svn commit: r337944 - head/sys/arm/conf

[John Baldwin]

On 8/17/18 3:53 PM, Ian Lepore wrote:
> On Fri, 2018-08-17 at 01:04 +0000, Conrad Meyer wrote:
>> Author: cem
>> Date: Fri Aug 17 01:04:02 2018
>> New Revision: 337944
>> URL: https://svnweb.freebsd.org/changeset/base/337944
>>
>> Log:
>>   arm: Define crypto option on platforms that include IPsec
>>   
>>   Missed in r337940.
>>   
>>   (It's not like there are any crypto files IPsec doesn't pull in, so it is
>>   unclear what not defining the crypto option was supposed to achieve.)
>>   
>>   Reported by:	np@
>>
>> Modified:
>>   head/sys/arm/conf/std.armv6
>>   head/sys/arm/conf/std.armv7
>>
>> Modified: head/sys/arm/conf/std.armv6
>> ==============================================================================
>> --- head/sys/arm/conf/std.armv6	Fri Aug 17 01:03:23 2018	(r337943)
>> +++ head/sys/arm/conf/std.armv6	Fri Aug 17 01:04:02 2018	(r337944)
>> @@ -9,6 +9,7 @@ options 	VIMAGE			# Subsystem virtualization, e.g. VNE
>>  options 	INET			# InterNETworking
>>  options 	INET6			# IPv6 communications protocols
>>  options 	TCP_HHOOK		# hhook(9) framework for TCP
>> +device		crypto			# IPSec && !crypto is nonsensical
>>  options 	IPSEC			# IP (v4/v6) security
>>  options 	SCTP			# Stream Control Transmission Protocol
>>  options 	FFS			# Berkeley Fast Filesystem
>>
>> Modified: head/sys/arm/conf/std.armv7
>> ==============================================================================
>> --- head/sys/arm/conf/std.armv7	Fri Aug 17 01:03:23 2018	(r337943)
>> +++ head/sys/arm/conf/std.armv7	Fri Aug 17 01:04:02 2018	(r337944)
>> @@ -9,6 +9,7 @@ options 	VIMAGE			# Subsystem virtualization, e.g. VNE
>>  options 	INET			# InterNETworking
>>  options 	INET6			# IPv6 communications protocols
>>  options 	TCP_HHOOK		# hhook(9) framework for TCP
>> +device		crypto			# IPSec && !crypto is nonsensical
>>  options 	IPSEC			# IP (v4/v6) security
>>  options 	SCTP			# Stream Control Transmission Protocol
>>  options 	FFS			# Berkeley Fast Filesystem
>>
> 
> What problem were you trying to solve with this change? Aside from
> putting a device statement into a file that, by design, only contains
> options, and besides adding it with a snarky comment rather than the
> canononical comment associated with that device from sys/conf/NOTES, I
> can't see offhand how this changes anything. Virtually everything that
> is dependent on the crypto device is actually specified as crypto |
> ipsec | ipsec_support, which seems like the correct way to implement
> "option IPSEC implies device crypto".

In other kernel config files we include crypto explicitly, but with a
less snarky comment (and in the device section), e.g. amd64's GENERIC:

# The crypto framework is required by IPSEC
device		crypto			# Required by IPSEC

That said, I think it would be better to instead move this up into the
existing list of pseudo devices that includes 'device ether', etc.

I've put a strawman up at https://reviews.freebsd.org/D16775

-- 
John Baldwin