On Thu, Jan 30, 2020 at 12:09:39AM -0800, Matthew Wilcox wrote:
> On Wed, Jan 29, 2020 at 11:46:09AM +1100, Dave Chinner wrote:
> > On Fri, Jan 24, 2020 at 05:35:47PM -0800, Matthew Wilcox wrote:
> > > From: "Matthew Wilcox (Oracle)" <willy@xxxxxxxxxxxxx>
> > >
> > > Use the new readahead operation in btrfs
> > >
> > > Signed-off-by: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx>
> > > Cc: linux-btrfs@xxxxxxxxxxxxxxx
> > > ---
> > > fs/btrfs/extent_io.c | 15 ++++-----------
> > > fs/btrfs/extent_io.h | 2 +-
> > > fs/btrfs/inode.c | 18 +++++++++---------
> > > 3 files changed, 14 insertions(+), 21 deletions(-)
> > >
> > > diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
> > > index 2f4802f405a2..b1e2acbec165 100644
> > > --- a/fs/btrfs/extent_io.c
> > > +++ b/fs/btrfs/extent_io.c
> > > @@ -4283,7 +4283,7 @@ int extent_writepages(struct address_space *mapping,
> > > return ret;
> > > }
> > >
> > > -int extent_readpages(struct address_space *mapping, struct list_head *pages,
> > > +unsigned extent_readahead(struct address_space *mapping, pgoff_t start,
> > > unsigned nr_pages)
> > > {
> > > struct bio *bio = NULL;
> > > @@ -4294,20 +4294,13 @@ int extent_readpages(struct address_space *mapping, struct list_head *pages,
> > > int nr = 0;
> > > u64 prev_em_start = (u64)-1;
> > >
> > > - while (!list_empty(pages)) {
> > > + while (nr_pages) {
> > > u64 contig_end = 0;
> > >
> > > - for (nr = 0; nr < ARRAY_SIZE(pagepool) && !list_empty(pages);) {
> > > - struct page *page = lru_to_page(pages);
> > > + for (nr = 0; nr < ARRAY_SIZE(pagepool) && nr_pages--;) {
> >
> > What is stopping nr_pages from going negative here, and then looping
> > forever on the outer nr_pages loop? Perhaps "while(nr_pages > 0) {"
> > would be better there?
>
> Ugh, nr_pages is unsigned, so that's no good. Maybe make this a more
> conventional loop ...
>
> while (nr_pages) {
> u64 contig_end = 0;
>
> for (nr = 0; nr < ARRAY_SIZE(pagepool); nr++) {
> struct page *page = readahead_page(mapping, start++);
>
> prefetchw(&page->flags);
> pagepool[nr] = page;
> contig_end = page_offset(page) + PAGE_SIZE - 1;
> if (--nr_pages == 0)
> break;
> }
Looks like it solves the problem :)
Cheers,
Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
