On Tue, Nov 10, 2020 at 10:41:14AM +0100, David Hildenbrand wrote:
> On 10.11.20 10:36, Alexander Gordeev wrote:
> >If for whatever reason the sub-PMD region to be used is less
> >than struct page size (e.g in the future), then it is possible
> >to overwrite beyond the region size.
> >
> >Signed-off-by: Alexander Gordeev <agordeev@xxxxxxxxxxxxx>
> >---
> > arch/s390/mm/vmem.c | 4 +++-
> > 1 file changed, 3 insertions(+), 1 deletion(-)
> >
> >diff --git a/arch/s390/mm/vmem.c b/arch/s390/mm/vmem.c
> >index 56ab9bb770f3a..d7f25884061f4 100644
> >--- a/arch/s390/mm/vmem.c
> >+++ b/arch/s390/mm/vmem.c
> >@@ -91,13 +91,15 @@ static void vmemmap_flush_unused_pmd(void)
> > static void __vmemmap_use_sub_pmd(unsigned long start, unsigned long end)
> > {
> >+ unsigned long size = min(end - start, sizeof(struct page));
> >+
> > /*
> > * As we expect to add in the same granularity as we remove, it's
> > * sufficient to mark only some piece used to block the memmap page from
> > * getting removed (just in case the memmap never gets initialized,
> > * e.g., because the memory block never gets onlined).
> > */
> >- memset(__va(start), 0, sizeof(struct page));
> >+ memset(__va(start), 0, size);
> > }
> > static void vmemmap_use_sub_pmd(unsigned long start, unsigned long end)
> >
>
> I don't really see a need for that. Can you spell out one possible
> configuration that would trigger that in the future? It's sounds
> very unlikely and I have the feeling there might be more to change
> at other points.
No configuration in mind. But dependency on struct page is the only
obstacle that prevents the whole thing to become generic (unless I
am missing something). Moreover, the memset() would not be needed
also - just a single non-PAGE_UNUSED word within [start..end) should
be enough.
> --
> Thanks,
>
> David / dhildenb
>
