JFFS2 Fix of panics caused by wrong condition for hole frag creation in write_begin

This fixes a regression introduced in commit
205c109a7a96d9a3d8ffe64c4068b70811fef5e8 when switching to
write_begin/write_end operations in JFFS2.

The page offset is miscalculated, leading to corruption of the fragment
lists and subsequently to memory corruption and panics.

[ Side note: the bug is a fairly direct result of the naming.  Nick was
  likely misled by the use of "offs", since we tend to use the notion of
  "offset" not as an absolute position, but as an offset _within_ a page
  or allocation.

  Alternatively, a "pgoff_t" is a page index, but not a byte offset -
  our VM naming can be a bit confusing.

  So in this case, a VM person would likely have called this a "pos",
  not an "offs", or perhaps talked about byte offsets rather than page
  offsets (since it's counted in bytes, not pages).    - Linus ]

Signed-off-by: Alexey Korolev <akorolev@infradead.org>
Signed-off-by: Vasiliy Leonenko <vasiliy.leonenko@mail.ru>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

diff --git a/fs/jffs2/file.c b/fs/jffs2/file.c
index f9c5dd6..dcc2734 100644 (file)
--- a/fs/jffs2/file.c
+++ b/fs/jffs2/file.c
@@ -129,7 +129,7 @@ static int jffs2_write_begin(struct file *filp, struct address_space *mapping,
        struct inode *inode = mapping->host;
        struct jffs2_inode_info *f = JFFS2_INODE_INFO(inode);
        pgoff_t index = pos >> PAGE_CACHE_SHIFT;
-       uint32_t pageofs = pos & (PAGE_CACHE_SIZE - 1);
+       uint32_t pageofs = index << PAGE_CACHE_SHIFT;
        int ret = 0;
 
        pg = __grab_cache_page(mapping, index);