Unsetting FMODE_NONOTIFY in fsnotify_open() is too late, since fsnotify_perm() is called before. If FMODE_NONOTIFY is set fsnotify_perm() will skip permission checks, so a user can still disable permission checks by setting this flag in an open() call. This patch corrects this by unsetting the flag before fsnotify_perm is called. Signed-off-by: Lino Sanfilippo <LinoSanfilippo@xxxxxx> --- This patch applies against 2.6.37-rc1 It corrects this unsufficient fix http://lkml.org/lkml/2010/10/29/125 I totally oversaw those fsnotify_perm() calls :(. Hopefully this one will set things right... diff -Nurp a/fs/namei.c b/fs/namei.c --- a/fs/namei.c 2010-11-01 12:54:12.000000000 +0100 +++ b/fs/namei.c 2010-11-05 10:53:57.000000000 +0100 @@ -1748,6 +1748,9 @@ struct file *do_filp_open(int dfd, const if (!(open_flag & O_CREAT)) mode = 0; + /* Must never be set by userspace */ + open_flag &= ~FMODE_NONOTIFY; + /* * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only * check for O_DSYNC if the need any syncing at all we enforce it's diff -Nurp a/include/linux/fsnotify.h b/include/linux/fsnotify.h --- a/include/linux/fsnotify.h 2010-11-01 12:54:12.000000000 +0100 +++ b/include/linux/fsnotify.h 2010-11-05 10:49:51.000000000 +0100 @@ -235,9 +235,6 @@ static inline void fsnotify_open(struct if (S_ISDIR(inode->i_mode)) mask |= FS_ISDIR; - /* FMODE_NONOTIFY must never be set from user */ - file->f_mode &= ~FMODE_NONOTIFY; - fsnotify_parent(path, NULL, mask); fsnotify(inode, mask, path, FSNOTIFY_EVENT_PATH, NULL, 0); } -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
