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

["HORIGUCHI NAOYA(堀口 直也)" <naoya.horiguchi@nec.com>]

On Thu, Sep 08, 2022 at 08:14:57AM +0200, David Hildenbrand wrote:
> On 08.09.22 04:47, HORIGUCHI NAOYA(堀口 直也) wrote:
> > On Wed, Sep 07, 2022 at 01:57:45PM -0700, Andrew Morton wrote:
> > > On Wed, 7 Sep 2022 21:11:57 +0900 Naoya Horiguchi <naoya.horiguchi@linux.dev> wrote:
> > > 
> > > > 
> > > > 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")
> > > 
> > > January of 2016!  Or was this a long-term bug which was recently
> > > exposed by some other change?
> > 
> > Maybe yes, I confirmed that this issue reproduces in kernel 5.10 (released
> > on Dec 13, 2020, which might not be so "recent"), but does not reproduce in
> > kernel 5.4.  So I think that marking "5.10+" for stable is fine.
> 
> 
> I fixed plenty of such issues in the past. For example:
> 
> aad5f69bc161 ("fs/proc/page.c: don't access uninitialized memmaps in fs/proc/page.c")
> 
> This one just wasn't found so far because it's triggered via
> a debug interface.
> 
> The correct fixes tag would be:
> 
> Fixes: f1dd2cd13c4b ("mm, memory_hotplug: do not associate hotadded memory to zones until online")      [visible after d0dc12e86b319]
> 
> Exposed "recently" -- 2018 ;)

Great, thank you for pinpointing the commit.

- Naoya Horiguchi