Re: Fwd: Re: Hibernation on i386 failure...

[Woody Suwalski]

Pavel Machek wrote:
> On Sat 2018-03-03 19:03:48, Woody Suwalski wrote:
> > Pavel Machek wrote:
> > > Hi!
> > >
> > > > > On Tue 2018-01-09 16:54:51, Woody Suwalski wrote:
> > > > > > Hey Pavel, so you have found that issue earlier... Do you have a patch to
> > > > > > try?
> > > > > Did I? I do not remember clearly.
> > > > >
> > > > > Anyway, here are my hibernation-related patches... Absolutely no
> > > > > guarantees.
> > > > Pavel, here is my attempt to convert your patch to 4.15 git, it builds
> > > > cleanly, and it does not work. If you have a moment to spare - could you
> > > > check if it is a classic case "everything is good, just does not
> > > > work"?
> > > Sorry for the delay.
> > >
> > > I've tested today's linux-next, and hibernation works for me, no
> > > patches needed.
> > >
> > > Best regards,
> > > 									Pavel
> > My January tests have showed that hibernation does work if I turn off the
> > CONFIG_RELOCATABLE option.
> > So can you confirm that you have it enabled and it works again? I will try
> > it on 4.16-rc4 tomorrow...
> I was actually testing with linux-next:
>
> commit 80253003d5da3b82d6bebfb755a04c24011e90e2
> Author: Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx>
> Date:   Fri Mar 2 14:17:11 2018 +1100
>
>      Add linux-next specific files for 20180302
>
> # git clone
>    git://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git
>    # git fetch linux-next
>    # git fetch --tags linux-next
>    git remote update
>    # then: git checkout next-20170201
>
> And yes, I do have CONFIG_RELOCATABLE on:
>
> CONFIG_RELOCATABLE=y
> # CONFIG_RANDOMIZE_BASE is not set
> CONFIG_X86_NEED_RELOCS=y
>
> Best regards,
> 									Pavel
Almost there. Yes I have noticed the linux-next, but I wanted to see 
first if the same does not work in the mainline - then it would have 
pointed to the difference between the branches.

However the important part is your "CONFIG_RANDOMIZE_BASE is not set". 
Try to turn it on and check hibernation. I have used the 
CONFIG_RELOCATABLE entry, because it turns off the randomized 
automatically, and on its own it I had no use for it. BTW. turning the 
relocatable field on by default also brings randomization on, so you 
must have specifically disable it. Why?

So the real issue is: can the 32-bit kernel use randomization, _and_ 
hibernate on linux-next???

Thanks, Woody