How to use kernel crypto

michi1 at michaelblizek.twilightparadox.com michi1 at michaelblizek.twilightparadox.com
Sun Jun 15 02:42:08 EDT 2014


Hi!

On 21:28 Sat 14 Jun     , Freeman Zhang wrote:
> Hi list,
> 
> Recently I'm learning to use kernel crypto. I find some examples but
> they are out of date.
> I manage to write a test program, trying to use aes to encrypt 'buf'
> ,but something goes wrong:
> 
> struct scatterlist sg;
> struct crypto_blkcipher *tfm;
> struct blkcipher_desc desc;
> unsigned char buf[10];
> char *key = "00112233445566778899aabbccddeeff";
> int keylen = 16;
> 
> memset(buf, 'A', 10);
> tfm = crypto_alloc_blkcipher("ecb(aes)", 0, CRYPTO_ALG_ASYNC);
> crypto_blkcipher_setkey(tfm,key,keylen);
> desc.tfm = tfm;
> desc.flags = 1;
> sg_init_one(&sg, buf, 10);
> crypto_blkcipher_encrypt(&desc, &sg, &sg, 10);
> sg_set_buf(&sg, buf,10);
> hexdump(buf,10);
> 
> The result of hexdump(buf) shows that 'buf' stay unchanged. What should
> I do to encrypt the buffer?

Crypto works differently that you probably think it does.

First of all, NEVER NEVER NEVER NEVER NEVER NEVER NEVER use ecb mode. Open
wikipedia to see why. This is one of the most basic mistakes you can make.

The reason why the you see plaintext is probably because the buffer size is
not a multiple of you aes block size (16 bytes). But I must admit that leaving
the data unencrypted instead of e.g. zeroing it does not sound like a good api
design to me...

	-Michi
-- 
programing a layer 3+4 network protocol for mesh networks
see http://michaelblizek.twilightparadox.com



More information about the Kernelnewbies mailing list