Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
random: use memmove instead of memcpy for remaining 32 bytes
In order to immediately overwrite the old key on the stack, before servicing a userspace request for bytes, we use the remaining 32 bytes of block 0 as the key. This means moving indices 8,9,a,b,c,d,e,f -> 4,5,6,7,8,9,a,b. Since 4 < 8, for the kernel implementations of memcpy(), this doesn't actually appear to be a problem in practice. But relying on that characteristic seems a bit brittle. So let's change that to a proper memmove(), which is the by-the-books way of handling overlapping memory copies. Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net> Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
- Loading branch information