On Tue, Apr 28, 2015 at 04:35:57AM +0200, Stephan Mueller wrote:
>
> In this case, shouldn't we just have a loop where:
>
> 1. from the given endpoint, we go a semiblock back
>
> 2. now we see how many bytes we get when fetching the SG list till the end,
>
> 3a. if answer from 2 is semiblock or
Am Dienstag, 28. April 2015, 09:10:47 schrieb Herbert Xu:
Hi Herbert,
> On Mon, Apr 27, 2015 at 04:34:19PM +0200, Stephan Mueller wrote:
> > Why do you think that will not work? I thought that the code works when
> > the
> > non-linear scatterlists are at least broken at an 8 byte boundary.
>
>
On Mon, Apr 27, 2015 at 04:58:51PM +0200, Stephan Mueller wrote:
>
> This memcmp implies that the final block->A from the decrypt is memcpy'ed to
> req->creq.info. I wanted to avoid any additional memcpy calls to not hurt
> performance even more.
I was hoping to directly use req->creq.info in th
On Mon, Apr 27, 2015 at 04:34:19PM +0200, Stephan Mueller wrote:
>
> Why do you think that will not work? I thought that the code works when the
> non-linear scatterlists are at least broken at an 8 byte boundary.
There is no guarantee that SG lists are set at 8-byte boundaries.
In fact, you need
Am Montag, 27. April 2015, 16:29:35 schrieb Herbert Xu:
Hi Herbert,
>On Sun, Apr 26, 2015 at 12:08:20AM +0200, Stephan Mueller wrote:
>> This patch implements the AES key wrapping as specified in
>> NIST SP800-38F and RFC3394.
>
>This is my attempt at turning kw into a givcipher. The encrypt
>pa
Am Montag, 27. April 2015, 16:26:07 schrieb Herbert Xu:
Hi Herbert,
>On Sun, Apr 26, 2015 at 12:08:20AM +0200, Stephan Mueller wrote:
>> +/*
>> + * Point to the end of the scatterlists to walk them
backwards.
>> + */
>> +src_walk.offset += src_nbyt
On Sun, Apr 26, 2015 at 12:08:20AM +0200, Stephan Mueller wrote:
> This patch implements the AES key wrapping as specified in
> NIST SP800-38F and RFC3394.
This is my attempt at turning kw into a givcipher. The encrypt
part is complete but untested as I gave up after finding the
reverse SG proble
On Sun, Apr 26, 2015 at 12:08:20AM +0200, Stephan Mueller wrote:
>
> + /*
> + * Point to the end of the scatterlists to walk them backwards.
> + */
> + src_walk.offset += src_nbytes;
> + dst_walk.offset += dst_nbytes;
This doesn't work.
