Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 13 Jan 2009 08:12:24 -0700
From:      John Hein <jhein@timing.com>
To:        Markus Hitter <mah@jump-ing.de>
Cc:        freebsd-embedded@freebsd.org
Subject:   Re: 7.1 Release broke my box
Message-ID:  <18796.44888.658271.953215@gromit.timing.com>
In-Reply-To: <78392165-0ACF-4473-8C34-72E5147B344C@jump-ing.de>
References:  <5A12A808-AC77-478A-A5E8-A13B63EEC3A1@jump-ing.de> <18794.45109.504512.894659@gromit.timing.com> <78392165-0ACF-4473-8C34-72E5147B344C@jump-ing.de>

next in thread | previous in thread | raw e-mail | index | archive | help
Markus Hitter wrote at 09:52 +0100 on Jan 12, 2009:
 > 
 > Am 12.01.2009 um 03:51 schrieb John Hein:
 > 
 > > Markus Hitter wrote at 01:31 +0100 on Jan 12, 2009:
 > >> Last Friday, after seeing ext2fs doesn't work with the 7.0 kernel, I
 > >> decided to do yet another step and upgrade to 7.1 Release. The
 > >> virtual machine for building tinybsd setup and ran flawlessly, but a
 > >> tinybsd image built from it doesn't like networking any longer. The
 > >> non-working box is a Geode-based Flepo Alpha, which uses the re
 > >> network driver.
 > >
 > > What happens if you build a kernel with the 7.0 version of
 > > sys/dev/re/if_re.c?  You may have to roll back sys/pci/if_rlreg.h too.
 > 
 > The exactly same (had to roll back sys/pci/if_rl.c as well).
 > 
 > > What happens to the link light while these things are happening?
 > 
 > The green light stays on, the yellow one flickers, just like expected.
 > 
 > Obviously, the transmission gets interrupted at some higher level. Do  
 > I really have to bisect the entire kernel with it's 10,000 revisions  
 > now? *sigh*

My guess was the first obvious one I could think of since I knew
changes happened in the re/rl drivers.  I can't think of anything else
obvious between 7.0 & 7.1 that would affect this (particularly since I
have an re that works fine in 7.1 - at least a 7.1 from, say,
September).

As far as bisecting goes, I recommend binary search by date on the
RELENG_7 branch rather than trying to rollback individual revs.  Note
that cvs 1.12 allows checking out by date (-D) on a branch
(-rRELENG_7) (but not 1.11).  Or you can use cvsup to help with that I
think.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?18796.44888.658271.953215>