Re: [PATCH v4 14/36] iomap: Support large pages in iomap_adjust_read_range

[Matthew Wilcox]

On Fri, May 22, 2020 at 08:24:38AM +1000, Dave Chinner wrote:
> > @@ -571,7 +572,6 @@ static int
> >  __iomap_write_begin(struct inode *inode, loff_t pos, unsigned len, int flags,
> >  		struct page *page, struct iomap *srcmap)
> >  {
> > -	struct iomap_page *iop = iomap_page_create(inode, page);
> >  	loff_t block_size = i_blocksize(inode);
> >  	loff_t block_start = pos & ~(block_size - 1);
> >  	loff_t block_end = (pos + len + block_size - 1) & ~(block_size - 1);
> > @@ -580,9 +580,10 @@ __iomap_write_begin(struct inode *inode, loff_t pos, unsigned len, int flags,
> >  
> >  	if (PageUptodate(page))
> >  		return 0;
> > +	iomap_page_create(inode, page);
> 
> What problem does this fix? i.e. if we can get here with an
> uninitialised page, why isn't this a separate bug fix. I don't see
> anything in this patch that actually changes behaviour, and there's
> nothing in the commit description to tell me why this is here,
> so... ???

I'm not fixing anything ... just moving the call to iomap_page_create()
from the opening stanza to down here because we no longer need a struct
iomap_page pointer in this function.