Date: Thu, 20 May 1999 11:50:31 -0700 (PDT) From: "Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com> To: pb@fasterix.freenix.org (Pierre Beyssac) Cc: pb@FreeBSD.org (Pierre Beyssac), cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/net if_var.h Message-ID: <199905201850.LAA02987@gndrsh.aac.dev.com> In-Reply-To: <19990520180356.A84590@fasterix.frmug.fr.net> from Pierre Beyssac at "May 20, 1999 06:03:56 pm"
next in thread | previous in thread | raw e-mail | index | archive | help
> On Thu, May 20, 1999 at 08:38:22AM -0700, Rodney W. Grimes wrote:
> > > MFC: change reference count in struct ifaddr to a u_int, to be able
> > > to handle more than 2^16 routes to the same interface.
> >
> > HEADSUP -- This will require compilation of many userland applications
> > that have to deal with networking code due to the change in the size
> > of the struct. In particular anything that prints route metrics or
> > tries to deal with them will barf :-(.
>
> netstat does work, as it uses another interface to get its metrics.
GOOD!
> I've been very careful to check that. Also, I've tested it under
> -current where I'm using gated and didn't have to recompile it
> either.
gated != all ports that may be dealing with ifaddr structs. I'm searching
now for anything that I might be running that could be effected by this.
Also are you using all possible protocols and/or interfaces of gated?
OSPF, BGP4, RIP I/II, EGP, etc... ?
> It doesn't seem to break route or ifconfig either.
^^^^^ This should be proven not to break it by source
code examination, not empirical testing.
> A quick grep through /usr/src/*sbin shows no occurence of "struct
> ifaddr".
What about structures that contain ifaddr's? Are there any?
> > Though I like this change, the interface change is going to cause me
> > and probably many other users great pains on all the boxes we just
> > upgraded to 3.2-STABLE or 3.2-RELEASE.
> > I would like to ask this be backed out until 3.3-RELEASE.
>
> It's purposely not in 3.2-RELEASE, I've been very careful to avoid
> risking breaking the release.
Most folks that upgrade to 3.2-RELEASE start to track 3.2-STABLE, and
you just created a user/kernel interface descepancy.
> I'd really prefer to back it out only if there is evidence of
> serious problems with a given system utility.
Scope should be > system utilities. If it breaks anything in the binaries
of the 3.2 release it should, IMHO, be backed out.
--
Rod Grimes - KD7CAX - (RWG25) rgrimes@gndrsh.aac.dev.com
Accurate Automation, Inc. Reliable computers for FreeBSD
http://www.aai.dnsmgr.com
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199905201850.LAA02987>