[Devel] Re: [ANNOUNCE] first stable release of OpenVZ kernel virtualization solution

Kirill Korotaev dev at sw.ru
Tue Dec 6 03:59:25 PST 2005


Ingo Molnar wrote:
> * Kirill Korotaev <dev at sw.ru> wrote:
> 
> 
>>Yes, we have a patch queue, though it is not available to outside yet :(
>>You can find patch-022stab0XX-core in SRC RPM, which includes the 
>>following parts without driver updates (which are included in -combined 
>>patch):
>>- mainstream fixes
>>- mainstream security fixes
>>- 4GB split (yours :) , patched by me)
>>- User beancounters (kernel/ub/*, include/ub/*). This includes 
>>accounting and limiting of VPSs.
>>- Virtualization itself (ve_struct, kernel/vecalls.c - main code for VPS 
>>start/stop, net/ipv4 - virtualization of TPC/IP and netfilters, 
>>drivers/net/venet* - virtual network device for VPS, virtual pids, etc.)
>>- fs/simfs - simple filesystem to fake VPS and return correct values on 
>>`df` and statfs() output.
>>- fs/vzdq* - 2-level disk quota.
>>- kernel/fairsched.c and kernel/sched.c - fair CPU scheduler.
>>
>>If you wish I can prepare these 8 patches for you a bit later.
> 
> 
> well ... in general for LKML review it's easier to have split up 
> patches.
ok. we'll prepare these 8 patches. Not a problem.

>>Actually I think we'll start doing our developement in git after some 
>>time.
> 
> the -rt tree has a similar size:
> 
>  799 files changed, 28782 insertions(+), 9714 deletions(-)
> 
> and GIT isnt the best way for me, it's actually having a quilt 
> repository of 110+ patches that works best for me. That way i can keep 
> pushing stuff upstream, and have the queue ordered by 'likelhood of 
> upstream merging' (putting the least likely items last). Quilt is also 
> extremely fast. (faster than GIT doing equivalent stuff)
do you sync your tree and resolve the conflicts for each new kernel version?

> GIT is best if you are an upstream maintainer and want to sync stuff to 
> Linus periodically. But it's not the best for separate trees.
We don't use quilt, but the developement way is the same actually:
we have a patch list with patches groupped by subsytem (mainstream, 
virtualization, resource management), some of them go upstream, some 
not, some are merged to avoid a lot of conflicts later.

Kirill




More information about the Devel mailing list