Re: [RFC v2 2/2] gfs2: Rework read and page fault locking

[Matthew Wilcox]

On Fri, Jul 03, 2020 at 12:38:01PM +0100, Matthew Wilcox wrote:
> > @@ -598,16 +562,9 @@ static void gfs2_readahead(struct readahead_control *rac)
> >  {
> >  	struct inode *inode = rac->mapping->host;
> >  	struct gfs2_inode *ip = GFS2_I(inode);
> > -	struct gfs2_holder gh;
> >  
> > -	gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &gh);
> > -	if (gfs2_glock_nq(&gh))
> > -		goto out_uninit;
> >  	if (!gfs2_is_stuffed(ip))
> >  		mpage_readahead(rac, gfs2_block_map);
> 
> I think you probably want to make this:
> 
> 	if (i_blocksize(page->mapping->host) == PAGE_SIZE &&
> 	    !page_has_buffers(page))
> 		error = iomap_readahead(rac, &gfs2_iomap_ops);
> 	else if (!gfs2_is_stuffed(ip))
> 		error = mpage_readahead(rac, gfs2_block_map);
> 
> ... but I understand not wanting to make that change at this point
> in the release cycle.

That was stupid.  I meant to write out:

	if (i_blocksize(rac->mapping->host) == PAGE_SIZE)
		error = iomap_readahead(rac, &gfs2_iomap_ops);
	else if (!gfs2_is_stuffed(ip))
		error = mpage_readahead(rac, gfs2_block_map);

Since the pages are freshly allocated, they can't have buffers, and
the mapping comes out of the readahead_control, not from the page.