work around for populate physmap hypercall

Hi.
This patch is work around for xencomm and the soft lockup
message.
However I need help before commit.

- It is necessary to test it before commit with much
memory.
  I haven't tested it with much memory because my
environment doesn't have
  such huge memory.
- I defined MEMORYOP_MAX_EXTENTS based on the xencomm
limitation.
  Howver I'm not sure the value is small enough to avoid the
soft lockup
  message because I couldn't produce the message on my
environment.
  It is necessary to determine the value with expriments.

Thanks,
-- 
yamahata

_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devellists.xensource.com
http://list
s.xensource.com/xen-ia64-devel

On Mon, 2007-08-20 at 12:23 +0900, Isaku Yamahata wrote:
> Hi.
> This patch is work around for xencomm and the soft
lockup message.
> However I need help before commit.
> 
> - It is necessary to test it before commit with much
memory.
>   I haven't tested it with much memory because my
environment doesn't have
>   such huge memory.
> - I defined MEMORYOP_MAX_EXTENTS based on the xencomm
limitation.
>   Howver I'm not sure the value is small enough to
avoid the soft lockup
>   message because I couldn't produce the message on my
environment.
>   It is necessary to determine the value with
expriments.

Hi Isaku,

   The defined MEMORYOP_MAX_EXTENTS is still a bit too big,
I get soft
lockups with large domains.  Further testing shows dividing
your
MEMORYOP_MAX_EXTENTS calculation by 4 gives me nearly 1
million extents,
and eliminates the soft lockup.  I've tested up to a 90G
domain with no
problems on startup.  Here's the macro I end up with:

#define MEMORYOP_MAX_EXTENTS                                
           
                ((((((PAGE_SIZE - sizeof(struct
xencomm_desc)) /        
                   sizeof(uint64_t)) - 2) * PAGE_SIZE) /    
           
                 
sizeof(*xen_guest_handle(kern_op.extent_start))) / 4)

Thanks,

	Alex

-- 
Alex Williamson                             HP Open Source
& Linux Org.


_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devellists.xensource.com
http://list
s.xensource.com/xen-ia64-devel

Thank you very much for testing/experimenting.
Attached the updated patch.
Only the constants and comments were changed.


On Mon, Aug 20, 2007 at 09:35:30AM -0600, Alex Williamson
wrote:
> On Mon, 2007-08-20 at 12:23 +0900, Isaku Yamahata
wrote:
> > Hi.
> > This patch is work around for xencomm and the soft
lockup message.
> > However I need help before commit.
> > 
> > - It is necessary to test it before commit with
much memory.
> >   I haven't tested it with much memory because my
environment doesn't have
> >   such huge memory.
> > - I defined MEMORYOP_MAX_EXTENTS based on the
xencomm limitation.
> >   Howver I'm not sure the value is small enough to
avoid the soft lockup
> >   message because I couldn't produce the message
on my environment.
> >   It is necessary to determine the value with
expriments.
> 
> Hi Isaku,
> 
>    The defined MEMORYOP_MAX_EXTENTS is still a bit too
big, I get soft
> lockups with large domains.  Further testing shows
dividing your
> MEMORYOP_MAX_EXTENTS calculation by 4 gives me nearly 1
million extents,
> and eliminates the soft lockup.  I've tested up to a
90G domain with no
> problems on startup.  Here's the macro I end up with:
> 
> #define MEMORYOP_MAX_EXTENTS                           
                
>                 ((((((PAGE_SIZE - sizeof(struct
xencomm_desc)) /        
>                    sizeof(uint64_t)) - 2) * PAGE_SIZE)
/                
>                  
sizeof(*xen_guest_handle(kern_op.extent_start))) / 4)
> 
> Thanks,
> 
> 	Alex
> 
> -- 
> Alex Williamson                             HP Open
Source & Linux Org.
> 
> 
> _______________________________________________
> Xen-ia64-devel mailing list
> Xen-ia64-devellists.xensource.com
> http://list
s.xensource.com/xen-ia64-devel
> 

-- 
yamahata

_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devellists.xensource.com
http://list
s.xensource.com/xen-ia64-devel

On Wed, 2007-08-22 at 14:47 +0900, Isaku Yamahata wrote:
> Thank you very much for testing/experimenting.
> Attached the updated patch.
> Only the constants and comments were changed.

   Applied.  Nice to have this limit removed.  Thanks,

	Alex

-- 
Alex Williamson                             HP Open Source
& Linux Org.


_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devellists.xensource.com
http://list
s.xensource.com/xen-ia64-devel