Re: [PATCH 1/2] Revert "xattr: mark variable as uninitialized to make both gcc and smatch happy"

[Aristeu Rozanski <aris@redhat.com>]

Sasha,

On Fri, Sep 14, 2012 at 09:35:53PM +0200, Sasha Levin wrote:
> This reverts commit 0142145ddb1d6c841be4eae2c7a32dd18ad34b24.
> 
> Short version:
> 
> Not initializing 'new_xattr' at the beginning of __simple_xattr_set() may lead to
> dereferencing it later on in the function.
> 
> 
> Long version:
> 
> The fix for the warnings generated by smatch due to 'new_xattr' being dereferenced
> without a check from being non-NULL is incorrect.
> 
> The problem is that the fix removed initialization of new_xattr with NULL, which
> meant that new_xattr could be anything at the beginning of __simple_xattr_set(),
> and might have not been initialized at any point throughout the function.
> 
> In case new_xattr does get left uninitialized ('value == 0' case) and XATTR_REPLACE
> being set, the fix will actually lead us to dereferencing new_xattr even if we wouldn't
> have done so before.
> 
> Why? Looking at the original code:
> 
>         if (flags & XATTR_REPLACE) {
>                 xattr = new_xattr;
>                 err = -ENODATA;
>         } else if (new_xattr) {
>                 list_add(&new_xattr->list, &xattrs->head);
>                 xattr = NULL;
>         }
> out:
>         spin_unlock(&xattrs->lock);
>         if (xattr) {
>                 kfree(xattr->name);
>                 kfree(xattr);
>         }
>         return err;

not to mention this:
	list_for_each_entry(xattr, &xattrs->head, list) {
		if (!strcmp(name, xattr->name)) {
			if (flags & XATTR_CREATE) {
				xattr = new_xattr;
				err = -EEXIST;
			} else if (new_xattr) {
				list_replace(&xattr->list, &new_xattr->list);
                                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
			} else {
				list_del(&xattr->list);
			}
			goto out;
		}
	}

Good catch.

-- 
Aristeu