Re: [PATCH -next 1/8] mm: memory_hotplug: add pfn_to_online_folio()

[Matthew Wilcox]

On Tue, Dec 13, 2022 at 08:13:31PM +0800, Kefeng Wang wrote:
> 
> On 2022/12/13 19:40, David Hildenbrand wrote:
> > On 13.12.22 10:27, Kefeng Wang wrote:
> > > Introduce a wrapper function pfn_to_online_folio(), which calls
> > > pfn_to_online_page() and returns the folio of the page found,
> > > or null if no page.
> > > 
> > > Signed-off-by: Kefeng Wang <wangkefeng.wang@xxxxxxxxxx>
> > > ---
> > >   include/linux/memory_hotplug.h | 7 +++++++
> > >   1 file changed, 7 insertions(+)
> > > 
> > > diff --git a/include/linux/memory_hotplug.h
> > > b/include/linux/memory_hotplug.h
> > > index 9fcbf5706595..e841e4fb52a7 100644
> > > --- a/include/linux/memory_hotplug.h
> > > +++ b/include/linux/memory_hotplug.h
> > > @@ -265,6 +265,13 @@ static inline void
> > > pgdat_kswapd_unlock(pg_data_t *pgdat) {}
> > >   static inline void pgdat_kswapd_lock_init(pg_data_t *pgdat) {}
> > >   #endif /* ! CONFIG_MEMORY_HOTPLUG */
> > >   +static inline struct folio *pfn_to_online_folio(unsigned long pfn)
> > > +{
> > > +    struct page *page = pfn_to_online_page(pfn);
> > > +
> > > +    return page ? page_folio(page) : NULL;
> > > +}
> > 
> > Who guarantees that page_folio() is safe and stable at that point?
> > 
> > IIRC, that's very tricky. We could have the page concurrently getting
> > freed and the folio dissolved.
> So the caller should consider this,  lock or get the folio in the caller?

The caller only has a pfn; it doesn't have the folio at the time of
the call.  David is right, this function cannot safely exist.