Xen devel - Oct 2005 - [PATCH][RESEND] Control panel changes for IA64/VTI

>From: Keir Fraser [mailto:Keir.Fraser@cl.cam.ac.uk]
>Sent: 2005年10月22日 14:26
>On 22 Oct 2005, at 04:20, Tian, Kevin wrote:
>
>> 	Attached are the necessary changes in control panel to support
>> ia64/vti domain builder. One is in common side, and the other is in vti
>> specific image builder. Please take a look, and if no objection,
I''d
>> like to ask Dan''s help to check into xen-ia64-unstable.hg
first.
>
>The common changes to libxc look ia64 specific. Why not stick all of
>them in ia64_stubs?
OK, see updated version and I moved all interfaces within "ifdef
__ia64__" to ia64_stubs.
>
>Adding code to memset to force into the TLB won''t be accepted.
Either
>those whole functions are ia64 specific or (better) replace use of
>mlock/munlock with map_to_hypervisor/unmap_from_hypervisor which do
>mlock/memset/munlock on ia64 and skip the memset on x86.
>
>  -- Keir
Actually those changes are all ia64 specific. With normal diff command, I can
use "-p" option to add one line like "@@ -322,46 +322,6 @@ int
xc_ia64_get_pfn_list(int xc_handle,". However I don''t know how to
produce same format with "hg diff". That brings some trouble to people
when similar code pieces exist in same file. ;-( I also removed memset in
xc_ia64_get_pfn_list since mlock seems to have the effect.

Thanks,
Kevin




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

BTW, the reason for an ia64 specific get_pfn_list (by Matt) is that machine
pages for domain on ia64 are allocated on demand. So there''s
requirement for parameters to contain start_pfn and nr_pages so that Xen can
allocate necessary pages for that specific range. On x86 side, since all the
pages are allocated when domain creation, get_pfn_list is aimed to query once
for all pages and so the parameter can be only nr_pages. Just the background
here. ;-)

Thanks,
Kevin
>-----Original Message-----
>From: xen-devel-bounces@lists.xensource.com
>[mailto:xen-devel-bounces@lists.xensource.com] On Behalf Of Tian, Kevin
>Sent: 2005年10月23日 11:47
>To: Keir Fraser
>Cc: xen-devel@lists.xensource.com; Xen-IA64-Devel
>Subject: [Xen-devel] [PATCH][RESEND] Control panel changes for IA64/VTI
>
>>From: Keir Fraser [mailto:Keir.Fraser@cl.cam.ac.uk]
>>Sent: 2005年10月22日 14:26
>>On 22 Oct 2005, at 04:20, Tian, Kevin wrote:
>>
>>> 	Attached are the necessary changes in control panel to support
>>> ia64/vti domain builder. One is in common side, and the other is in
vti
>>> specific image builder. Please take a look, and if no objection,
I''d
>>> like to ask Dan''s help to check into xen-ia64-unstable.hg
first.
>>
>>The common changes to libxc look ia64 specific. Why not stick all of
>>them in ia64_stubs?
>
>OK, see updated version and I moved all interfaces within "ifdef
__ia64__" to
>ia64_stubs.
>
>>
>>Adding code to memset to force into the TLB won''t be accepted.
Either
>>those whole functions are ia64 specific or (better) replace use of
>>mlock/munlock with map_to_hypervisor/unmap_from_hypervisor which do
>>mlock/memset/munlock on ia64 and skip the memset on x86.
>>
>>  -- Keir
>
>Actually those changes are all ia64 specific. With normal diff command, I
can use "-p"
>option to add one line like "@@ -322,46 +322,6 @@ int
xc_ia64_get_pfn_list(int
>xc_handle,". However I don''t know how to produce same format
with "hg diff". That
>brings some trouble to people when similar code pieces exist in same file.
;-( I also
>removed memset in xc_ia64_get_pfn_list since mlock seems to have the effect.
>
>Thanks,
>Kevin

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

On 23 Oct 2005, at 04:53, Tian, Kevin wrote:
> BTW, the reason for an ia64 specific get_pfn_list (by Matt) is that 
> machine pages for domain on ia64 are allocated on demand. So
there''s
> requirement for parameters to contain start_pfn and nr_pages so that 
> Xen can allocate necessary pages for that specific range. On x86 side, 
> since all the pages are allocated when domain creation, get_pfn_list 
> is aimed to query once for all pages and so the parameter can be only 
> nr_pages. Just the background here. ;-)
We could have a common interface then. All the uses of get_pfn_list in 
common or x86 code will have start_pfn==0.

