mm/filemap: don't allow partially uptodate page for pipes
authorEryu Guan <guaneryu@gmail.com>
Thu, 10 Nov 2016 18:46:35 +0000 (10:46 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 11 Nov 2016 16:12:37 +0000 (08:12 -0800)
Starting from 4.9-rc1 kernel, I started noticing some test failures of
sendfile(2) and splice(2) (sendfile0N and splice01 from LTP) when
testing on sub-page block size filesystems (tested both XFS and ext4),
these syscalls start to return EIO in the tests.  e.g.

  sendfile02    1  TFAIL  :  sendfile02.c:133: sendfile(2) failed to return expected value, expected: 26, got: -1
  sendfile02    2  TFAIL  :  sendfile02.c:133: sendfile(2) failed to return expected value, expected: 24, got: -1
  sendfile02    3  TFAIL  :  sendfile02.c:133: sendfile(2) failed to return expected value, expected: 22, got: -1
  sendfile02    4  TFAIL  :  sendfile02.c:133: sendfile(2) failed to return expected value, expected: 20, got: -1

This is because that in sub-page block size cases, we don't need the
whole page to be uptodate, only the part we care about is uptodate is OK
(if fs has ->is_partially_uptodate defined).

But page_cache_pipe_buf_confirm() doesn't have the ability to check the
partially-uptodate case, it needs the whole page to be uptodate.  So it
returns EIO in this case.

This is a regression introduced by commit 82c156f85384 ("switch
generic_file_splice_read() to use of ->read_iter()").  Prior to the
change, generic_file_splice_read() doesn't allow partially-uptodate page
either, so it worked fine.

Fix it by skipping the partially-uptodate check if we're working on a
pipe in do_generic_file_read(), so we read the whole page from disk as
long as the page is not uptodate.

I think the other way to fix it is to add the ability to check & allow
partially-uptodate page to page_cache_pipe_buf_confirm(), but that is
much harder to do and seems gain little.

Link: http://lkml.kernel.org/r/1477986187-12717-1-git-send-email-guaneryu@gmail.com
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/filemap.c

index c7fe2f1..50b52fe 100644 (file)
@@ -1732,6 +1732,9 @@ find_page:
                        if (inode->i_blkbits == PAGE_SHIFT ||
                                        !mapping->a_ops->is_partially_uptodate)
                                goto page_not_up_to_date;
+                       /* pipes can't handle partially uptodate pages */
+                       if (unlikely(iter->type & ITER_PIPE))
+                               goto page_not_up_to_date;
                        if (!trylock_page(page))
                                goto page_not_up_to_date;
                        /* Did it get truncated before we got the lock? */