Alexey Dobriyan <[email protected]> writes:
> On Fri, Feb 13, 2009 at 12:45:03PM +0100, Ingo Molnar wrote:
>>
>> * Alexey Dobriyan <[email protected]> wrote:
>>
>> > On Fri, Feb 13, 2009 at 11:27:32AM +0100, Ingo Molnar wrote:
>> > > Merging checkpoints instead might give them the incentive to get
>> > > their act together.
>> >
>> > Knowing how much time it takes to beat CPT back into usable shape every time
>> > big kernel rebase is done, OpenVZ/Virtuozzo have every single damn incentive
>> > to have CPT mainlined.
>>
>> So where is the bottleneck? I suspect the effort in having forward ported
>> it across 4 major kernel releases in a single year is already larger than
>> the technical effort it would take to upstream it. Any unreasonable upstream
>> resistence/passivity you are bumping into?
>
> People were busy with netns/containers stuff and OpenVZ/Virtuozzo bugs.
Yes. Getting the namespaces particularly the network namespace finished
has consumed a lot of work.
Then we have a bunch of people helping with ill conceived patches that seem
to wear out the patience of people upstream. Al, Greg kh, Linus.
The whole recent ressurection of the question of we should have a clone
with pid syscall.
Eric
Quoting Eric W. Biederman ([email protected]):
> Alexey Dobriyan <[email protected]> writes:
>
> > On Fri, Feb 13, 2009 at 12:45:03PM +0100, Ingo Molnar wrote:
> >>
> >> * Alexey Dobriyan <[email protected]> wrote:
> >>
> >> > On Fri, Feb 13, 2009 at 11:27:32AM +0100, Ingo Molnar wrote:
> >> > > Merging checkpoints instead might give them the incentive to get
> >> > > their act together.
> >> >
> >> > Knowing how much time it takes to beat CPT back into usable shape every time
> >> > big kernel rebase is done, OpenVZ/Virtuozzo have every single damn incentive
> >> > to have CPT mainlined.
> >>
> >> So where is the bottleneck? I suspect the effort in having forward ported
> >> it across 4 major kernel releases in a single year is already larger than
> >> the technical effort it would take to upstream it. Any unreasonable upstream
> >> resistence/passivity you are bumping into?
> >
> > People were busy with netns/containers stuff and OpenVZ/Virtuozzo bugs.
>
> Yes. Getting the namespaces particularly the network namespace finished
> has consumed a lot of work.
>
> Then we have a bunch of people helping with ill conceived patches that seem
> to wear out the patience of people upstream. Al, Greg kh, Linus.
>
> The whole recent ressurection of the question of we should have a clone
> with pid syscall.
/me points
Alexey started it :)
But, Linus asks to start with simple checkpoint/restart patches. Oren's
basic patchset pretty much does that, though, right? Patches 1-7 just
do a basic single task. 8-10 add simple open files. 11, 13 and 14 do
external checkpoint and multiple tasks.
Are these an ok place to start, or do these need to be simplified even
more?
-serge