Re: Memory corruption during hibernation since 2.6.31

[Balbir Singh <balbir@linux.vnet.ibm.com>]

* Hugh Dickins <hughd@google.com> [2010-07-29 21:32:35]:

> On Thu, Jul 29, 2010 at 9:18 PM, Balbir Singh <balbir@linux.vnet.ibm.com> wrote:
> >
> > Couldn't we reuse PF_* flags to differentiate between the paths, if
> > that is what it eventually boils down to? On an unrelated note, I was
> > looking at shrink_all_memory() and wondering if swappiness really
> > mattered there.
> 
> So far as the swap-reuse issue goes, I don't see that a PF_ flag
> would help: the threads that already worry about such issues do
> the set_freezable()/try_to_freeze() thing, and won't get into
> trouble anyway; we don't want to force every other thread to
> have to do something special now, better just check global state
> in the very few places its needed.
>

We already do that with PF_MEMALLOC in several places. If the goal is
to avoid resuing the swap entry at all times after hibernate, the
solution needs to be a global state solution like you suggest.
 
> On the unrelated note: better in an unrelated thread!
>

Sure :)
 
-- 
	Three Cheers,
	Balbir