Re: [dm-devel] [PATCH 07/11] crypto: xcbc: Remove VLA usage

[Kees Cook]

On Wed, Jun 20, 2018 at 4:46 PM, Eric Biggers <ebiggers3@xxxxxxxxx> wrote:
> On Wed, Jun 20, 2018 at 12:04:04PM -0700, Kees Cook wrote:
>> In the quest to remove all stack VLA usage from the kernel[1], this uses
>> the maximum blocksize and adds a sanity check.
>>
>> [1] https://lkml.kernel.org/r/CA+55aFzCG-zNmZwX4A2FQpadafLfEzK6CC=qPXydAacU1RqZWA@xxxxxxxxxxxxxx
>>
>> Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
>> ---
>>  crypto/xcbc.c | 5 ++++-
>>  1 file changed, 4 insertions(+), 1 deletion(-)
>>
>> diff --git a/crypto/xcbc.c b/crypto/xcbc.c
>> index 25c75af50d3f..016481b1f3ba 100644
>> --- a/crypto/xcbc.c
>> +++ b/crypto/xcbc.c
>> @@ -65,7 +65,10 @@ static int crypto_xcbc_digest_setkey(struct crypto_shash *parent,
>>       int bs = crypto_shash_blocksize(parent);
>>       u8 *consts = PTR_ALIGN(&ctx->ctx[0], alignmask + 1);
>>       int err = 0;
>> -     u8 key1[bs];
>> +     u8 key1[CRYPTO_ALG_MAX_BLOCKSIZE];
>> +
>> +     if (WARN_ON(bs > sizeof(key1)))
>> +             return -EINVAL;
>
> Similarly, why not MAX_CIPHER_BLOCKSIZE?
>
> Also, xcbc_create() only allows a 16-byte block size, and you made the API
> enforce the maximum for algorithms anyway.  So I think the extra check here
> isn't very worthwhile.

Is the "parent" argument in crypto_xcbc_digest_setkey() always going
to be the "alg" from xcbc_create()? I couldn't convince myself that
was true. If it is, then yes, this VLA can trivially made to be 16,
but it seemed like they were separate instances...

-Kees

-- 
Kees Cook
Pixel Security