mm: do not access page->mapping directly on page_endio
authorMinchan Kim <minchan@kernel.org>
Fri, 24 Feb 2017 22:59:59 +0000 (14:59 -0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 12 Mar 2017 05:37:25 +0000 (06:37 +0100)
commitc5c893e7c44ed8a9dee23cb7e93e93816bae4be4
tree0bd3bd6de03f54c128eaeb385e3095fa8301b368
parent66f43a5768799c1b33fde22fe2231645bb743e0f
mm: do not access page->mapping directly on page_endio

commit dd8416c47715cf324c9a16f13273f9fda87acfed upstream.

With rw_page, page_endio is used for completing IO on a page and it
propagates write error to the address space if the IO fails.  The
problem is it accesses page->mapping directly which might be okay for
file-backed pages but it shouldn't for anonymous page.  Otherwise, it
can corrupt one of field from anon_vma under us and system goes panic
randomly.

swap_writepage
  bdev_writepage
    ops->rw_page

I encountered the BUG during developing new zram feature and it was
really hard to figure it out because it made random crash, somtime
mmap_sem lockdep, sometime other places where places never related to
zram/zsmalloc, and not reproducible with some configuration.

When I consider how that bug is subtle and people do fast-swap test with
brd, it's worth to add stable mark, I think.

Fixes: dd6bd0d9c7db ("swap: use bdev_read_page() / bdev_write_page()")
Signed-off-by: Minchan Kim <minchan@kernel.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
mm/filemap.c