cryptocheck: Expand the set of sizes tested by -z.

[jhb]

Summary

Test individual sizes up to the max encryption block length as well as
a few sizes that include 1 full block and a partial block before
doubling the size.

Sponsored by: Netflix

[cem]

tools/tools/crypto/cryptocheck.c

1738

16 isn’t a partial block for any 128-bit cipher, like AES?

[jhb]

tools/tools/crypto/cryptocheck.c

1738

16 isn’t a partial block for any 128-bit cipher, like AES?

Ugh, only for Chacha I guess. I could lower the increment to 8.

[cem]

tools/tools/crypto/cryptocheck.c

1738

8 sounds good to me. No one should be using 64-bit block ciphers anymore (DES).

[jhb]

Altered step in second block to 8.

[cem]

Merci

[markj]

tools/tools/crypto/cryptocheck.c

231

I don't quite understand the relationship between EALG_MAX_BLOCK_LEN and the number of sizes we use. We test

• EALG_MAX_BLOCK_LEN sizes, plus
• EALG_MAX_BLOCK_LEN / 8 sizes, plus
• log2(240 * 1024) - log2(EALG_MAX_BLOCK_LEN * 2) - 1 sizes

so with the current max block length of 128 == EALG_MAX_BLOCK_LEN * 2 it just happens to work.

[jhb]

tools/tools/crypto/cryptocheck.c

231

Before it was also a bit of a guess at a number (the magic 128). The assertion _should_ ensure that we don't overflow the array in practice. It's mostly just laziness to avoid having to deal with malloc. Alternatively this could be in C++ with a std::vector<> which would avoid the need for static sizes here as well as the helper variables (naad_sizes and nsizes). I mostly just used N * 2 I as I know we needed at least N + 1 and my plan was to let '-z' assert if it was too small and if so bump it up to N * 3, etc. That is roughly the plan I used when I used 128 before though it was more of a guess that worked the first time.