On 2020-04-16 13:02, Stuart Henderson wrote:
> On 2020/04/16 12:38, satmeir wrote:
>> On 2020-04-16 12:20, Stuart Henderson wrote:
>>> On 2020/04/16 11:34, satmeir wrote:
>>>> Updated to 2.31.0
>>>> @solene: Tested patch this time on amd64. Please let me know if it works.
>>>
>>> The emailed patch is line-wrapped and does not apply.
>>>
>>> patch: **** malformed patch at line 233: ${DL_LIB} ${Boost_LIBRARIES}
>>> ${OPENSSL_LIBRARIES} ${ZLIB_LIBRARY}
>>>
>>>> What's the OpenBSD convention about including AESNI Support for machines
>>>> that support it?
>>>
>>> That can be enabled as long as
>>>
>>> 1) detection of cpu features is at runtime not build time (so it doesn't
>>> break when running on x86 machines that do not have AESNI even if built on a
>>> machine with AESNI)
>>>
>>> and
>>>
>>> 2) it does not break non-x86 architectures
>>>
>>>
>>
>> Ah, now I get why the patch isn't applying.
>>
>> I've attached it as a file.
>
> Thanks, that applies now.
>
>> AESNI: Do any examples from the ports tree come to mind that I could
>> check out?
>
> The only one I can think of right now is net/strongswan which builds
> AESNI as a separate plugin on x86. All you'll see from the port
> is how to cope with software that creates additional files on certain
> archs (which won't be necessary unless it's creating additional files
> to be installed). Basically just check that it's doing detection via\
> cpuid at runtime rather than making decisions based on the build cpu
> environment.
>
I think I'll hold off on implementing AESNI for now, as I don't have the
skills or time currently to do it securely. Maybe I'll get it in in the
future, but I think we should proceed with the latest patch.
--
satmeir
use pgp
92E1 AF2A D62E 7B46 00EE DE82 C3C3 BBA2 91DD DB9F
