Re: [PATCH 4.14 023/159] mm/sparsemem: Allocate mem_section at runtime for CONFIG_SPARSEMEM_EXTREME=y

[Greg Kroah-Hartman]

On Mon, Jan 08, 2018 at 09:47:23AM +0100, Michal Hocko wrote:
> On Mon 08-01-18 08:53:08, Greg KH wrote:
> > On Sun, Jan 07, 2018 at 02:23:09PM +0100, Michal Hocko wrote:
> > > On Sun 07-01-18 13:44:02, Mike Galbraith wrote:
> > > > On Sun, 2018-01-07 at 11:18 +0100, Michal Hocko wrote:
> > > > > On Sun 07-01-18 10:11:15, Greg KH wrote:
> > > > > > On Sun, Jan 07, 2018 at 06:14:22AM +0100, Mike Galbraith wrote:
> > > > > > > On Fri, 2017-12-22 at 09:45 +0100, Greg Kroah-Hartman wrote:
> > > > > > > > 4.14-stable review patch.  If anyone has any objections, please let me know.
> > > > > > > 
> > > > > > > FYI, this broke kdump, or rather the makedumpfile part thereof.
> > > > > > >  Forward looking wreckage is par for the kdump course, but...
> > > > > > 
> > > > > > Is it also broken in Linus's tree with this patch?  Or is there an
> > > > > > add-on patch that I should apply to 4.14 to resolve this issue there?
> > > > > 
> > > > > This one http://lkml.kernel.org/r/1513932498-20350-1-git-send-email-bhe@xxxxxxxxxx
> > > > > I guess.
> > > > 
> > > > That won't unbreak kdump, else master wouldn't be broken.  I don't care
> > > > deeply, or know if anyone else does, I'm just reporting it because I
> > > > met it and chased it down.
> > > 
> > > OK, I didn't notice that d8cfbbfa0f7 ("mm/sparse.c: wrong allocation
> > > for mem_section") made it in after rc6. I am still wondering why
> > > 83e3c48729 ("mm/sparsemem: Allocate mem_section at runtime for
> > > CONFIG_SPARSEMEM_EXTREME=y") made it into the stable tree in the first
> > > place.
> > 
> > It was part of the prep for the KTPI code from what I can tell.
> 
> I do not see a direct relation, to be honest. It is more related to
> 5-level page tables but I might be missing some subtle relation.
> 
> > If you
> > think it should be reverted, just let me know and I'll be glad to do so.
> 
> This seems to be affecting Linus tree as well so it needs to get
> resolved. I would suggest reverting in stable for the mean time.
> If you really need it in the stable tree then you can pull it back later
> with all the follow up fixes.

Ok, I've now reverted it, thanks.

greg k-h