[ANNOUNCE] kvm-79 release

[ANNOUNCE] kvm-79 release

[Avi Kivity]

This is the first release to fully support pci device assignment.  You 
can assign a pci device to qemu on the command line, or hot-plug it in 
via the monitor.  Note that at this time, Linux 2.6.28 is required on 
the host.

Upstream qemu recently gained kvm support.  At this time a lot is 
missing in upstream (smp, performance) so this the upstream capabilities 
are not used yet.  Over time we will switch to using qemu upstream for 
more functionality.

Changes from kvm-78:
- merge qemu-svn
   - fix qcow2 problems with scsi
   - 'info chardev' monitor command
- device assignment userspace (Amit Shah, Muli Ben Yehudah, Ben-Ami Yassour,
   Weidong Han, Or Sagi, Nir Peleg, Glauber Costa, Xiantao Zhang)
- multiple processor infrastructure for ppc (Hollis Blanchard)
- fix error handling in eventfd() emulation (Mark McLoughlin)
- handle large mtu with virtio-net (Mark McLoughlin)
- move x86 specific device assignment code to x86 files (Christian Ehrhardt)
- fix 'pci_add' command descriptor (Weidong Han)
- enable pci function level reset for device assignment (Sheng Yang)
- fix incorrect handling of aliases gfns (Izik Eidus)
- ppc optimizations (Hollis Blanchard, Christian Ehrhardt)
- disallow guests from setting memory type when using EPT (Sheng Yang)
- fix cpuid leaf 11 loop termination (Nitin A Kamble)
- fix cpuid multiple leaf iteration (Nitin A Kamble)
- fix ia64 uniprocessor build (Xiantao Zhang)
- fix ppc Kbuild constraints (Hollis Blanchard)
- fix pit initialization memory leak

Notes:
     If you use the modules bundled with kvm-79, you can use any version
of Linux from 2.6.16 upwards.  You may also use kvm-79 userspace with
the kvm modules provided by Linux 2.6.25 or above.  Some features may
only be available in newer releases.

http://kvm.qumranet.com

-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.

Re: [ANNOUNCE] kvm-79 release

[Farkas Levente]

[-- Attachment #1: Type: text/plain, Size: 822 bytes --]

Avi Kivity wrote:
> This is the first release to fully support pci device assignment.  You
> can assign a pci device to qemu on the command line, or hot-plug it in
> via the monitor.  Note that at this time, Linux 2.6.28 is required on
> the host.
> 
> Upstream qemu recently gained kvm support.  At this time a lot is
> missing in upstream (smp, performance) so this the upstream capabilities
> are not used yet.  Over time we will switch to using qemu upstream for
> more functionality.

hi,
- guest fedora-9 latest kernel-2.6.26.6-79.fc9.i686 still not boot,
- guest centos-5 x86_64's kernel-2.6.18-92.1.17.el5 still gives the same
crash as kvm-78 (screenshot attached),
- guest mandrake-10 still can't boot
so imho still not the best release:-(

-- 
  Levente                               "Si vis pacem para bellum!"

[-- Attachment #2: devel-x86-64-kvm-79.png --]
[-- Type: image/png, Size: 28634 bytes --]

Re: [ANNOUNCE] kvm-79 release

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1248 bytes --]

Farkas Levente wrote:
> Avi Kivity wrote:
>> This is the first release to fully support pci device assignment.  You
>> can assign a pci device to qemu on the command line, or hot-plug it in
>> via the monitor.  Note that at this time, Linux 2.6.28 is required on
>> the host.
>>
>> Upstream qemu recently gained kvm support.  At this time a lot is
>> missing in upstream (smp, performance) so this the upstream capabilities
>> are not used yet.  Over time we will switch to using qemu upstream for
>> more functionality.
> 
> hi,
> - guest fedora-9 latest kernel-2.6.26.6-79.fc9.i686 still not boot,
> - guest centos-5 x86_64's kernel-2.6.18-92.1.17.el5 still gives the same
> crash as kvm-78 (screenshot attached),

Those two should be fine with this patch [1]. You are welcome to test it
and be the first (mmmh...) to provide feedback on it.

> - guest mandrake-10 still can't boot

Can't help here, your test boots fine on OpenSuse's 2.6.25.18-0.2 with
kvm.git for me as well.

