[bug report] ext4: Add support for blocksize < pagesize in dioread_nolock

[bug report] ext4: Add support for blocksize < pagesize in dioread_nolock

[Dan Carpenter]

Hello Ritesh Harjani,

The patch c8cc88163f40: "ext4: Add support for blocksize < pagesize
in dioread_nolock" from Oct 16, 2019, leads to the following static
checker warning:

fs/ext4/inode.c:2390 mpage_process_page() error: 'io_end_vec' dereferencing possible ERR_PTR()
fs/ext4/inode.c:2557 mpage_map_and_submit_extent() error: 'io_end_vec' dereferencing possible ERR_PTR()
fs/ext4/inode.c:3677 ext4_end_io_dio() error: 'io_end_vec' dereferencing possible ERR_PTR()

fs/ext4/inode.c
  2371          bh = head = page_buffers(page);
  2372          do {
  2373                  if (lblk < mpd->map.m_lblk)
  2374                          continue;
  2375                  if (lblk >= mpd->map.m_lblk + mpd->map.m_len) {
  2376                          /*
  2377                           * Buffer after end of mapped extent.
  2378                           * Find next buffer in the page to map.
  2379                           */
  2380                          mpd->map.m_len = 0;
  2381                          mpd->map.m_flags = 0;
  2382                          io_end_vec->size += io_end_size;
  2383                          io_end_size = 0;
  2384  
  2385                          err = mpage_process_page_bufs(mpd, head, bh, lblk);
  2386                          if (err > 0)
  2387                                  err = 0;
  2388                          if (!err && mpd->map.m_len && mpd->map.m_lblk > lblk) {
  2389                                  io_end_vec = ext4_alloc_io_end_vec(io_end);
                                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
This allocation can fail.

  2390                                  io_end_vec->offset = mpd->map.m_lblk << blkbits;
                                        ^^^^^^^^^^^^^^^^^^
Oops

  2391                          }
  2392                          *map_bh = true;
  2393                          goto out;
  2394                  }
  2395                  if (buffer_delay(bh)) {
  2396                          clear_buffer_delay(bh);
  2397                          bh->b_blocknr = pblock++;
  2398                  }
  2399                  clear_buffer_unwritten(bh);
  2400                  io_end_size += (1 << blkbits);
  2401          } while (lblk++, (bh = bh->b_this_page) != head);

regards,
dan carpenter

[PATCH 1/1] ext4: Add error handling for io_end_vec struct allocation

[Ritesh Harjani]

This patch adds the error handling in case of any memory allocation
failure for io_end_vec. This was missing in original
patch series which enables dioread_nolock for blocksize < pagesize.

Fixes: c8cc88163f40 ("ext4: Add support for blocksize < pagesize in dioread_nolock")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Ritesh Harjani <riteshh@linux.ibm.com>
---
 fs/ext4/inode.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 381813205f99..de70f19bfa7e 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -2240,6 +2240,10 @@ static int mpage_process_page(struct mpage_da_data *mpd, struct page *page,
 				err = 0;
 			if (!err && mpd->map.m_len && mpd->map.m_lblk > lblk) {
 				io_end_vec = ext4_alloc_io_end_vec(io_end);
+				if (IS_ERR(io_end_vec)) {
+					err = PTR_ERR(io_end_vec);
+					goto out;
+				}
 				io_end_vec->offset = mpd->map.m_lblk << blkbits;
 			}
 			*map_bh = true;
@@ -2405,8 +2409,11 @@ static int mpage_map_and_submit_extent(handle_t *handle,
 	loff_t disksize;
 	int progress = 0;
 	ext4_io_end_t *io_end = mpd->io_submit.io_end;
-	struct ext4_io_end_vec *io_end_vec = ext4_alloc_io_end_vec(io_end);
+	struct ext4_io_end_vec *io_end_vec;
 
+	io_end_vec = ext4_alloc_io_end_vec(io_end);
+	if (IS_ERR(io_end_vec))
+		return PTR_ERR(io_end_vec);
 	io_end_vec->offset = ((loff_t)map->m_lblk) << inode->i_blkbits;
 	do {
 		err = mpage_map_one_extent(handle, mpd);
-- 
2.21.0

Re: [bug report] ext4: Add support for blocksize < pagesize in dioread_nolock

[Ritesh Harjani]

Thanks Dan for reporting this.

On 11/6/19 1:55 PM, Dan Carpenter wrote:
> Hello Ritesh Harjani,
> 
> The patch c8cc88163f40: "ext4: Add support for blocksize < pagesize
> in dioread_nolock" from Oct 16, 2019, leads to the following static
> checker warning:
> 
> fs/ext4/inode.c:2390 mpage_process_page() error: 'io_end_vec' dereferencing possible ERR_PTR()
> fs/ext4/inode.c:2557 mpage_map_and_submit_extent() error: 'io_end_vec' dereferencing possible ERR_PTR()
> fs/ext4/inode.c:3677 ext4_end_io_dio() error: 'io_end_vec' dereferencing possible ERR_PTR()

ext4_end_io_dio func is removed on recent ext4 master branch.
It got removed in ext4 iomap DIO patches. So my patch
(which is based on today's ext4 master branch) does not covers
for ext4_end_io_dio().

-ritesh

Re: [PATCH 1/1] ext4: Add error handling for io_end_vec struct allocation

[Theodore Y. Ts'o]

On Wed, Nov 06, 2019 at 03:08:09PM +0530, Ritesh Harjani wrote:
> This patch adds the error handling in case of any memory allocation
> failure for io_end_vec. This was missing in original
> patch series which enables dioread_nolock for blocksize < pagesize.
> 
> Fixes: c8cc88163f40 ("ext4: Add support for blocksize < pagesize in dioread_nolock")
> Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
> Signed-off-by: Ritesh Harjani <riteshh@linux.ibm.com>

Applied, thanks.

					- Ted