Re: suspend2 merge (was Re: [Suspend2-devel] Re: CFS and suspend2:hang in atomic copy)

From: Nigel Cunningham
Date: Thu Apr 26 2007 - 05:45:38 EST


On Thu, 2007-04-26 at 00:27 -0700, David Lang wrote:
> On Thu, 26 Apr 2007, Nigel Cunningham wrote:
> > Hi.
> >
> > On Thu, 2007-04-26 at 01:33 +0200, Olivier Galibert wrote:
> >> On Wed, Apr 25, 2007 at 11:50:45AM -0700, Linus Torvalds wrote:
> >>> .. but if the alternative is a feature that just isn't worth it, and
> >>> likely to not only have its own bugs, but cause bugs elsewhere? (And yes,
> >>> I believe STD is both of those. There's a reason it's called "STD". Go
> >>> to google and type "STD" and press "I'm feeling lucky". Google is God).
> >>
> >> If it was correctly designed, it would be possible to change the
> >> hardware or even the kernel through a STD cycle. And that would be
> >> damn interesting on servers.
> >
> > Those are different issues - hardware hot/cold plugging for the first.
> >
> > Changing the kernel through a cycle - that's not a design fault. The
> > problem there is that the kernel and it's associated data structures are
> > part of the state. Changing the kernel and keeping the image would
> > require exactly correspondence in data structures, memory map and so on.
> > That's why the same kernel is required.
> that depends on exactly what you save in your snapshot.
> one approach is to try and save absolutly everything in ram (this is the current
> approach)
> if you do this then you do need to use the same kernel for the reasons that you
> list.
> however, you could also decide to only save the information about processes on
> the system (i.e. what you absolutly have to) and let the kernel re-initialize
> itself (along with it's devices) then you could use a different kernel safely.
> doing this should also save you a significant amount of storage when makeing
> your snapshot

Well, there is cryopid for individual processes. I suppose you could
potentially try doing a mass cryopiding. That would make things a lot
more complicated though. I'm not saying it's not doable.



Attachment: signature.asc
Description: This is a digitally signed message part