Re: [Qemu-devel] Re: [PATCH 1/6] Use correct types to enable > 2G support

[Philip Boulain]

On 1 Feb 2008, at 16:09, M. Warner Losh wrote:
> In message: <address@hidden>
>             Robert William Fuller <address@hidden> writes:
> : Avi Kivity wrote:
> : > Anthony Liguori wrote:
> : >> I think I'll change this too into a single qemu_ram_alloc.   
> That will
> : >> fix the bug with KVM when using -kernel and large memory  
> anyway :-)
> : > Won't that cause all of the memory in the hole to be wasted?
> : Linux doesn't commit mapped memory until it's faulted.  As for other
> : platforms, who knows?

It would appear that modern Windows also overcommits:

"This memory isn’t allocated until the application explicitly uses  
it.  Once the application uses the page, it becomes committed."
"When an application touches a virtual memory page (reads/write/ 
programmatically commits) the page becomes a committed page.  It is  
now backed by a physical memory page.  This will usually be a  
physical RAM page, but could eventually be a page in the page file on  
the hard disk, or it could be a page in a memory mapped file on the  
hard disk."
 -- http://blogs.msdn.com/ntdebugging/archive/2007/10/10/the-memory- 
shell-game.aspx

So it looks like you could get away with this on the two big host  
platforms.

> Most BSDs are also similarly overcommitted. 95% of the users think
> this is a feature, but the other 5 argue 20 times harder sometimes :-(

Some of us don't like the idea that our operating systems lie about  
how many resources they have available, then have to club innocent  
processes over the head when their lies catch up with them. ;)

Phil