[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Publicity] Blog: Xen ARM and DT vs ACPI



> I'm not going to put a subscriber link in the blog but perhaps I'll include the plain link with "(subscribers only until $DATETIME)". I think that is workable. Maybe say somethink like "LWN.net published an article summarizing this discussion, which unfortunately for now is only available to subscribers (until ....)".

I would insert a section called "How Xen uses Device Tree". Many readers may not have followed Xen on ARM that closely. The article may thus dive into too specifics to quickly. I would move "The platform configuration mechanism supported by Xen today is Device Tree. Device Tree is a good fit for our requirements and we will continue to support it as our primary hardware description mechanism." to the beginning of this section and then talk a bit how Xen on ARM uses Device tree.

Probably also worth pointing to http://xenproject.org/developers/teams/arm-hypervisor.html which contains overview material for Xen on ARM. I will add links to the latest videos to that page later today.

I would rename "Xen on ARMâs Position" to "The Xen Project's view on the ACPI vs DT Discussion"

I would start reminding people on how we as a community work: i.e. we work closely with the Linux kernel and other open source upstreams. Maybe worth pointing out one or two lessons we learned from forking upstreams. It also, worth pointing out that we work with Linaro.org and have consistently made the point that ACPI needs to first get accepted into the Linux kernel before we can support it. And that means it needs to be ready enough for Linux maintainers to support it. Any other approach would create pain for users, distros and other downstreams.

> Itâs possible, perhaps even likely, that we will eventually find ourselves needing support ACPI as well as a secondary
> configuration mechanism.
Maybe better: Given that a number of operating system vendors and hardware vendors care about ACPI on ARM and are pushing hard for it, it is possible, perhaps even likely, that we will need to support ACPI in Xen on ARM as secondary configuration mechanism.

> However this will be driven by the existence of actual platforms which only support ACPI. On systems which support both
> ACPI and DT we will continue to prefer Device Tree.
Better: On systems which support both ACPI and DT we will continue to prefer Device Tree. Once ARM hardware platforms that only support ACPI are available, we will obviously need to support ACPI as primary configuration.

> Note that this only applies to Xen on ARM. Xen on x86 will clearly continue to support ACPI so long as it remains
> the dominant hardware description on that platform.
Xen on x86 does and will clearly continue to support ACPI so long as it remains the dominant hardware description on that platform.

The rest of this section could be called something like "Why ACPI and Type I Hypervisors are not a good architectural match"

Our experience with ACPI on x86 taught us that ACPI and Type I Hypervisors are architecturally not well matched. This is also
true for ACPI on ARM.

> While it doesnât seem wise for Xen to be pioneering the use of ACPI on the ARM platform in any case it is also pretty clear > already that the ACPI model is not a good fit for a Type I hypervisor such as Xen, because it essentially enforces a model > where the hypervisor, the kernel, the OSPM (the ACPI term for the bit of an OS which speaks ACPI) and the device drivers
> all must reside in the same privileged entity.
The main reason for this is that the OSPM (the ACPI entity for the portion of an OS which exposes ACPI), the hypervisor and the Dom0? kernel must reside in the same privileged entity. <Maybe an example or a picture showing what would need to run in the same privileged entity (BTW is this a ring?, the term is a little fuzzy at least for me) would help - am willing to draw a nice looking one from a sketch if that helps>

> This impedance mismatch is probably not insurmountable but experience with ACPI on x86 Xen suggests that the resulting
> architecture is not very agreeable.
This mismatch is not insurmountable, but experience with ACPI on x86 Xen suggests that the resulting architecture is not very agreeable. <Maybe also say something about implications for Xen: extra code, bloat, ...>

> If you are a silicon or platform vendor and you care about virtualization and Xen in particular, I encourage you to provide
> a device tree binary for your platform
Thus, if you are a silicon or platform vendor and you care about supporting Type I Hypervisors, I encourage you to provide Device Tree alongside ACPI in your platform. Of course Xen and Linux are not just about servers : both communities have a sizeable community of embedded vendors, for which DT is most likely more suitable than ACPI as configuration mechanism <not sure whether this is true - delete as needed>

Maybe add: More information on Xen on ARM
<Add a link to http://xenproject.org/developers/teams/arm-hypervisor.html, the wiki and maybe a couple of links to XenSummit videos. Maybe http://www.youtube.com/watch?v=Xa33qR0Zopo, http://www.slideshare.net/xen_com_mgr/sruk-xen-presentation2013v7dualandroid, http://www.youtube.com/watch?v=po1IeElg8tg&feature=youtu.be, http://www.youtube.com/watch?v=FiZeOA4qdkQ, http://www.slideshare.net/xen_com_mgr/live-migration-in-xen-arm-samsungjaeyong ... I should have that video shortly>

Cheers
Lars

On 22/11/2013 10:01, Ian Campbell wrote:
Oh good, I was looking/hoping for an LWN link to give some context.
Shame it won't be open to all when the post goes out. I'm not going to
put a subscriber link in the blog but perhaps I'll include the plain
link with "(subscribers only until $DATETIME)".

