > Alas, it's not a problem: my suggestion wasnt to simulate 64 TB of
> > RAM. My suggestion was to create a sparse physical memory map (in a
> > virtual machine) that spreads ~1GB of RAM all around the 64 TB
> > physical address space. That will test whether the kernel is able to
> > map and work with such physical addresses. (which will cover most of
> > the issues)
> >
> > A good look at /debug/x86/dump_pagetables with such a system booted
> > up would be nice as well - to make sure every virtual memory range
> > is in its proper area, and that there's enough free space around
> > them.
> >
>
> We're working on simulating this at Intel. We should hopefully be
> able to test this next week.

Wow, very nice!

It would be nice to do it on a KVM basis and submit the
weird-memory-layout submission to the KVM tree. It would be helpful
with the reproduction of weird, memory layout dependent bugs too for
example. Plus we could create a test facility that randomizes the
physical memory layout (with a given fragmentation level).

Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/