> so imho still not the best release:-(
> 

Bugs happen, and some corner cases may remain unfixed for a longer time
as they happen to be hard to track down.

Jan

[1] http://permalink.gmane.org/gmane.comp.emulators.kvm.devel/24409


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 257 bytes --]

Re: [ANNOUNCE] kvm-79 release

[Charles Duffy]

Jan Kiszka wrote:
> Farkas Levente wrote:
>> - guest fedora-9 latest kernel-2.6.26.6-79.fc9.i686 still not boot,
>> - guest centos-5 x86_64's kernel-2.6.18-92.1.17.el5 still gives the same
>> crash as kvm-78 (screenshot attached),
> 
> Those two should be fine with this patch [1]. You are welcome to test it
> and be the first (mmmh...) to provide feedback on it.

I've been able to reproduce the issue on CentOS 5 x86_64 guests quite 
reliably; after applying the patch to kvm-79 (with the necessary 
filename munging), this issue does indeed appear to be resolved.

Re: [ANNOUNCE] kvm-79 release

[Martin Vogt]

Hello,


[...]
>Notes:
>If you use the modules bundled with kvm-79, you can use any version
>of Linux from 2.6.16 upwards.

Nope.
I tested kvm-79 on a SLES 10 with 2.6.16,
in the hope that my previously reported bug:

"hrtimer_start: Unknown symbol in module"

was fixed. (It wasnt)

Now I have this "one-liner patch(tm)" here which works for me:


diff -u external-module-compat-comm.h.org external-module-compat-comm.h
--- external-module-compat-comm.h.org   2008-11-12 15:51:55.000000000 +0100
+++ external-module-compat-comm.h       2008-11-12 15:52:05.000000000 +0100
@@ -601,7 +601,7 @@

 static inline void hrtimer_start_expires(struct hrtimer *timer, int mode)
 {
-       hrtimer_start(timer, timer->expires, mode);
+       hrtimer_start_p(timer, timer->expires, mode);
 }



Is this a correct fix?

- I could boot suse 11.1 beta4 with it(which didnt work with kvm-77)

Im really not sure if this works corretly and I dont want have broken
modules in my kernel :(

Is hrtimer_start_p already initialized when  hrtimer_start_expires is
called? (help)

regards,

Martin

Re: [ANNOUNCE] kvm-79 release

[Avi Kivity]

Martin Vogt wrote:
> Hello,
>
>
> [...]
>   
>> Notes:
>> If you use the modules bundled with kvm-79, you can use any version
>> of Linux from 2.6.16 upwards.
>>     
>
> Nope.
> I tested kvm-79 on a SLES 10 with 2.6.16,
> in the hope that my previously reported bug:
>
> "hrtimer_start: Unknown symbol in module"
>
> was fixed. (It wasnt)
>
> Now I have this "one-liner patch(tm)" here which works for me:
>
>
> diff -u external-module-compat-comm.h.org external-module-compat-comm.h
> --- external-module-compat-comm.h.org   2008-11-12 15:51:55.000000000 +0100
> +++ external-module-compat-comm.h       2008-11-12 15:52:05.000000000 +0100
> @@ -601,7 +601,7 @@
>
>  static inline void hrtimer_start_expires(struct hrtimer *timer, int mode)
>  {
> -       hrtimer_start(timer, timer->expires, mode);
> +       hrtimer_start_p(timer, timer->expires, mode);
>  }
>
>
>
> Is this a correct fix?
>
>   

It is correct -- I added hrtimer_start_expires without taking 2.6.16 
into account.

> - I could boot suse 11.1 beta4 with it(which didnt work with kvm-77)
>
> Im really not sure if this works corretly and I dont want have broken
> modules in my kernel :(
>
> Is hrtimer_start_p already initialized when  hrtimer_start_expires is
> called? (help)
>   

Yes, hrtimer_start_p is initialized very early on.

Please send a signed-off-by line and I will apply.


-- 
error compiling committee.c: too many arguments to function

Re: [ANNOUNCE] kvm-79 release

[Avi Kivity]

Martin Vogt wrote:
>
> Do I need git for this?
> Is this something informal, does it need some special diff format?
>
>   

Your patch is fine and I applied it.  Thanks.

-- 
error compiling committee.c: too many arguments to function