Re: syslogd starting before IPv6 network is up

Nicolas Blais wrote:
> On Sat, May 23, 2009 at 3:51 PM, Willem Jan Withagen <wjw_at_digiware.nl>wrote:
> 
>> Nicolas Blais wrote:
>>
>>> I migrated one of my 8-CURRENT (Thu May 21) box to IPv6 and since then, I
>>> get the following message at boot:
>>>
>>> syslogd: bind: Can't assign requested address
>>> syslogd: bind: Can't assign requested address
>>> syslogd:
>>> child pid 250 exited with return code 1
>>>
>>> and syslogd obviously fails to start. Apparently this problem has been
>>> reported before, but no actual solution other than start syslogd from
>>> rc.local was given.
>>>
>> It is not because  net.inet6.ip6.v6only is off?
>>
>> Because than you otherwise get ip4 in ip6 mapping.
>> It should be of in recent releases, but in 5.x I remember it being on.

> Perhaps you are right, but that interface requires ip4 to connect to the
> tunnel broker and become ip6-capable. I don't understand why that sysctl
> should have an effect on the rcorder. I checked and net.inet6.ip6.v6only: 1
> 
> Here's my dmesg:
> 
> /etc/rc: WARNING: failed to start syslogd
> Starting Network: lo0 em0.
> May 22 15:59:53 pflogd[620]: [priv]: msg PRIV_OPEN_LOG received
> pf enabled
....
> IPv4 mapped IPv6 address support=NO

Well that should not be the case given what you have.

You could use lsof to see what has the socket open already.
Or start syslog with strace and see where it fails.

Or first check when booting without firewall

--WjW