Re: git: 5ae69e2f10da - stable/13 - Import the WireGuard driver from zx2c4.com.

On 11/14/22 10:56 AM, Helge Oldach wrote:
> Helge Oldach wrote on Mon, 14 Nov 2022 18:21:23 +0100 (CET):
>> John Baldwin wrote on Mon, 14 Nov 2022 17:55:26 +0100 (CET):
>>> On 11/12/22 9:42 AM, Helge Oldach wrote:
>>>> John Baldwin wrote on Fri, 11 Nov 2022 23:04:20 +0100 (CET):
>>>>> The branch stable/13 has been updated by jhb:
>>>>>
>>>>> URL: https://cgit.FreeBSD.org/src/commit/?id=5ae69e2f10dacd750b0f0e9284fefcbe97eb0476
>>>>>
>>>>> commit 5ae69e2f10dacd750b0f0e9284fefcbe97eb0476
>>>>> Author:     John Baldwin <jhb@FreeBSD.org>
>>>>> AuthorDate: 2022-10-28 20:36:12 +0000
>>>>> Commit:     John Baldwin <jhb@FreeBSD.org>
>>>>> CommitDate: 2022-11-11 21:44:11 +0000
>>>>>
>>>>>       Import the WireGuard driver from zx2c4.com.
>>>>
>>>> Is perhaps an #include <sys/proc.h> missing in if_wg.c?
>>>>
>>>> I'm getting below error on stable/13-n252920-0baacb12181. After manually adding said header include, it compiles cleanly.
>>>
>>> Do you have a custom kernel config?
>>
>> Sort of:
>> include MINIMAL
>> nooption XENHVM
>> nodevice xenpci
> 
> Adding two more data points:
> 
> make KERNCONF=GENERIC compiles without issues.
> 
> The difference seems to be that GENERIC has option VIMAGE but MINIMAL has not: If I add option VIMAGE to my lightly-customized MINIMAL config, it compiles nicely as well.
> 
> VIMAGE triggers inclusion of <sys/proc.h>. However if_wg should not depend on VIMAGE of course. So I suspect the error I'm seeing is a side effect, and sys/proc.h should probably be included verbatim in if_wg.c to be on the safe side. (As mentioned, it is actually included upstream.)

I will add the include.  However, upstream has a lot of excessive #includes that I
trimmed when importing into head.  I had to add about 3-4 when merging back to
stable/13, but it seems I need to add one more.

There are also some missing bits for atomic_load/store_bool for the sanitizer
kernel builds on stable/13 (but the commit to fix that in atomic_san.h doesn't
merge cleanly, so I'm going to ask Mark about that one)

-- 
John Baldwin