Re: [Xen-devel] Balloon driver bug in increase_reservation

[Wei Liu <wei.liu2@xxxxxxxxxx>]

On Wed, Sep 04, 2013 at 02:31:13PM +0100, Stefano Stabellini wrote:
[...]
> > > 
> > > - if the guest is autotranslate, then the BUG_ON is already irrelevant;
> > > - if the guest is not autotranslate, then we are sure that the p2m of
> > >   the page is pointing to a scratch page;
> > > 
> > > either way the BUG_ON is wrong.
> > 
> > Is it there to guard other misuse of the function? IOW can we be
> > confident that all calls are legit?
> 
> the ballooned_pages list would need to be broken for this BUG_ON to
> trigger
> 

Right.

> 
> > Messing up the P2M table looks very hard to debug?
> >
> > > Otherwise could simply implement a is_balloon_scratch_page function that
> > > checks whether a given pfn corresponds to any of the scratch pages (it
> > > doesn't need to be the scratch page of this cpu).
> > 
> > That's quite expensive IMHO, especially when you have lots of CPU's and
> > lots of ballooned pages.
> 
> as a compromise we could run a check on the pfn only on debug runs?

Actually I'm out of idea now so I'm happy to implement a solution that
everybody agrees...

Wei.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel