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

[M. Warner Losh]

In message: <16046.56283.631906.102138 at grasshopper.cs.duke.edu>
            Andrew Gallatin <gallatin at cs.duke.edu> writes:
: 
: M. Warner Losh writes:
:  > 
:  > Comments?
: 
: Mostly OK by me, except for the // style comments.

I added those at the last second, and they are already out of my
tree.  They are bogus, of course.

: Also, I'm not sure that checking for mutex_owned in the ioctl function
: is right.  Picture ifconfig'ing an interface down (or starting
: tcpdump) while you're undergoing a DOS attach.  Can you check for
: suspened instead?

All that mtx_owned does is say 'does the current thread own this
lock'.  How does the DOS scenario relate to this?  In that case, the
current thread wouldn't own the lock (although another might) and
would eventually acquire it to do the tcpdump/ioctl.  I don't object
to checking suspend, but I'm curious as why you think that mtx_owned
might present a probelm.

: I'd still like to wrangle a generic, safe way of detaching interfaces
: out of John though ;)

I would too.  That work kinda dove-tails into the newbus locking I'm
working on.

Warner