btrfs: calculate number of eb pages properly in csum_tree_block
authorDavid Sterba <dsterba@suse.com>
Wed, 28 Jul 2021 16:00:24 +0000 (18:00 +0200)
committerDavid Sterba <dsterba@suse.com>
Thu, 29 Jul 2021 11:01:04 +0000 (13:01 +0200)
commit7280305eb57dd32735f795ed4ee679bf9854f9d0
treef3b169d1e5110807a30e7b1c4ce39b5db4a45c5c
parentb2a616676839e2a6b02c8e40be7f886f882ed194
btrfs: calculate number of eb pages properly in csum_tree_block

Building with -Warray-bounds on systems with 64K pages there's a
warning:

  fs/btrfs/disk-io.c: In function ‘csum_tree_block’:
  fs/btrfs/disk-io.c:226:34: warning: array subscript 1 is above array bounds of ‘struct page *[1]’ [-Warray-bounds]
    226 |   kaddr = page_address(buf->pages[i]);
        |                        ~~~~~~~~~~^~~
  ./include/linux/mm.h:1630:48: note: in definition of macro ‘page_address’
   1630 | #define page_address(page) lowmem_page_address(page)
        |                                                ^~~~
  In file included from fs/btrfs/ctree.h:32,
                   from fs/btrfs/disk-io.c:23:
  fs/btrfs/extent_io.h:98:15: note: while referencing ‘pages’
     98 |  struct page *pages[1];
        |               ^~~~~

The compiler has no way to know that in that case the nodesize is exactly
PAGE_SIZE, so the resulting number of pages will be correct (1).

Let's use num_extent_pages that makes the case nodesize == PAGE_SIZE
explicitly 1.

Reported-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/disk-io.c