cvs commit: src/sys/dev/fxp if_fxp.c if_fxpvar.h

[Nate Lawson]

On Tue, 6 May 2003, Peter Jeremy wrote:
> On Fri, Apr 25, 2003 at 06:24:08PM -0700, Nate Lawson wrote:
> >I have run various versions of the patch for about 3 weeks and the final
> >version with no changes for about a week, all without Giant.  The reason
> >why I did not see ifnet problems even though I processed ~400M packets was
> >because all ifnet processing happened to be with the fxp lock held and my
> >laptop only had one network interface.  This is not an intentional part of
> >the patch; it is not an attempt to protect ifnet with a local fxp lock!  
> 
> What about the loopback interface (and maybe other virtual interfaces)?
> Do you have lo0 disabled or does something else protect it?

ifnet locking is the answer