On Fri, 2013-11-22 at 01:34 +0000, Russell Pavlicek wrote:
You all might find this interesting:

ACPI for ARM?  By Jonathan Corbet
http://lwn.net/SubscriberLink/574439/8083696c52b414d3/

Even if you don't have a subscription to LWN, you should be able to use the 
link (it is normally subscriber-only content, but they allow subscribers to 
create open links like this one).

Russ Pavlicek
Xen Project Evangelist, Citrix Systems
Home Office: +1-301-829-5327
Mobile: +1-240-397-0199
UK VoIP: +44 1223 852 894

-----Original Message-----
From: Lars Kurth
Sent: Thursday, November 21, 2013 12:37 PM
To: Russell Pavlicek; Ian Campbell; #XenOrg
Cc: Dario Faggioli; Lars Kurth; publicity@xxxxxxxxxxxxxxxxxxxx
Subject: RE: Blog: Xen ARM and DT vs ACPI

Hi,

I forgot about Thanksgiving. I had a look through and it is not really 
controversial. But the piece is not very clear either. And because of the Linux 
controversy there is a danger that somebody from the press will quote Ian, 
which could backfire. So it is worth to be clear.

I can spend some time giving feedback tomorrow morning

Given that next week are US holidays, Fridays are usually slow, maybe we can 
wait for a week.

I had a bit of a press search and only 
http://www.pcworld.com/article/2062220/usenix-red-hat-prepares-for-64bit-arm-servers.html
 came up.

Lars

-----Original Message-----
From: Russell Pavlicek
Sent: 21 November 2013 17:26
To: Lars Kurth; Ian Campbell; #XenOrg
Cc: Dario Faggioli; Lars Kurth; publicity@xxxxxxxxxxxxxxxxxxxx
Subject: RE: Blog: Xen ARM and DT vs ACPI

FWIW, if it goes out Tuesday it is unlikely to get any US readership at all.

Thanksgiving week is next week.  Thursday & Friday are holidays for most 
everyone (except the unfortunate folks who work retail or food service).  Wednesday 
is frequently a travel day (many places close around lunchtime to allow people to 
begin travelling to family locations for the holiday).  Tuesday is either a very 
slow day or hideously hectic, depending on the industry.  Either way, not many 
people will be paying attention.

I'd pull the trigger on the piece whenever it makes sense for the piece.  I'll 
push my blog to external sources and do what I can for readership.  Since the 
ARM piece is important, push it when everyone thinks it is ready (whenever that 
is).  But if you wait until next week, I'd suggest maybe Monday (and even that 
is weak) or wait one more week.  Later in the week makes sense only if US 
people are not a major part of the potential audience (seems unlikely, but you 
folks know better than I).

Just my opinion.  Thanks,

Russ Pavlicek
Xen Project Evangelist, Citrix Systems
Home Office: +1-301-829-5327
Mobile: +1-240-397-0199
UK VoIP: +44 1223 852 894


-----Original Message-----
From: Lars Kurth
Sent: Thursday, November 21, 2013 12:16 PM
To: Ian Campbell; #XenOrg
Cc: Dario Faggioli; Lars Kurth; publicity@xxxxxxxxxxxxxxxxxxxx
Subject: RE: Blog: Xen ARM and DT vs ACPI

Adding publicity.

Russell's post just went out, so this week would be unwise now.

Maybe early next week: Tue?  That would also give us a chance to review the 
post thoroughly. This is a sensitive topic and we don't want to hurt ourselves 
unintentionally (which seems to be happening to vendors that are pushing their 
angle too hard). We don't really want that to happen to us.

I would even suggest to wait until Stefano is back from holidays and had a 
chance to catch up on ongoing discussions.

Lars

-----Original Message-----
From: Ian Campbell
Sent: 21 November 2013 17:07
To: #XenOrg
Cc: Dario Faggioli; Lars Kurth
Subject: Blog: Xen ARM and DT vs ACPI

We'd like to say something about our position here, so I've drafted
this: http://blog.xen.org/?p=8347&preview=true

Ideally it would be posted soon, ie.. tomorrow. Does that fit in with the 
schedule?

Ian




_______________________________________________
Publicity mailing list
Publicity@xxxxxxxxxxxxxxxxxxxx
http://lists.xenproject.org/cgi-bin/mailman/listinfo/publicity

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.