Re: [PATCH v4 3/6] mm: migrate: split folio_migrate_mapping()

[Matthew Wilcox <willy@infradead.org>]

On Thu, Jun 06, 2024 at 09:24:37AM +0800, Kefeng Wang wrote:
> On 2024/6/6 8:54, Jane Chu wrote:
> > [..]
> > > -int folio_migrate_mapping(struct address_space *mapping,
> > > -		struct folio *newfolio, struct folio *folio, int extra_count)
> > > +static void __folio_migrate_mapping(struct address_space *mapping,
> > > +		struct folio *newfolio, struct folio *folio, int expected_cnt)
> > >   {
> > >   	XA_STATE(xas, &mapping->i_pages, folio_index(folio));
> > >   	struct zone *oldzone, *newzone;
> > > -	int dirty;
> > > -	int expected_count = folio_expected_refs(mapping, folio) + extra_count;
> > >   	long nr = folio_nr_pages(folio);
> > >   	long entries, i;
> > > +	int dirty;
> > >   	if (!mapping) {
> > > -		/* Anonymous page without mapping */
> > 
> > If 'mapping' was NULL, the first line would blow up while dereferencing
> > 'mapping->i_pages'.
> 
> 0->i_pages is wrong, but &(0->i_pages) is legal, and then xas->xa = NULL.

Uhh, it's not NULL, but it will be a small integer (offsetof(struct
address_space, i_pages)).