debug.mpsafenet=1 vs. user/group rules [Re: kern/106805: ...]

[Christian S.J. Peron]

Tai-hwa Liang wrote:
> On Fri, 29 Dec 2006, Christian S.J. Peron wrote:
>> Max,
>>
>> I have replied to this mail and I guess it has been lost, as I have 
>> had no response.  Although this technically makes
>> the problem harmless, all you are doing is moving the lock order 
>> reversal from pf+inpcb to pfil+inpcb. The
>
>   I probably missed something; however, with Max's patch applied, I 
> did not see any pf related LOR on a WITNESS + INVARIANT enabled 
> -STABLE box during
> last two weeks [..]

You won't see it on -STABLE because it doesn't exist. We switched from a 
home rolled reader/writer type lock (with no WITNESS semantics), to a 
standard read/write locking API in -CURRENT.