Re: Socket problems with latest -current

From: Doug Barton <dougb_at_FreeBSD.org>
Date: Fri, 15 Dec 2006 11:09:18 -0800
Bruce A. Mah wrote:
> If memory serves me right, Doug Barton wrote:
>> gnn_at_freebsd.org wrote:
>>> At Mon, 11 Dec 2006 01:35:17 -0800,
>>> Doug Barton wrote:
>>>> Doug Barton wrote:
>>>>> With sources cvsup'ed around 20:35 PST on 8 December everything works
>>>>> just fine. With sources cvsup'ed early this morning, 12:42 am PST on
>>>>> 10 December, I'm getting a lot of errors related to sockets:
>>>> I found the problem, it's the change made in version 1.4 of
>>>> rc.d/auto_linklocal. Reverting only that change, and using otherwise
>>>> up to date -current sources allows things to work just fine. Adding
>>>> that change on the same exact system causes the breakage I described
>>>> in my previous message.
>>>>
>>>> It's probably worth mentioning that I exactly fit the criteria from
>>>> the 1.4 commit message, I have INET6 in my kernel, but at the moment I
>>>> have ipv6_enable=no in rc.conf.
>>>>
>>>> This change should be backed out of HEAD and RELENG_6[_2] until the
>>>> cause of this breakage is understood.
>>> I'll take a quick look at this too.  
>> Thanks. I installed a totally generic RELENG_6 system from scratch
>> yesterday, and confirmed that the problem I had in -current exists in
>> -stable too. This is a serious issue, given that the default
>> installation will have INET6 in the kernel, but ipv6_enable=no in
>> /etc/defaults/rc.conf.
>>
>> It would be preferable of course to fix the actual problem, but
>> between "it works" and "it is fully RFC compliant," I have to pick the
>> former.
> 
> Hey Doug--
> 
> I'm trying to reproduce this problem but without success.  I did an
> install of 6.2-RC1/i386 from CD into a Parallels VM, then manually
> updated /etc/rc.d/auto_linklocal to version 1.1.2.3.2.1 (same as 1.4 you
> tested with). 

I'm not sure if they're relevant or not, but the differences between
what you did and what I did are:

1. I installed on a real box
2. I installed 6.1-RELEASE from CD, then cvsup'ed to RELENG_6.

One other possible difference, I'm on a Core 2 Duo, running i386 SMP.

You might also try enabling named for your testing. The default
configuration should give you a simple local resolver which you can
then test things like 'dig _at_127.0.0.1 www.yahoo.com'

hth,

Doug


> I'm running a GENERIC kernel with ipv6_enable="NO" set in
> /etc/rc.conf.  I set sshd_enable="YES" and sendmail_enable="YES" and
> they both seem to be behaving normally.
> 
> vm2# ifconfig
> ed0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
>         inet WW.XX.YY.ZZ netmask 0xffffff00 broadcast WW.XX.YY.255
>         ether 00:5d:54:14:12:1a
>         media: Ethernet autoselect (10baseT/UTP)
> lo0: flags=8049<UP,LOOPBACK,RUNNING,MULTICAST> mtu 16384
>         inet6 fe80::1%lo0 prefixlen 64 scopeid 0x2
>         inet6 ::1 prefixlen 128
>         inet 127.0.0.1 netmask 0xff000000
> 
> I also remember lightly testing this change before it was committed, so
> I'm kind of puzzled by the problems you saw.
> 
> Do you have any other information that you think might be useful?
> 
> Thanks!
> 
> Bruce.
> 


-- 

    This .signature sanitized for your protection
Received on Fri Dec 15 2006 - 18:14:42 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:04 UTC