On Wed, Jan 17, 2024 at 08:55:27PM +0100, Alexander Bluhm wrote:
> On Wed, Jan 17, 2024 at 10:21:43AM +0300, Vitaliy Makkoveev wrote:
> > It has no corresponding wake(9).
> 
> OK bluhm@
> 
> Should we replace NET_UNLOCK(); tsleep_nsec(); NET_LOCK();
> with rwsleep_nsec(&netlock)?
> 

I don't think this this is better. We don't use `netlock' directly
except the sosleep_nsec() which seems to be the special case.

I slightly started to use wg(4) too, so may be I'll remove this hack
completely.

> > Index: sys/net/if_wg.c
> > ===================================================================
> > RCS file: /cvs/src/sys/net/if_wg.c,v
> > retrieving revision 1.35
> > diff -u -p -r1.35 if_wg.c
> > --- sys/net/if_wg.c	1 Jan 2024 18:47:02 -0000	1.35
> > +++ sys/net/if_wg.c	17 Jan 2024 07:19:22 -0000
> > @@ -518,7 +518,7 @@ wg_peer_destroy(struct wg_peer *peer)
> >  			continue;
> >  		}
> >  		NET_UNLOCK();
> > -		tsleep_nsec(sc, PWAIT, "wg_ifq", 1000);
> > +		tsleep_nsec(&nowake, PWAIT, "wg_ifq", 1000);
> >  		NET_LOCK();
> >  	}
> >  	NET_UNLOCK();
>