Re: kern/58168: IFF_OACTIVE semantics and replacement plan is unclear

[mlelstv%serpens.de@localhost (Michael van Elst)]

The following reply was made to PR kern/58168; it has been noted by GNATS.

From: mlelstv%serpens.de@localhost (Michael van Elst)
To: gnats-bugs%netbsd.org@localhost
Cc: 
Subject: Re: kern/58168: IFF_OACTIVE semantics and replacement plan is unclear
Date: Thu, 18 Apr 2024 12:05:22 -0000 (UTC)

 mrg%eterna23.net@localhost (matthew green) writes:
 
 > that check, if needed, should also be moved into the device
 > if_start() if not already there.
 
 That check in net/if.c exists only to optimize away the call of the
 start routine. The common boilerplate for such a (legacy) start routine
 is:
 
 static void
 xxx_start(struct ifnet *ifp) {
 
         if ((ifp->if_flags & (IFF_RUNNING|IFF_OACTIVE)) != IFF_RUNNING)
                 return;
 
 	...
 }
 
 to skip the start when the interface is either unavailable or busy.
 
 But when you require to take extra locks for checking an mp-safe
 driver flag, the additional overhead of the call itself becomes
 neglible.