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

[Pavel Machek <pavel@ucw.cz>]

Hi!

> > > Why? Becuase there is no _room_ for inconsistency. There's nothing to be 
> > > "inconsistent with", since any changes to memory (by things like DMA or 
> > > other setup that happens while the suspend process is going on) is by 
> > > _definition_ consistent with the resume image (becasue there is no 
> > > separate image).
> > 
> > Do you propose to keep DMAs running while suspending-to-RAM?
> 
> What part of "suspend a chip" do you have trouble with?
> 
> DMA obviously does *not* happen with a suspended device. There's no need 
> to turn DMA even off - it just doesn't happen!

Ok, I guess I'll have nightmares of DMA controllers doing DMAs from
chips that are no longer there tonight.

> > > For example, the whole myth that "freeze" needs to shut off DMA is a total 
> > > and utter *myth*. It needs nothing of the sort at all. Rather than shut 
> > > off DMA and try to make the hardware be wevy wevy quiet while it's hunting 
> > > wabbits, it's a lot easier to just do nothing at all on "freeze",
> > 
> > No. Sorry, you are wrong here. 
> > 
> > Remember that during resume we run
> > 
> > freeze()
> > copy old data into memory
> > thaw()
> > 
> > Now, if the old kernel left DMAs running, it could be overwriting
> > the data we are copying in.
> 
> The *thaw* needs to happen with devices quiescent. 
> 
> But that sure doesn't have anythign to do with the "snapshot()" path. In 
> fact, you'll have rebooted the machine in between.

Only the fact that we are currently using same device call during
snapshot() and during restore(). We obviously could do _5_ device
calls

(suspend/resume/freeze/quiesce_disable_dma/thaw)

...but that looks like too many calls to me.

> So what does that have to do with "snapshotting"?

I'm not comfortable with memory I'm copying changing under my hands
because of some DMA. It just looks like asking for trouble. I _think_
we can get away with DMA running during snapshot, because driver may
not assume anything about the DMA result before it got completion
interrupt, but... 

									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/