[regression] fib6_del() bug from 2.6.34-rc1 still present in 2.6.34

[Julien BLACHE]

Hi,

[subscribed to lkml but not netdev, Cc me on replies]

I'm seeing a warning in fib6_del() that is very close to what was
reported by Emil S Tantilov back in march/april for 2.6.34-rc1:

<http://kerneltrap.org/mailarchive/linux-netdev/2010/4/9/6274401/thread>

It looks like there hasn't been a fix, other than what was mentioned in
this thread for net-next and Emil reported that it did not fix it for
him. So it looks like it's still there, alive and kicking.

This is the warning I'm getting:

------------[ cut here ]------------
WARNING: at net/ipv6/ip6_fib.c:1160 fib6_del+0x506/0x5b0()
Hardware name: MacBookPro2,2
Modules linked in: sco bnep rfcomm l2cap crc16 cpufreq_userspace cpufreq_powersave cpufreq_conservative nfsd nfs lockd auth_rpcgss sunrpc uinput btusb ath9k ath9k_common mac80211 ath9k_hw ath isight_firmware joydev cfg80211 i2c_i801 ohci1394 ieee1394 [last unloaded: scsi_wait_scan]
Pid: 4020, comm: ifconfig Not tainted 2.6.34 #1
Call Trace:
 [<ffffffff810389d3>] ? warn_slowpath_common+0x73/0xb0
 [<ffffffff8142f956>] ? fib6_del+0x506/0x5b0
 [<ffffffff8102ceb3>] ? __wake_up+0x43/0x70
 [<ffffffff813c68ef>] ? netlink_broadcast+0x21f/0x410
 [<ffffffff8142c2ab>] ? __ip6_del_rt+0x4b/0x80
 [<ffffffff8142c436>] ? ip6_del_rt+0x26/0x30
 [<ffffffff81426dff>] ? __ipv6_ifa_notify+0x15f/0x200
 [<ffffffff81428d99>] ? addrconf_ifdown+0x159/0x350
 [<ffffffff8142915d>] ? addrconf_notify+0xed/0x920
 [<ffffffff81043d33>] ? lock_timer_base+0x33/0x70
 [<ffffffff810445ab>] ? mod_timer+0x11b/0x1a0
 [<ffffffff81054826>] ? notifier_call_chain+0x46/0x70
 [<ffffffff813b1ae5>] ? __dev_notify_flags+0x65/0x90
 [<ffffffff813b1b4b>] ? dev_change_flags+0x3b/0x70
 [<ffffffff813fd2a2>] ? devinet_ioctl+0x602/0x750
 [<ffffffff813a12ea>] ? T.945+0x1a/0x50
 [<ffffffff813a1589>] ? sock_ioctl+0x59/0x2a0
 [<ffffffff810bee55>] ? vfs_ioctl+0x35/0xd0
 [<ffffffff810bf018>] ? do_vfs_ioctl+0x88/0x570
 [<ffffffff810bf549>] ? sys_ioctl+0x49/0x80
 [<ffffffff810023eb>] ? system_call_fastpath+0x16/0x1b
---[ end trace b5a833c8e5539431 ]---

I can reliably reproduce it on both ath9k and sky2 with the
following sequence:

 # ifconfig eth0 up
 # ifconfig eth0 add 2001:7a8:5dd7:123::12/64
 # ifconfig eth0 down
 # ifconfig eth0 up
 # ifconfig eth0 add 2001:7a8:5dd7:123::12/64 <=== fails
 # ifconfig eth0 down <=== triggers the warning

Note that this sequence is equivalent to:
 # ifup eth0
 # ifdown eth0
 # ifup eth0 (will fail because it cannot add the v6 address)
 # ifconfig eth0 down

This regression breaks ifupdown as it always tries to add the v6 address
when configuring the interface. It's a behaviour change compared to
previous kernel versions.

It looks like triggering this warning a couple times (3-4) in a row ends
up locking up the machine, too.

I can test patches etc.

JB.

-- 
Julien BLACHE                                   <http://www.jblache.org> 
<jb@xxxxxxxxxxx>                                  GPG KeyID 0xF5D65169
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/