Overhead incurred due to Xen for a page fault

Overhead incurred due to Xen for a page fault

[pak333]

[-- Attachment #1.1: Type: text/plain, Size: 176 bytes --]

I would like to get some help on estimating the cost of page fault handling in Xen. Can someone help me identify where I need to put the hooks into the xen source.

Thanks
Kris

[-- Attachment #1.2: Type: text/html, Size: 254 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Overhead incurred due to Xen for a page fault

[Randy Thelen]

pak333@comcast.net wrote:

> I would like to get some help on estimating the cost of page fault  
> handling in Xen. Can someone help me identify where I need to put  
> the hooks into the xen source.

Check arch/x86/traps.c, look for the function:

asmlinkage int do_page_fault(struct cpu_user_regs *regs)

It has some hooks for measuring counts.  You can add timing information.

Note that page faults are fundamentally handled by the guest O/S, Xen  
simply forwards them.

-- Randy

Re: Overhead incurred due to Xen for a page fault

[Dave Winchell]

Fault path:

vmx_vmexit_handler
   vmx_do_page_fault
      shadow_fault (shadow_fault_64)

Put memory based trace code in shadow_fault_64, or, I presume, 
shadow_fault_32.
You'll see shadow_fault_64 get entered multiple times. I believe the 
faults are as
follows:
  1. The data page fault.
  2. The fault for the gos accessing the pte.
  3. The second fault for the data page.

This is for fully virtualized mode. I have not been looking at 
paravirtualized.
I see page fault times at 20-25 usec.

regards,
Dave



pak333@comcast.net wrote:

> I would like to get some help on estimating the cost of page fault 
> handling in Xen. Can someone help me identify where I need to put the 
> hooks into the xen source.
>  
> Thanks
> Kris
>
>------------------------------------------------------------------------
>
>_______________________________________________
>Xen-devel mailing list
>Xen-devel@lists.xensource.com
>http://lists.xensource.com/xen-devel
>  
>

Re: Overhead incurred due to Xen for a page fault

[pak333]

[-- Attachment #1.1: Type: text/plain, Size: 474 bytes --]

Hi, 
Thanks for the pointer for the do_page_fault function!

I need to find out whether the page fault was raised by GuestOS or HostOS inside do_page_fault() function. Is there a way to identify this ?
If so how?

Thanks


-------------- Original message -------------- 
From: pak333@comcast.net 

I would like to get some help on estimating the cost of page fault handling in Xen. Can someone help me identify where I need to put the hooks into the xen source.

Thanks
Kris

[-- Attachment #1.2: Type: text/html, Size: 787 bytes --]

[-- Attachment #2: Type: message/rfc822, Size: 615 bytes --]

[-- Attachment #2.1.1: Type: text/plain, Size: 138 bytes --]

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

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

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