Re: [PATCH v3] x86/e820: Fix handling of subpage regions when calculating nosave ranges

[Myrrh Periwinkle]

On 4/7/25 07:53, Myrrh Periwinkle wrote:

> On 4/7/25 01:36, Ingo Molnar wrote:
>
> > * Ingo Molnar<mingo@xxxxxxxxxx> wrote:
> >
> > > * Myrrh Periwinkle<myrrhperiwinkle@xxxxxxxxxxx> wrote:
> > >
> > > > The current implementation of e820__register_nosave_regions suffers 
> > > > from
> > > > multiple serious issues:
> > > >   - The end of last region is tracked by PFN, causing it to find holes
> > > >     that aren't there if two consecutive subpage regions are present
> > > >   - The nosave PFN ranges derived from holes are rounded out 
> > > > (instead of
> > > >     rounded in) which makes it inconsistent with how explicitly 
> > > > reserved
> > > >     regions are handled
> > > >
> > > > Fix this by:
> > > >   - Treating reserved regions as if they were holes, to ensure 
> > > > consistent
> > > >     handling (rounding out nosave PFN ranges is more correct as the
> > > >     kernel does not use partial pages)
> > > >   - Tracking the end of the last RAM region by address instead of 
> > > > pages
> > > >     to detect holes more precisely
> > > >
> > > > Cc:stable@xxxxxxxxxxxxxxx
> > > > Fixes: e5540f875404 ("x86/boot/e820: Consolidate 'struct e820_entry 
> > > > *entry' local variable names")
> > > So why is this SHA1 indicated as the root cause? AFAICS that commit
> > > does nothing but cleanups, so it cannot cause such regressions.
> > BTW.:
> >
> >   A) "It was the first random commit that seemed related, sry"
> >   B) "It's a 15 years old bug, but I wanted to indicate a fresh, 
> > 8-year old bug to get this into -stable. Busted!"
>
> You got me :) How did you know that this is a 15 years old bug? 
> (although I didn't think the age of the bug a patch fixes would affect 
> its chances of getting to -stable)
>
> This specific revision was picked since it's the latest one that this 
> patch can be straightforwardly applied to (there is a (trivial) merge 
> conflict with -stable, though).
>
> Later, I managed to track the buggy logic back to 1c10070a55a3 ("i386: 
> do not restore reserved memory after hibernation"), which I believe is 
> the very first occurrence of this bug. If you prefer, I can send a v4 
> with a more correct Fixes: tag (or feel free to do so yourself when 
> applying this patch).

I did some more digging and it seems like the buggy logic actually 
appeared all the way back in e8eff5ac294e ("[PATCH] Make swsusp avoid 
memory holes and reserved memory regions on x86_64") back when x86_64 
was a separate port, which was copied later by the i386 port in the 
commit I mentioned above, which would make this a 19 year old bug 
instead of 15.

> > ... are perfectly fine answers in my book. :-)
> >
> > I'm glad about the fixes, I'm just curious how the Fixes tag came about.
> >
> > Thanks,
> >
> >     Ingo
>
> Regards,
>
> Myrrh