BETA3 showstoppers (read: critical bugs)

Matthias Andree ma at
Mon Sep 6 10:03:40 PDT 2004


5.3-BETA3 (cvsupped around 1000 UTC today) appears to exhibit multiple
critical problems (haven't yet had the time to figure if its mpsafenet
related, but I only have an Athlon XP 2500+ uniprocessor machine):

- netgraph broken:

  1. "ngctl list" lists nothing in spite of an installed ether.bridge.

  2. bridge stats broken

  3. user-ppp does not find "ether" for ng_pppoe, see PR #71424

- rp-pppoe recently goes into 100% busy wait loop, truss shows:

  select(5,{0 3 4},0,0,0) = 1
  in a tight loop, which is kind of bogus. Network seems fine though at
  first connect. This regression was introduced recently, too.

- rp-pppoe can't reconnect if it's killed (not yet researched into),
  apparently related to other network or netgraph problems. It appears
  as though rp-pppoe would not see the reply packets from the network
  and may be an xl(4) bug.

- xl(4) autoconfig appears to break netgraph, was broken a few weeks ago
  and discussed here without solution

  workaround: ifconfig xl1 media 10baseT/UTP

  I cannot checked if this has been fixed, depends on 71424 being fixed

- a mounted ext2 file system (read-only sufficient) still causes FreeBSD
  to not flush its own ufs super blocks at shut down, causing an fsck
  run at next reboot.
  Workaround: umount all ext2 file systems before reboot/halt.

If I can help to have these fixed, test patches, do some debugging work
whatever, so that BETA4 fixes these, let me know. I know and program in
C but I am not acquainted with the FreeBSD kernel, so please don't
instruct me too tersely.

Matthias Andree

Encrypted mail welcome: my GnuPG key ID is 0x052E7D95 (PGP/MIME preferred)

More information about the freebsd-current mailing list