Mailing List Archive: [PATCH] [shadow] Disable higher level pagetables early unshadow only when the "process dying" hypercall is used.

[PATCH] [shadow] Disable higher level pagetables early unshadow only when the "process dying" hypercall is used.

Nov 16, 2011, 5:44 AM

Post #1 of 3 (146 views)

This patch fixes a performance problem in fully virtualized guests.

Should be backported to xen-4.0 and xen-4.1. This patch applies (with
slightly different offsets) to all three versions.

Signed-off-by: Gianluca Guida <gianluca.guida@citrix.com>
Tested-by: Jan Beulich <jbeulich@suse.com>

diff -r 068d3d55ce6e xen/arch/x86/mm/shadow/multi.c
--- a/xen/arch/x86/mm/shadow/multi.c Tue Nov 01 19:03:38 2011 +0000
+++ b/xen/arch/x86/mm/shadow/multi.c Wed Nov 16 14:30:57 2011 -0800
@@ -2723,8 +2723,9 @@
|| ( !v->domain->arch.paging.shadow.pagetable_dying_op
&&
v->arch.paging.shadow.last_emulated_mfn_for_unshadow == mfn_x(gmfn) )
)
&& sh_mfn_is_a_page_table(gmfn)
- && !(mfn_to_page(gmfn)->shadow_flags
- & (SHF_L2_32|SHF_L2_PAE|SHF_L2H_PAE|SHF_L4_64)) )
+ && (!v->domain->arch.paging.shadow.pagetable_dying_op ||
+ !(mfn_to_page(gmfn)->shadow_flags
+ & (SHF_L2_32|SHF_L2_PAE|SHF_L2H_PAE|SHF_L4_64))) )
{
perfc_incr(shadow_early_unshadow);
sh_remove_shadows(v, gmfn, 1, 0 /* Fast, can fail to unshadow */ );

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: [PATCH] [shadow] Disable higher level pagetables early unshadow only when the "process dying" hypercall is used. [ In reply to ]

dunlapg at umich

Dec 14, 2011, 7:58 AM

Post #2 of 3 (136 views)

Permalink

On Wed, Nov 16, 2011 at 1:44 PM, Gianluca Guida <glguida@gmail.com> wrote:
> This patch fixes a performance problem in fully virtualized guests.

What performance problem, and how does it fix it? The c/s doesn't
have any information about either, so it's hard for me to evaluate
whether I should pull it into the XenServer patchqueue...

>
> Should be backported to xen-4.0 and xen-4.1. This patch applies (with
> slightly different offsets) to all three versions.
>
> Signed-off-by: Gianluca Guida <gianluca.guida@citrix.com>
> Tested-by: Jan Beulich <jbeulich@suse.com>
>
> diff -r 068d3d55ce6e xen/arch/x86/mm/shadow/multi.c
> --- a/xen/arch/x86/mm/shadow/multi.c Tue Nov 01 19:03:38 2011 +0000
> +++ b/xen/arch/x86/mm/shadow/multi.c Wed Nov 16 14:30:57 2011 -0800
> @@ -2723,8 +2723,9 @@
> || ( !v->domain->arch.paging.shadow.pagetable_dying_op
> &&
> v->arch.paging.shadow.last_emulated_mfn_for_unshadow == mfn_x(gmfn) )
> )
> && sh_mfn_is_a_page_table(gmfn)
> - && !(mfn_to_page(gmfn)->shadow_flags
> - & (SHF_L2_32|SHF_L2_PAE|SHF_L2H_PAE|SHF_L4_64)) )
> + && (!v->domain->arch.paging.shadow.pagetable_dying_op ||
> + !(mfn_to_page(gmfn)->shadow_flags
> + & (SHF_L2_32|SHF_L2_PAE|SHF_L2H_PAE|SHF_L4_64))) )
> {
> perfc_incr(shadow_early_unshadow);
> sh_remove_shadows(v, gmfn, 1, 0 /* Fast, can fail to unshadow */ );
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: [PATCH] [shadow] Disable higher level pagetables early unshadow only when the "process dying" hypercall is used. [ In reply to ]

glguida at gmail

Dec 14, 2011, 8:19 AM

Post #3 of 3 (133 views)

Permalink

Hi George,

On Wed, Dec 14, 2011 at 7:58 AM, George Dunlap <dunlapg@umich.edu> wrote:
> On Wed, Nov 16, 2011 at 1:44 PM, Gianluca Guida <glguida@gmail.com> wrote:
>> This patch fixes a performance problem in fully virtualized guests.
>
> What performance problem, and how does it fix it? The c/s doesn't
> have any information about either, so it's hard for me to evaluate
> whether I should pull it into the XenServer patchqueue...

Have a look at this:

http://old-list-archives.xen.org/archives/html/xen-devel/2011-11/msg00650.html

A good question is whether the performance improvement that brought
Stefano to implement the hypercall in Linux was due to the original
patch creating a regression or if it really helps even after the
performance regression is fixed.

I wasn't here on the Xen world while the patch was tested on unstable,
so i have no idea about it.

Gianluca

--
It was a type of people I did not know, I found them very strange and
they did not inspire confidence at all. Later I learned that I had been
introduced to electronic engineers.
E. W. Dijkstra

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel