Re: short read from /dev/urandom

[Ulrich Drepper]

On Fri, 14 Jan 2005 05:56:41 +0000 (UTC), David Wagner
<daw@xxxxxxxxxxxxxxxxxxxxxxxx> wrote:

> True.  Arguably, the solution is to fix the documentation.

The problem is that no-short-reads behavior has been documented for a
long time and so programs might, correctly so, use

    while (read(fd, buf, sizeof buf) == -1)
      continue;

Image a program doing this.  It provides the possibility for a local
attack.  If one can determine the content of the to-be-filled buffer
before the 'read', then an attacker could limit the randomness in the
buffer after the read by sending signals to the program.

Not breaking the ABI is more important than symmetry.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/