The patches look okay, although the Python bits aren''t particularly 
pretty.

  -- Keir


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

>From: Keir Fraser [mailto:Keir.Fraser@cl.cam.ac.uk]
>Sent: 2005年10月23日 16:15
>
>On 23 Oct 2005, at 04:53, Tian, Kevin wrote:
>
>> BTW, the reason for an ia64 specific get_pfn_list (by Matt) is that
>> machine pages for domain on ia64 are allocated on demand. So
there''s
>> requirement for parameters to contain start_pfn and nr_pages so that
>> Xen can allocate necessary pages for that specific range. On x86 side,
>> since all the pages are allocated when domain creation, get_pfn_list
>> is aimed to query once for all pages and so the parameter can be only
>> nr_pages. Just the background here. ;-)
>
>We could have a common interface then. All the uses of get_pfn_list in
>common or x86 code will have start_pfn==0.
OK. I can do it later when there''s no other requirement for an ia64
specific version. I believe that can be done just after current domU enable on
ia64 will get solved at hand.
>
>The patches look okay, although the Python bits aren''t particularly
>pretty.
>
>  -- Keir
Sorry for bad Python skill. ;-) So Dan, please help check in.

Thanks,
Kevin

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

On Mon, Oct 24, 2005 at 09:19:33AM +0800, Tian, Kevin wrote:
> >From: Keir Fraser [mailto:Keir.Fraser@cl.cam.ac.uk]
> >
> >The patches look okay, although the Python bits aren''t
particularly
> >pretty.
> >
> >  -- Keir
> 
> Sorry for bad Python skill. ;-) So Dan, please help check in.
Actually executing your code would be a good start.

--- a/tools/python/xen/xend/image.py	Sat Oct 15 03:09:43 2005
+++ b/tools/python/xen/xend/image.py	Fri Oct 21 15:35:26 2005
@@ -58,6 +58,7 @@
     """
 
     ostype = None
+    page_kb = 4
 
 
     def __init__(self, vm, imageConfig, deviceConfig):
@@ -366,11 +367,13 @@
 
     def getDomainMemory(self, mem):
         """@see ImageHandler.getDomainMemory"""
+	if os.uname[4] == ''ia64'':
+	    page_kb = 16
         # for ioreq_t and xenstore
         static_pages = 2

This is going to fail with a TypeError exception, because os.uname is a
function.  You want os.uname()[4].

Ewan.

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

Ewan, thanks for pointing out following obvious error, and I need to make patch
more carefully next time. ;-)

Attached the updated version. To reduce the mail size, I skipped attaching the
other untouched one.

Thanks,
Kevin
>From: Ewan Mellor [mailto:ewan@xensource.com]
>Sent: 2005年10月24日 18:07
>On Mon, Oct 24, 2005 at 09:19:33AM +0800, Tian, Kevin wrote:
>
>> >From: Keir Fraser [mailto:Keir.Fraser@cl.cam.ac.uk]
>> >
>> >The patches look okay, although the Python bits aren''t
particularly
>> >pretty.
>> >
>> >  -- Keir
>>
>> Sorry for bad Python skill. ;-) So Dan, please help check in.
>
>Actually executing your code would be a good start.
>
>--- a/tools/python/xen/xend/image.py	Sat Oct 15 03:09:43 2005
>+++ b/tools/python/xen/xend/image.py	Fri Oct 21 15:35:26 2005
>@@ -58,6 +58,7 @@
>     """
>
>     ostype = None
>+    page_kb = 4
>
>
>     def __init__(self, vm, imageConfig, deviceConfig):
>@@ -366,11 +367,13 @@
>
>     def getDomainMemory(self, mem):
>         """@see
ImageHandler.getDomainMemory"""
>+	if os.uname[4] == ''ia64'':
>+	    page_kb = 16
>         # for ioreq_t and xenstore
>         static_pages = 2
>
>This is going to fail with a TypeError exception, because os.uname is a
>function.  You want os.uname()[4].
>
>Ewan.

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