Re: [PATCH v1] drivers/base/memory.c: Don't access uninitialized memmaps in soft_offline_page_store()

[David Hildenbrand]

On 11.10.19 08:13, Naoya Horiguchi wrote:
> On Thu, Oct 10, 2019 at 04:12:00PM +0200, David Hildenbrand wrote:
> > Uninitialized memmaps contain garbage and in the worst case trigger kernel
> > BUGs, especially with CONFIG_PAGE_POISONING. They should not get
> > touched.
> >
> > Right now, when trying to soft-offline a PFN that resides on a memory
> > block that was never onlined, one gets a misleading error with
> > CONFIG_PAGE_POISONING:
> >    :/# echo 5637144576 > /sys/devices/system/memory/soft_offline_page
> >    [   23.097167] soft offline: 0x150000 page already poisoned
> >
> > But the actual result depends on the garbage in the memmap.
> >
> > soft_offline_page() can only work with online pages, it returns -EIO in
> > case of ZONE_DEVICE. Make sure to only forward pages that are online
> > (iow, managed by the buddy) and, therefore, have an initialized memmap.
> >
> > Add a check against pfn_to_online_page() and similarly return -EIO.
> >
> > Fixes: f1dd2cd13c4b ("mm, memory_hotplug: do not associate hotadded memory to zones until online") # visible after d0dc12e86b319
> > Cc: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> > Cc: "Rafael J. Wysocki" <rafael@xxxxxxxxxx>
> > Cc: Michal Hocko <mhocko@xxxxxxxxxx>
> > Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> > Signed-off-by: David Hildenbrand <david@xxxxxxxxxx>
> > ---
> >   drivers/base/memory.c | 3 +++
> >   1 file changed, 3 insertions(+)
> >
> > diff --git a/drivers/base/memory.c b/drivers/base/memory.c
> > index 6bea4f3f8040..55907c27075b 100644
> > --- a/drivers/base/memory.c
> > +++ b/drivers/base/memory.c
> > @@ -540,6 +540,9 @@ static ssize_t soft_offline_page_store(struct device *dev,
> >   	pfn >>= PAGE_SHIFT;
> >   	if (!pfn_valid(pfn))
> >   		return -ENXIO;
> > +	/* Only online pages can be soft-offlined (esp., not ZONE_DEVICE). */
> > +	if (!pfn_to_online_page(pfn))
> > +		return -EIO;
>
> Acked-by: Naoya Horiguchi <n-horiguchi@xxxxxxxxxxxxx>
>
> I think this check could be placed in soft_offline_page(), but that requires
> a few more unrelated lines of changes due to the mismatch on type of parameter
> between memory_failure() and soft_offline_page(),  This is not your problem,
> and I plan to do some cleanup on related interfaces, so this patch is fine.

Thanks,

well I think when you come via madvise(), you are always guaranteed to  
hold a reasonable page in your hands. Only when converting from  
arbitrary pfns, we have to watch out. But yeah, feel free to cc me on  
cleanups :)

> - Naoya Horiguchi


--

Thanks,

David / dhildenb