Re: [PATCH v2] mm/huge_memory: use pfn_to_online_page() in split_huge_pages_all()

[Kirill A. Shutemov]

On Thu, Sep 08, 2022 at 01:11:50PM +0900, Naoya Horiguchi wrote:
> From: Naoya Horiguchi <naoya.horiguchi@xxxxxxx>
> 
> NULL pointer dereference is triggered when calling thp split via debugfs
> on the system with offlined memory blocks.  With debug option enabled,
> the following kernel messages are printed out:
> 
>   page:00000000467f4890 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x121c000
>   flags: 0x17fffc00000000(node=0|zone=2|lastcpupid=0x1ffff)
>   raw: 0017fffc00000000 0000000000000000 dead000000000122 0000000000000000
>   raw: 0000000000000000 0000000000000000 00000001ffffffff 0000000000000000
>   page dumped because: unmovable page
>   page:000000007d7ab72e is uninitialized and poisoned
>   page dumped because: VM_BUG_ON_PAGE(PagePoisoned(p))
>   ------------[ cut here ]------------
>   kernel BUG at include/linux/mm.h:1248!
>   invalid opcode: 0000 [#1] PREEMPT SMP PTI
>   CPU: 16 PID: 20964 Comm: bash Tainted: G          I        6.0.0-rc3-foll-numa+ #41
>   ...
>   RIP: 0010:split_huge_pages_write+0xcf4/0xe30
> 
> This shows that page_to_nid() in page_zone() is unexpectedly called for an
> offlined memmap.
> 
> Use pfn_to_online_page() to get struct page in PFN walker.
> 
> Fixes: 49071d436b51 ("thp: add debugfs handle to split all huge pages")
> Signed-off-by: Naoya Horiguchi <naoya.horiguchi@xxxxxxx>
> Co-developed-by: David Hildenbrand <david@xxxxxxxxxx>
> Signed-off-by: David Hildenbrand <david@xxxxxxxxxx>
> Reviewed-by: Yang Shi <shy828301@xxxxxxxxx>
> Acked-by: Michal Hocko <mhocko@xxxxxxxx>
> Reviewed-by: Miaohe Lin <linmiaohe@xxxxxxxxxx>
> Reviewed-by: Oscar Salvador <osalvador@xxxxxxx>
> Cc: <stable@xxxxxxxxxxxxxxx> # 5.10+

Looks good:

Acked-by: Kirill A. Shutemov <kirill.shutemov@xxxxxxxxxxxxxxx>

But it makes me think if there's other similar cases. "page is offline" is
rather obscure case that rarely covered by routine testing. Otherwise the
bug would not survive for 6 years.

After quick look, kvm_pfn_to_refcounted_page() looks suspicious.
kdb_getphys() too.

Maybe we should make pfn_valid() false for offline pages and introduce
other check that allows offline pages which can be used in codepaths that
deal with offline pages explicitly.

-- 
  Kiryl Shutsemau / Kirill A. Shutemov