[CentOS] Patched KVM RPMs (solves 'network shutdown under heavy load')

[Timo Schoeler]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

thus Timo Schoeler spake:
> thus Farkas Levente spake:
>> On 03/22/2010 12:12 PM, Timo Schoeler wrote:
>>> -----BEGIN PGP SIGNED MESSAGE-----
>>> Hash: SHA1
>>>
>>> thus Farkas Levente spake:
>>>> On 03/22/2010 12:03 PM, Timo Schoeler wrote:
>>>>> Hi list,
>>>>>
>>>>> due to the necessity of having a patched KVM running, I build some new
>>>>> RPMs. However, only
>>>>>
>>>>> kmod-kvm-83-105.27.x86_64.rpm
>>>>>
>>>>> is important to replace the 'original' one as it's the only package that
>>>>> contains the patch. I didn't rename the packages, though, hopefully in
>>>>> order not to break the upgrade path provided to us.
>>>>>
>>>>> The patch adresses following problem:
>>>>>
>>>>> 'Fix a race condition where qemu finds that there are not enough virtio
>>>>> ring buffers available and the guest make more buffers available before
>>>>> qemu can enable notifications.' [0], the thread startet as 'network
>>>>> shutdown under heavy load' [1].
>>>>>
>>>>> KVM already contains the fix (since February 19th, IIRC), CentOS'
>>>>> packages doesn't.
>>>>>
>>>>> The RPMs can be found here:
>>>>>
>>>>> http://www.riscworks.net/CentOS/KVM/fix-virt-io-ring-buffer-race/
>>>> does it have an upstream (redhat) bugzilla entry?
>>>> it'd be better if they fix it...
>>> Sure, but we had the problem that it constantly crashed several
>>> customer's domUs, so we had to fix it ASAP. Just wanted to make it more
>>> convenient for people that have the same problem and can _not_ wait
>>> until it's fixed upstream.
>> then at least and src.rpm would be useful...
> 
> Okay, I'll see when I got time to create one...
> 
> Timo

Done, can be found in

http://www.riscworks.net/CentOS/KVM/fix-virt-io-ring-buffer-race/SRPM

now.

Timo
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)

iD8DBQFLqbqifg746kcGBOwRAvwiAJ9hqwVw7bQ3LtYOyu3uwoQMUsxzmACfS/fn
aEiJv9xg8HS7BDqcMloTrhs=
=2qZP
-----END PGP SIGNATURE-----