vfs: __read_cache_page should use gfp argument rather than GFP_KERNEL
authorDave Kleikamp <dave.kleikamp@oracle.com>
Wed, 21 Dec 2011 17:05:48 +0000 (11:05 -0600)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 22 Dec 2011 01:02:46 +0000 (17:02 -0800)
lockdep reports a deadlock in jfs because a special inode's rw semaphore
is taken recursively.  The mapping's gfp mask is GFP_NOFS, but is not
used when __read_cache_page() calls add_to_page_cache_lru().

Signed-off-by: Dave Kleikamp <dave.kleikamp@oracle.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/filemap.c

index c106d3b..5f0a3c9 100644 (file)
@@ -1828,7 +1828,7 @@ repeat:
                page = __page_cache_alloc(gfp | __GFP_COLD);
                if (!page)
                        return ERR_PTR(-ENOMEM);
-               err = add_to_page_cache_lru(page, mapping, index, GFP_KERNEL);
+               err = add_to_page_cache_lru(page, mapping, index, gfp);
                if (unlikely(err)) {
                        page_cache_release(page);
                        if (err == -EEXIST)
@@ -1925,10 +1925,7 @@ static struct page *wait_on_page_read(struct page *page)
  * @gfp:       the page allocator flags to use if allocating
  *
  * This is the same as "read_mapping_page(mapping, index, NULL)", but with
- * any new page allocations done using the specified allocation flags. Note
- * that the Radix tree operations will still use GFP_KERNEL, so you can't
- * expect to do this atomically or anything like that - but you can pass in
- * other page requirements.
+ * any new page allocations done using the specified allocation flags.
  *
  * If the page does not get brought uptodate, return -EIO.
  */