On Mon, Oct 18, 2021 at 04:17:57PM +0200, Andrej Shadura wrote:

On 18/10/2021 16:15, Alan Stern wrote:

quoted

On Mon, Oct 18, 2021 at 02:21:43PM +0200, Andrej Shadura wrote:

quoted

The previous commit fixed handling of incomplete packets but broke error
handling: offsetof returns an unsigned value (size_t), but when compared
against the signed return value, the return value is interpreted as if
it were unsigned, so negative return values are never less than the
offset.

Fixes: 22d65765f211 ("HID: u2fzero: ignore incomplete packets without data")
Fixes: 42337b9d4d95 ("HID: add driver for U2F Zero built-in LED and RNG")
Signed-off-by: Andrej Shadura <redacted>
---
  drivers/hid/hid-u2fzero.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/hid/hid-u2fzero.c b/drivers/hid/hid-u2fzero.c
index d70cd3d7f583..5145d758bea0 100644
--- a/drivers/hid/hid-u2fzero.c
+++ b/drivers/hid/hid-u2fzero.c

@@ -200,7 +200,7 @@ static int u2fzero_rng_read(struct hwrng *rng, void *data,
  	ret = u2fzero_recv(dev, &req, &resp);
  	/* ignore errors or packets without data */
-	if (ret < offsetof(struct u2f_hid_msg, init.data))
+	if (ret < 0 || ret < offsetof(struct u2f_hid_msg, init.data))

Although the patch description does a good job of explaining what's
happening, someone merely reading the code will most likely not
understand.

One alternative is to add a comment.  Another is simply to force a
signed integer comparison:

	if (ret < (ssize_t) offsetof(...

I have considered that, but I thought that is actually less readable than
having two conditions. I’m curious that you say "ignore errors or packets
without data" is not clear enough — how would you reword that without
inflating it too much?

You misunderstand.  The existing comment is clear enough.  But the code 
itself is misleading:

	if (ret < 0 || ret < offsetof(...

looks redundant.  Someone reading it for the first time will 
automatically think: "If ret < 0 then certainly it is < the offset of 
some internal field.  So why perform two comparisons when one is 
enough?"

To help such a reader understand what is happening, you could add a 
comment like:

	/*
	 * offsetof returns an unsigned value, so the comparison with
	 * ret uses unsigned arithmetic and won't detect a negative
	 * error value.  We need a separate test for errors.
	 */

If you think a comment like this is preferable to a typecast, fine.

Another alternative is:

	ret -= offsetof(...);
	if (ret < 0)
		return 0;

which may look more complicated but allows you to simplify the max3 
computation in the next line.

Alan Stern