On Mon, 2017-03-20 at 10:47 +0800, Huang, Ying wrote:
> Hi, Tim,
>
> Tim Chen <tim.c.chen@xxxxxxxxxxxxxxx> writes:
>
> >
> > On Tue, Mar 14, 2017 at 05:25:29PM +0800, Huang, Ying wrote:
> > >
> > > +struct page *do_swap_page_readahead(struct vm_fault *vmf,
> > > + struct vma_swap_readahead *swap_ra,
> > > + swp_entry_t fentry,
> > > + struct page *fpage)
> > > +{
> > > + struct blk_plug plug;
> > > + struct vm_area_struct *vma = vmf->vma;
> > > + struct page *page;
> > > + unsigned long addr;
> > > + pte_t *pte, pentry;
> > > + gfp_t gfp_mask;
> > > + swp_entry_t entry;
> > > + int i, alloc = 0, count;
> > > + bool page_allocated;
> > > +
> > > + addr = vmf->address & PAGE_MASK;
> > > + blk_start_plug(&plug);
> > > + if (!fpage) {
> > > + fpage = __read_swap_cache_async(fentry, GFP_HIGHUSER_MOVABLE,
> > > + vma, addr, &page_allocated);
> > > + if (!fpage) {
> > > + blk_finish_plug(&plug);
> > > + return NULL;
> > > + }
> > > + if (page_allocated) {
> > > + alloc++;
> > > + swap_readpage(fpage);
> > > + }
> > Do you need to add here a put_page as there's a get_page
> > in __read-swap_cache_async?
> I don't call put_page() here because the page will be mapped to process
> page table.
>
> >
> > put_page(fpage);
> >
> > I think there is no put_page on the returned page when you return from
> > do_swap_page_readahead.
> In the original swapin_readahead(), the read_swap_cache_async() will be
> called for the fault swap entry again in the end of the function, and
> pug_page() is not called there.
>
I missed the second call to read_swap_cache_async in swapin_readahead.
You're right that we should keep the reference on the faulted page and not call
put_page on fpage here.
Thanks.
Tim
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
