[bug report] btrfs: introduce read_extent_buffer_subpage()

[Dan Carpenter <dan.carpenter@oracle.com>]

Hello Qu Wenruo,

The patch 5c60a522f1ea: "btrfs: introduce
read_extent_buffer_subpage()" from Jan 16, 2021, leads to the
following static checker warning:

	fs/btrfs/extent_io.c:5797 read_extent_buffer_subpage()
	info: return a literal instead of 'ret'

fs/btrfs/extent_io.c
  5780  static int read_extent_buffer_subpage(struct extent_buffer *eb, int wait,
  5781                                        int mirror_num)
  5782  {
  5783          struct btrfs_fs_info *fs_info = eb->fs_info;
  5784          struct extent_io_tree *io_tree;
  5785          struct page *page = eb->pages[0];
  5786          struct bio *bio = NULL;
  5787          int ret = 0;
  5788  
  5789          ASSERT(!test_bit(EXTENT_BUFFER_UNMAPPED, &eb->bflags));
  5790          ASSERT(PagePrivate(page));
  5791          io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree;
  5792  
  5793          if (wait == WAIT_NONE) {
  5794                  ret = try_lock_extent(io_tree, eb->start,
  5795                                        eb->start + eb->len - 1);
  5796                  if (ret <= 0)
  5797                          return ret;

If try_lock_extent() fails to get the lock and returns 0, then is
returning zero here really the correct behavior?  It feels like there
should be some documentation because this behavior is unexpected.

  5798          } else {
  5799                  ret = lock_extent(io_tree, eb->start, eb->start + eb->len - 1);
  5800                  if (ret < 0)
  5801                          return ret;
  5802          }
  5803  
  5804          ret = 0;
  5805          if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags) ||
  5806              PageUptodate(page) ||
  5807              btrfs_subpage_test_uptodate(fs_info, page, eb->start, eb->len)) {
  5808                  set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
  5809                  unlock_extent(io_tree, eb->start, eb->start + eb->len - 1);
  5810                  return ret;
  5811          }

regards,
dan carpenter