On Wed, Jan 20, 2021 at 04:02:10PM +0000, Matthew Wilcox wrote:
> On Wed, Jan 20, 2021 at 05:05:10PM +0200, Mike Rapoport wrote:
> > On Tue, Jan 19, 2021 at 08:22:13PM +0000, Matthew Wilcox wrote:
> > > On Thu, Dec 03, 2020 at 08:29:44AM +0200, Mike Rapoport wrote:
> > > > +static vm_fault_t secretmem_fault(struct vm_fault *vmf)
> > > > +{
> > > > + struct address_space *mapping = vmf->vma->vm_file->f_mapping;
> > > > + struct inode *inode = file_inode(vmf->vma->vm_file);
> > > > + pgoff_t offset = vmf->pgoff;
> > > > + vm_fault_t ret = 0;
> > > > + unsigned long addr;
> > > > + struct page *page;
> > > > + int err;
> > > > +
> > > > + if (((loff_t)vmf->pgoff << PAGE_SHIFT) >= i_size_read(inode))
> > > > + return vmf_error(-EINVAL);
> > > > +
> > > > + page = find_get_page(mapping, offset);
> > > > + if (!page) {
> > > > +
> > > > + page = secretmem_alloc_page(vmf->gfp_mask);
> > > > + if (!page)
> > > > + return vmf_error(-ENOMEM);
> > >
> > > Just use VM_FAULT_OOM directly.
> >
> > Ok.
> >
> > > > + err = add_to_page_cache(page, mapping, offset,
> > > > vmf->gfp_mask);
> > > > + if (unlikely(err))
> > > > + goto err_put_page;
> > >
> > > What if the error is EEXIST because somebody else raced with you to add
> > > a new page to the page cache?
> >
> > Right, for -EEXIST I need a retry here, thanks.
> >
> > > > + err = set_direct_map_invalid_noflush(page, 1);
> > > > + if (err)
> > > > + goto err_del_page_cache;
> > >
> > > Does this work correctly if somebody else has a reference to the page
> > > in the meantime?
> >
> > Yes, it does. If somebody else won the race that page was dropped from the
> > direct map and this call would be essentially a nop. And anyway, the very
> > next patch changes the way pages are removed from the direct map ;-)
>
> What I'm thinking is:
>
> thread A page faults
> doesn't find page
> allocates page
> adds page to page cache
> thread B page faults
> does find page in page cache
> set direct map invalid fails
> deletes from page cache
> ... ?
Hmm, this is not nice indeed...
--
Sincerely yours,
Mike.
