Re: [PATCH v3 5/6] netfs, 9p, afs, ceph: Use folios

[Matthew Wilcox <willy@infradead.org>]

On Wed, Nov 03, 2021 at 02:58:12PM +0000, David Howells wrote:
> Matthew Wilcox <willy@infradead.org> wrote:
> 
> > > +	len = (size >= start + gran) ? gran : size - start;
> > 
> > This seems like the most complicated way to write this ... how about:
> > 
> >         size_t len = min_t(loff_t, isize - start, folio_size(folio));
> 
> I was trying to hedge against isize-start going negative.  Can this code race
> against truncate?  truncate_setsize() changes i_size *before* invalidating the
> pages.

We should check for isize < start separately, and skip the writeback
entirely.

> > >  static int afs_symlink_readpage(struct file *file, struct page *page)
> > >  {
> > > -	struct afs_vnode *vnode = AFS_FS_I(page->mapping->host);
> > > +	struct afs_vnode *vnode = AFS_FS_I(page_mapping(page)->host);
> > 
> > How does swap end up calling readpage on a symlink?
> 
> Um - readpage is called to read the symlink.

But the only reason to use page_mapping() instead of page->mapping
is if you don't know that the page is in the page cache.  You know
that here, so I don't understand why you changed it.

> > > -	page_endio(page, false, ret);
> > > +	page_endio(&folio->page, false, ret);
> > 
> > We need a folio_endio() ...
> 
> I think we mentioned this before and I think you said you had or would make a
> patch for it.  I can just create a wrapper for it if that'll do.

Probably better to convert it and put a page_endio wrapper in
folio-compat.c