[PATCH] Use Intel AES instruction set if it exists - v4

[Jeremy Allison]

On Thu, Sep 07, 2017 at 01:49:22AM +0200, Stefan Metzmacher wrote:
> Am 06.09.2017 um 22:07 schrieb Jeremy Allison:
> > On Wed, Sep 06, 2017 at 12:10:17PM -0700, Jeremy Allison via samba-technical wrote:
> >> On Wed, Sep 06, 2017 at 08:52:58PM +0200, Stefan Metzmacher wrote:
> >>> Am 06.09.2017 um 18:15 schrieb Jeremy Allison:
> >>>> On Wed, Sep 06, 2017 at 03:38:20PM +0200, Stefan Metzmacher wrote:
> >>>>> Hi Jeremy,
> >>>>>
> >>>>> can you please check the indentation in various wscript files,
> >>>>> they're python scripts and should not use tabs, 4 spaces.
> >>>>>
> >>>>> Also running "git show eb1f476657c" in a recent kernel checkout,
> >>>>> gives a failure.
> >>>>
> >>>> That's a Samba hash, not a kernel one.
> >>>
> >>> No, that's the hash of the file content and I'd assume that doesn't
> >>> change if we just copy the file.
> >>
> >> Ah, got it. We do need to change it to compile as a -fPIC library
> >> for Samba.
> >>
> >>> Thanks!
> >>
> >> I've now got a patchset that starts from a pristine kernel
> >> check-in with a known git refspec and shows what is modified
> >> from there.
> >>
> >> Testing now.
> > 
> > Here it is. Please review. I'm hoping v4 will be the last one,
> > fingers crossed !
> 
> What's the reason to take such an old version?

The modifications to move data to read-only sections and
other changes made it impossible for me to make this into
a Samba shared library. The version I took was the latest
one that contained all the code we use that is easy to
make a shared library (check out the small patch to the
pristine source code that follows the initial check-in).

> There were some fixes in the patchset up to
> cce2ea8d90fec72ed954830e73f5d4995cc79193 (crypto: aesni - add generic
> gcm(aes))

We're not currently using those. I think we should get
that when we do the migration to GNUtls.