Re: Please apply d2155fe54ddb to 5.10 and 5.4

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Tue, Apr 22, 2025 at 05:58:26PM +0800, Qingfang Deng wrote:
> Hi David,
> 
> 
> On Tue, Apr 22, 2025 at 3:04 PM David Hildenbrand <david@redhat.com> wrote:
> > > Let me post it again:
> > >
> > > Please consider applying d2155fe54ddb ("mm: compaction: remove
> > > duplicate !list_empty(&sublist) check") to 5.10 and 5.4, as it
> > > resolves a -Wdangling-pointer warning in recent GCC versions:
> > >
> > > In function '__list_cut_position',
> > >      inlined from 'list_cut_position' at ./include/linux/list.h:400:3,
> > >      inlined from 'move_freelist_tail' at mm/compaction.c:1241:3:
> > > ./include/linux/list.h:370:21: warning: storing the address of local
> > > variable 'sublist' in '*&freepage_6(D)->D.15621.D.15566.lru.next'
> > > [-Wdangling-pointer=]
> >
> > The commit looks harmless. But I don't see how it could fix any warning?
> >
> > I mean, we replace two !list_empty() checks by a single one ... and the
> > warning is about list_cut_position() ?
> 
> I have no idea, actually. Maybe the double !list_empty() confuses the
> compiler, making it think `sublist` can be referenced out of the
> scope?

That is odd, are you sure this isn't a compiler bug?