[dm-devel] [PATCH v4 11/14] treewide: Prepare to remove VLA usage for AHASH_REQUEST_ON_STACK

[Kees Cook]

On Fri, Jul 13, 2018 at 3:14 AM, Arnd Bergmann <arnd at arndb.de> wrote:
> On Fri, Jul 13, 2018 at 8:00 AM, Kees Cook <keescook at chromium.org> wrote:
>> On Thu, Jul 12, 2018 at 10:20 PM, Herbert Xu
>> <herbert at gondor.apana.org.au> wrote:
>>> On Thu, Jul 12, 2018 at 10:17:29PM -0700, Kees Cook wrote:
>>>> What is the correct value to use for AHASH_REQUEST_ON_STACK?
>>>
>>> As I said to arrive at a fixed value you should examine all sync
>>> ahash algorithms (e.g., all shash ones plus ahash ones marked as
>>> sync if there are any).
>>
>> The "value" for the ahash I understand: it has a request size
>> (tfm->reqsize, as set by crypto_ahash_set_reqsize()) what should be
>> used to measure the shash value? (And how does this relate to the
>> value returned by crypto_ahash_reqsize()?) The closest clue I can find
>> is this:
>>
>> crypto_init_shash_ops_async() does:
>>         crt->reqsize = sizeof(struct shash_desc) + crypto_shash_descsize(shash);
>>
>> and that gets called from crypto_ahash_init_tfm(), so if it starts
>> with the above reqsize and adds to it with a call to
>> crypto_ahash_set_reqsize() later, we'll have that maximum?
>>
>> So, do I want to calculate this answer as:
>>
>> sizeof(struct shash_desc) + max(shash descsize) + max(ahash-sync reqsize) ?
>> 16 + 360 + 0
>
> I arrived at the same number, looking at all the sizes in shash,
> The largest I found are sha3_state (360 bytes) and s390_sha_ctx
> (336 bytes), everything else is way smaller.

Excellent. Thanks for double-checking this. :)

>
>> It's 0 above because if I look at all the callers of
>> crypto_ahash_set_reqsize() that do wrapping, all are ASYNC.
>>
>> So, should this really just be 376? Where is best to validate this
>> size, as it seems checking in crypto_ahash_set_reqsize() is
>> inappropriate?
>
> How about crypto_init_shash_ops_async()?

Ah yes, that looks good. Nice find!

After my ahash to shash conversions, only ccm is left as an ahash
user, since it actually uses sg. But with the hard-coded value reduced
to 376, this doesn't trip the frame warnings any more. :)

I'll send an updated series soon.

-Kees

-- 
Kees Cook
Pixel Security