Re: aes: self-test failed

[Harold Gutch <logix%foobar.franken.de@localhost>]

Hi,

On Sat, Mar 18, 2023 at 08:16:07AM +0900, Mori Hiroki wrote:
> [   1.0300028] BearSSL aes_ct AES-192-CBC enc: self-test failed
> [   1.0300028] was: 144 bytes @ 0xc04f794d
> [...]
> [   1.0300028] expected: 144 bytes @ 0xc037f8f8
> [...]

The two buffers differ in the following way:

0x53: 0x4f -> 0x5f (0x10)
0x57: 0x8d -> 0xe4 (0x69)
0x5b: 0x40 -> 0x8f (0xcf)
0x5f: 0x49 -> 0x6b (0x22)
0x63: 0x7e -> 0xeb (0x95)
0x67: 0x03 -> 0x9c (0x9f)
0x6b: 0x30 -> 0xb0 (0x80)
0x6f: 0x22 -> 0x45 (0x9e)

(location, value in first buffer, value in second buffer, xor of the
two)


> [   1.0300028] BearSSL aes_ct AES-128-XTS enc: self-test failed
> [   1.0300028] was: 144 bytes @ 0xc04f794d
> [...]
> [   1.0300028] expected: 144 bytes @ 0xc037fa80
> [...]

and these differ as follows:

0x13: 0x6c -> 0xbe (0xd2)
0x17: 0xe8
0x1b: 0x8d -> 0x81 (0x0c)
0x1f: 0xc9 -> 0x9c (0x55)
0x23: 0x8d -> 0xf5 (0x78)
0x27: 0xc9 -> 0x5f (0x96)
0x2b: 0x6c -> 0xfa (0x96)
0x2f: 0xc4 -> 0x00 (0xc4)

0x17 is 0xe8 both times - but this might be just by coincidence.

I don't see a pattern in the changes, but the locations are all at
offsets 4N+3.  Is this perhaps a signedness issue with 32 bit values?


regards,
  Harold