Re: [PATCH v12 01/12] lib: introduce copy_struct_{to,from}_user helpers

On Thu, Sep 05, 2019 at 01:17:38PM +0200, Rasmus Villemoes wrote:

On 05/09/2019 13.05, Christian Brauner wrote:

quoted

On Thu, Sep 05, 2019 at 06:19:22AM +1000, Aleksa Sarai wrote:

quoted

quoted

+	if (unlikely(!access_ok(dst, usize)))
+		return -EFAULT;
+
+	/* Deal with trailing bytes. */
+	if (usize < ksize) {
+		if (memchr_inv(src + size, 0, rest))
+			return -EFBIG;
+	} else if (usize > ksize) {
+		if (__memzero_user(dst + size, rest))
+			return -EFAULT;

Is zeroing that memory really our job? Seems to me we should just check
it is zeroed.

Of course it is, otherwise you'd require userspace to clear the output
buffer it gives us, which in the majority of cases is wasted work. It's
much easier to reason about if we just say "the kernel populates [uaddr,
uaddr + usize)".

I don't really mind either way so sure. :)