Re: [PATCH] crypto/testmgr: don't copy from source IV too much

[Andrey Ryabinin <aryabinin@odin.com>]

On 09/03/2015 04:20 PM, Herbert Xu wrote:
> On Thu, Sep 03, 2015 at 02:32:00PM +0300, Andrey Ryabinin wrote:
>> While the destination buffer 'iv' is MAX_IVLEN size,
>> the source 'template[i].iv' could be smaller. Thus
>> copying it via memcpy() leads to invalid memory access.
>> Use strlcpy() instead.
>>
>> Signed-off-by: Andrey Ryabinin <aryabinin@odin.com>
> 
> Thanks for the patch.  Unfortunately it's broken because the IV
> is not a string and can contain NULs.  So either fix it by using
> the real ivsize,

So I've tried to use crypto_ablkcipher_ivsize(tfm) for that.

But noticed that some algs don't set ivsize (which makes it zero).
E.g. "ecb-cast6-avx" doesn't set it, but test vectors (cast6_enc_tv_template[], cast6_dec_tv_template[])
have .iv of 16bytes.

So I'm not sure what part is wrong here.
Is it wrong to use crypto_ablkcipher_ivsize(tfm) to get ivsize here?
Is it bug in 'ecb-cast6-avx'?
Or maybe something else?


> or change template[i].iv to a char array.
> 
> Cheers,
>