KVM call minutes July 31th
Hi
Today minutes:
- Avi fixed and got 9p working
- QEMU 1.2 Test Day (stefan)
* Let's find -rc bugs and ensure the release is stable
* We've done this in the past and have a wiki template but can discuss suggestions on the call
* A date?
- 1.2 plans for CPU model versioning/compatibility (eduardo) (global properties vs QOM vs qdev)
how to do it ? configuration file? moving back to the code? different external interface from internal one
Eduardo/Andreas/Anthony, <fill here>
[libvirt] QEMU CPU model versioning/compatibility (was Re: KVM call minutes July 31th)
Quoted 4 lines · Juan Quintela
- 1.2 plans for CPU model versioning/compatibility (eduardo) (global properties vs QOM vs qdev)
how to do it ? configuration file? moving back to the code? different external interface from internal one
(CCing libvir-list)
So, the problem is (please correct me if I am wrong):
- libvirt expects the CPU models from the current config file to be provided by QEMU. libvirt won't write a cpudef config file.
- Fixing compatibility problems on the CPU models (even the ones that are in config files) are expected to be QEMU's responsibility.
Moving the CPU models inside the code is a step backwards, IMO. I don't think loading some kind of data from an external file (provided and maintained by QEMU itself) should be considered something special and magic, and make the data there a second-class citizen (that QEMU refuses to give proper support and keep compatibility).
But if it will make us stop running in circles every time something related to those files needs to be changed (remember the -no-user-config discussion?), I think I will accept that.
Signature · 1 line
Eduardo
Re: QEMU CPU model versioning/compatibility (was Re: KVM call minutes July 31th)
Quoted 15 lines · Eduardo Habkost
Quoted 4 lines · Juan Quintela
- 1.2 plans for CPU model versioning/compatibility (eduardo) (global properties vs QOM vs qdev)
how to do it ? configuration file? moving back to the code? different external interface from internal one(CCing libvir-list)
So, the problem is (please correct me if I am wrong):
- libvirt expects the CPU models from the current config file to be provided by QEMU. libvirt won't write a cpudef config file.
- Fixing compatibility problems on the CPU models (even the ones that are in config files) are expected to be QEMU's responsibility.
Quoted 5 lines
Moving the CPU models inside the code is a step backwards, IMO. I don't think loading some kind of data from an external file (provided and maintained by QEMU itself) should be considered something special and magic, and make the data there a second-class citizen (that QEMU refuses to give proper support and keep compatibility).
I agree.
Quoted 3 lines
But if it will make us stop running in circles every time something related to those files needs to be changed (remember the -no-user-config discussion?), I think I will accept that.
The issue is that we have a lot of machinery for backwards compatibility in the code, and none in cpu definitions parser. Yes we could mark up the cpu definitions so it could be text driven, but that's effort that could be spent elsewhere. Moving it to a data-driven C implementation that can rely on the existing backwards compat code is a good tradeoff IMO.
Signature · 1 line
error compiling committee.c: too many arguments to function
Quoted 4 lines · Juan Quintela
- 1.2 plans for CPU model versioning/compatibility (eduardo) (global properties vs QOM vs qdev)
how to do it ? configuration file? moving back to the code? different external interface from internal one
Another question:
Anthony, you mentioned that you had a series to convert the CPU to be actual qdev device objects. Do you have a pointer for it? I couldn't find it.
All I remember (from a call 2 or 3 weeks ago) was that the conversion was difficult because it would make *-user pull qdev, and that was not a trivial task.
Signature · 1 line
Eduardo
Am 31.07.2012 17:40, schrieb Eduardo Habkost:
Quoted 8 lines · Juan Quintela
Quoted 4 lines
- 1.2 plans for CPU model versioning/compatibility (eduardo) (global properties vs QOM vs qdev)
how to do it ? configuration file? moving back to the code? different external interface from internal oneAnother question:
Quoted 7 lines
Anthony, you mentioned that you had a series to convert the CPU to be actual qdev device objects. Do you have a pointer for it? I couldn't find it.
All I remember (from a call 2 or 3 weeks ago) was that the conversion was difficult because it would make *-user pull qdev, and that was not a trivial task.
We had a KVM call on that topic. After that I was on vacation, we had our Hackweek and this week is the one I started diving into QOM again - Anthony said he already had something new and would post that.
I have added a feature item on the 1.2 feature page for the Soft Freeze.
Andreas
Signature · 2 lines
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg
----- Original Message -----
Quoted 12 lines
From: "Eduardo Habkost" <[email protected]>
To: "Juan Quintela" <[email protected]>, "Anthony Liguori" <[email protected]>
Cc: [email protected], "KVM devel mailing list" <[email protected]>
Sent: Tuesday, July 31, 2012 5:40:14 PM
Subject: Re: [Qemu-devel] KVM call minutes July 31thQuoted 4 lines · Juan Quintela
- 1.2 plans for CPU model versioning/compatibility (eduardo) (global properties vs QOM vs qdev)
how to do it ? configuration file? moving back to the code? different external interface from internal one
Quoted 6 lines
Another question:
Anthony, you mentioned that you had a series to convert the CPU to be
actual qdev device objects. Do you have a pointer for it? I couldn't find it.
it was http://lists.gnu.org/archive/html/qemu-devel/2012-06/msg03847.html
but there is another earlier proposal by Andreas to make CPU as DEVICE only for softmmu targets.
So consensus wasn't reached yet.
If we are to to use global props creating CPU subclasses, we might have to make CPU as DEVICE for *-user target as well and pull in qdev deps into it.
Quoted 9 lines
All I remember (from a call 2 or 3 weeks ago) was that the conversion was difficult because it would make *-user pull qdev, and that was not a
trivial task.--
Eduardo
Am 31.07.2012 18:06, schrieb Igor Mammedov:
Quoted 19 lines
----- Original Message -----
Quoted 12 lines
From: "Eduardo Habkost" <[email protected]>
To: "Juan Quintela" <[email protected]>, "Anthony Liguori" <[email protected]>
Cc: [email protected], "KVM devel mailing list" <[email protected]>
Sent: Tuesday, July 31, 2012 5:40:14 PM
Subject: Re: [Qemu-devel] KVM call minutes July 31thQuoted 4 lines · Juan Quintela
- 1.2 plans for CPU model versioning/compatibility (eduardo) (global properties vs QOM vs qdev)
how to do it ? configuration file? moving back to the code? different external interface from internal oneQuoted 6 lines
Another question:
Anthony, you mentioned that you had a series to convert the CPU to be
actual qdev device objects. Do you have a pointer for it? I couldn't find it.
Quoted 7 lines
it was http://lists.gnu.org/archive/html/qemu-devel/2012-06/msg03847.html
but there is another earlier proposal by Andreas to make CPU as DEVICE only for softmmu targets.
So consensus wasn't reached yet.If we are to to use global props creating CPU subclasses, we might have to make CPU as DEVICE for *-user target as well and pull in qdev deps into it.
We don't version the user emulators, so there's no compatibility properties to set if we follow the machine versioning scheme. I expect Nehalem to resolve to Nehalem-1.2 but there's no strict need to have a Nehalem-1.1 since there's no concept of migration.
Anything involving qdev in *-user will not make it in time for 1.2! :)
Andreas
Signature · 2 lines
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg