[lkml]   [2017]   [Sep]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH] crypto: caam - properly set IV after {en,de}crypt
On 9/6/2017 1:14 PM, Gilad Ben-Yossef wrote:
> On Tue, Sep 5, 2017 at 6:33 PM, Horia Geantă <> wrote:
>> On 8/14/2017 10:59 AM, Gilad Ben-Yossef wrote:
>>> Hi,
>>> On Thu, Jun 29, 2017 at 1:19 PM, Horia Geantă <> wrote:
>>>> On 6/28/2017 4:42 PM, Horia Geantă wrote:
>>>>> On 6/28/2017 4:27 PM, David Gstir wrote:
>>>>>> Certain cipher modes like CTS expect the IV (req->info) of
>>>>>> ablkcipher_request (or equivalently req->iv of skcipher_request) to
>>>>>> contain the last ciphertext block when the {en,de}crypt operation is done.
>>>>>> This is currently not the case for the CAAM driver which in turn breaks
>>>>>> e.g. cts(cbc(aes)) when the CAAM driver is enabled.
>>>>>> This patch fixes the CAAM driver to properly set the IV after the
>>>>>> {en,de}crypt operation of ablkcipher finishes.
>>>>>> This issue was revealed by the changes in the SW CTS mode in commit
>>>>>> 0605c41cc53ca ("crypto: cts - Convert to skcipher")
>>>>>> Cc: <> # 4.8+
>>>>>> Signed-off-by: David Gstir <>
>>>>> Reviewed-by: Horia Geantă <>
>>>> Btw, instead of updating the IV in SW, CAAM engine could be programmed
>>>> to do it - by saving the Context Register of the AES accelerator.
>>>> Unfortunately this would require changes in quite a few places: shared
>>>> descriptor, HW S/G generation logic, IV dma (un)mapping and maybe others.
>>>> So it's better to have this fix now (which, considering size, is
>>>> appropriate for -stable) and later, if needed, offload IV updating in HW.
>>> My apologies for reviving this thread from the dead, but doesn't the patch fail
>>> for in-place decryption since we are copying from req->dst after
>>> the operation is done, and therefore it no longer contains the ciphertext?
>> You are right, thanks! Will follow up with a fix.
>> Though cts(cbc(aes)) in particular is working, see below.
>>> I'm asking since I ran into a similar issue in the ccree driver and thought
>>> to deploy a similar fix but could not convince myself why this works.
>> IIUC cts(cbc(aes)) in-place decryption (with cbc(aes) offloaded to CAAM
>> engine) works since SW implementation of cts, when performing the
>> ciphertext stealing phase in cts_cbc_decrypt() does not use req->iv, but
>> a previously value, saved before staring decryption in crypto_cts_decrypt():
>> if (cbc_blocks <= 1)
>> memcpy(space, req->iv, bsize);
>> else
>> scatterwalk_map_and_copy(space, req->src, offset - 2 * bsize,
>> bsize, 0);
> Is that not a performance bug in software CTS than? I mean all those
> transformation
> drivers doing that extra copy and possibly malloc and free to save the
> data for the info
> and than have the CTS implementation ignore that and do its own memory copy?
AFAICT, in case cbc_blocks > 1 cts saves the second from last full
block, while underlying cbc subrequest saves the last full block.


 \ /
  Last update: 2017-09-07 12:13    [W:0.126 / U:2.140 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site