[Xen-ia64-devel] RE: Paravirtualization changes in the Linux kernel ?

["Magenheimer, Dan (HP Labs Fort Collins)" <dan.magenheimer@xxxxxx>]

Yes, agreed.  I think it is a reasonable goal for the Xen/ia64 community

to see the paravirtualization changes required for Xen/ia64 in the same

Linux release as the Xen/x86 changes, whatever release that turns out

to be.

 

Dan

---

From: jean-paul.pigache@xxxxxxxx [mailto:jean-paul.pigache@xxxxxxxx]
Sent: Wednesday, February 15, 2006 7:52 AM
To: Magenheimer, Dan (HP Labs Fort Collins)
Cc: xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
Subject: Paravirtualization changes in the Linux kernel ?

Dan,

well... it appears that we are still very far from seeing ia64 Xen paravirtualization changes in the standard Linux kernel.

I have copied the xen-ia64-devel list to share your answers.
Thank you.
Jean-Paul

"Magenheimer, Dan (HP Labs Fort Collins)" <dan.magenheimer@xxxxxx> 13/02/2006 18:09

Pour :        <jean-paul.pigache@xxxxxxxx>         cc :        "Yang, Fred" <fred.yang@xxxxxxxxx>, "Yoshi. Oguchi" <y-oguchi@xxxxxxxxxxxxxx>         Objet :        RE: Réf. : RE: Réf. : RE: Xen/ia64 progress

Hi Jean-Paul --
 
There are (at least) three groups of Linux paravirtualization changes, let's call them:
1) virtual cpu paravirtualization changes
2) virtual memory paravirtualization changes
3) virtual driver paravirtualization changes
 
All three groups need to be pushed upstream.  None have been accepted
into Linux yet.
 
On x86, the virtual cpu and virtual memory changes have been discussed and
cleaned up on the xen-merge list, but I haven't seen any any target date or target
Linux version for pushing upstream... it is already too late to get them into 2.6.16.
 
On ia64, I proposed a Linux/ia64 patch for the virtual cpu part last August:
http://www.gelato.unsw.edu.au/archives/linux-ia64/0508/15089.html
Tony Luck declined to accept it: "I'm waiting to see how the integration in other
architectures goes.  I don't see that there is a big rush to push this into
the base."  At that time, we thought that the Xen/x86 changes might
be pushed upstream in 2.6.14 or 2.6.15 so I was eager to get the
Xen/ia64 changes in as well.  Tony's caution proved prescient as the
Xen/ia64 virtual cpu paravirtualization changes are not useful without
the Xen virtual drivers (except only to boot dom0).
 
(Note however that the Xen/ia64 virtual cpu patch is very clean and has applied
nicely to 2.6.14 and 2.6.15 and 2.6.16-rc2.)
 
With domU being VP and dom0 being P==M, there are no virtual memory
paravirtualization changes required for Xen/ia64.  When dom0 (and driver
domains) switches to VP+DMA, there will be some changes required.
Until we know the extent of those changes, it is hard to guess how difficult
it will be to push them upstream.
 
I haven't seen any attempts to submit the Xen VBD and VNIF drivers into
Linux.  This concerns me because I think there will be many objections
and long discussions from the Linux driver community.

Thanks,
Dan
 
P.S. Is there any reason this discussion is not on the public list?  Feel
free to edit/reply to the list if you prefer.

---

From: jean-paul.pigache@xxxxxxxx [mailto:jean-paul.pigache@xxxxxxxx]
Sent: Monday, February 13, 2006 9:35 AM
To: Magenheimer, Dan (HP Labs Fort Collins)
Cc: Yang, Fred; Yoshi. Oguchi
Subject: Réf. : RE: Réf. : RE: Xen/ia64 progress


Dan,

could you give a status about pushing the changes for paravirtualization into the standard Linux kernel ?
Do you know if the current discussions with the Linux community include X86 (and EM64T and X64) and IPF ?
Or do you think that there a risk than the Linux community accepts the changes for X86 (and EM64T and X64) only, in a first step ?

Jean-Paul

_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel