Re: [RFT] kvm with Windows optimization

[Avi Kivity <avi-atKUWr5tajBWk0Htik3J/w@public.gmane.org>]

Anthony Liguori wrote:
>
> Using the following patch, I'm getting really good results.  

> --- a/kernel/mmu.c	2007-10-25 12:36:18.000000000 -0500
> +++ b/kernel/mmu.c	2007-10-25 17:09:55.000000000 -0500
> @@ -280,7 +280,7 @@
>  	if (r)
>  		goto out;
>  	r = mmu_topup_memory_cache(&vcpu->mmu_rmap_desc_cache,
> -				   rmap_desc_cache, 1);
> +				   rmap_desc_cache, 2);
>  	if (r)
>  		goto out;
>  	r = mmu_topup_memory_cache_page(&vcpu->mmu_page_cache, 8);
>   


I think I know what's happening.

We get a write page fault on a page that's not shadowed.  Turns out it's 
a page table page, so we map it read-only, and also rmap it.

Since it's a page table page, we emulate the write and try to map the 
just-written pte as well, since it's likely somebody will access the 
memory pointed-to by the pte.  This consumes another rmap entry, and we 
go boom.

This didn't happen previously, because we didn't rmap read-only pages, 
so the first mapping did not consume an rmap entry.  With swapping 
enabled, we do map read only pages, so we see this.


The correct fix is probably to call mmu_topup_memory_pages() from where 
we do the emulated write, but this seems difficult.

-- 
error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/