vfs: allocate page instead of names_cache buffer in mount_block_root

First, it's incorrect to call putname() after __getname_gfp() since the
bare __getname_gfp() call skips the auditing code, while putname()
doesn't.

mount_block_root allocates a PATH_MAX buffer via __getname_gfp, and then
calls get_fs_names to fill the buffer. That function can call
get_filesystem_list which assumes that that buffer is a full page in
size. On arches where PAGE_SIZE != 4k, then this could potentially
overrun.

In practice, it's hard to imagine the list of filesystem names even
approaching 4k, but it's best to be safe. Just allocate a page for this
purpose instead.

With this, we can also remove the __getname_gfp() definition since there
are no more callers.

Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

1 files changed, 4 insertions, 3 deletions

diff --git a/init/do_mounts.c b/init/do_mounts.c
index d3f0aeed2d39..f8a66424360d 100644
--- a/init/do_mounts.c
+++ b/init/do_mounts.c
@@ -353,8 +353,9 @@ static int __init do_mount_root(char *name, char *fs, int flags, void *data)
 
 void __init mount_block_root(char *name, int flags)
 {
-        char *fs_names = __getname_gfp(GFP_KERNEL
-                | __GFP_NOTRACK_FALSE_POSITIVE);
+        struct page *page = alloc_page(GFP_KERNEL |
+                                        __GFP_NOTRACK_FALSE_POSITIVE);
+        char *fs_names = page_address(page);
         char *p;
 #ifdef CONFIG_BLOCK
         char b[BDEVNAME_SIZE];
@@ -406,7 +407,7 @@ retry:
 #endif
         panic("VFS: Unable to mount root fs on %s", b);
 out:
-        putname(fs_names);
+        put_page(page);
 }
  
 #ifdef CONFIG_ROOT_NFS