[ 3.8.y.z extended stable ] Patch "mm: avoid reinserting isolated balloon pages into LRU lists" has been added to staging queue

[Rafael Aquini]

On Thu, Oct 10, 2013 at 08:36:34AM -0400, Luiz Capitulino wrote:
> On Wed,  9 Oct 2013 13:51:23 -0700
> Kamal Mostafa <kamal at canonical.com> wrote:
> 
> > This is a note to let you know that I have just added a patch titled
> > 
> >     mm: avoid reinserting isolated balloon pages into LRU lists
> > 
> > to the linux-3.8.y-queue branch of the 3.8.y.z extended stable tree 
> > which can be found at:
> > 
> >  http://kernel.ubuntu.com/git?p=ubuntu/linux.git;a=shortlog;h=refs/heads/linux-3.8.y-queue
> > 
> > This patch is scheduled to be released in version 3.8.13.11.
> > 
> > If you, or anyone else, feels it should not be added to this tree, please 
> > reply to this email.
> 
> I guess it should be OK to add it, but I'm not sure you can trigger
> the bug in question in that old kernel. Only latest code made it possible
> to trigger the bug, and even then it's a corner case.
> 
> Rafael, what do you think?

I think we need it even though. Although we won't see the panic, as it only happens after merging 
commit e2be15f6c3eecedfbe1550cca8d72c5057abbbd2, the bug is still there, unfortunately.

For kernels pre-e2be15f this balloon page 'leak' to LRU lists would go unnoticed, 
as the 'leaked' pages would eventually be reclaimed when reaching the bottom of
LRU inactive lists. OTOH, as far as the balloon driver page counters are concerned,
there will be pages yet isolated, therefore the driver could get stuck at unload time,
while deflating the balloon, as it would keep waiting for a chance to free these pages 
that had already been freed elsewhere.

Btw, Luiz, perhaps we could think of a way on making balloon pages LRU eligible for the sake of 
your autoballooning project, but this is subject for another thread :)

Cheers!
-- Rafael