Link UP/DOWN problem with re0 on FreeBSD 7.1

[Graeme Dargie]

-----Original Message-----
From: FreeBSD [mailto:freebsd at optiksecurite.com] 
Sent: 04 February 2009 21:55
Cc: freebsd-questions at freebsd.org
Subject: Re: Link UP/DOWN problem with re0 on FreeBSD 7.1

FreeBSD a écrit :
> Graeme Dargie a écrit :
>> If you do a dmesg are you also showing a watchdog time out for the nic ?
>>
>> I only ask as I am having the exact same problem with the exact same 
>> card and I have yet to find a solution, if I come across something I 
>> will let you know.
>>
>> Regards
>> Graeme
> 
> Not a single time...sorry.
> 
>>
>> -----Original Message-----
>> From: FreeBSD [mailto:freebsd at optiksecurite.com] Sent: 26 January 2009 
>> 18:58
>> To: freebsd-questions at freebsd.org
>> Subject: Re: Link UP/DOWN problem with re0 on FreeBSD 7.1
>>
>> FreeBSD a écrit :
>>> Hi everyone,
>>>
>>> Just to put you in context, I applied the following patch to make the 
>>> card available:
>>>
>>> SVN rev 186389 on 2008-12-22 00:46:22Z by yongari
>>>
>>> Since we don't request reset for rlphy(4), the link state 'UP'
>>> event from mii(4) may not be delivered if valid link was already
>>> established. To address the issue, check current link state after
>>> driving MII_TICK. This should fix a regression introduced in
>>> r185753 on fast ethernet controllers.
>>>
>>> ---
>>>
>>> I don't have any issue related to that anymore. The problem is that I 
>>> get link UP/DOWN a few times per hour on 3 identical machines that I 
>>> dumped/restored. They are all pluged in a Cisco switch that works 
>>> fine for every other PCs.
>>>
>>> Jan 26 06:09:15 term005 kernel: re0: link state changed to DOWN
>>> Jan 26 06:09:17 term005 kernel: re0: link state changed to UP
>>>
>>> I tried to switch cables, but I got the same result.
>>>
>>> There is the pciconf -lv output:
>>>
>>> re0 at pci0:3:0:0: class=0x020000 card=0x02831028 chip=0x816810ec 
>>> rev=0x02 hdr=0x00
>>>     vendor     = 'Realtek Semiconductor'
>>>     device     = 'RTL8168/8111 PCI-E Gigabit Ethernet NIC'
>>>     class      = network
>>>     subclass   = ethernet
>>>
>>> There is the output of vmstat -i:
>>>
>>> interrupt                          total       rate
>>> irq18: re0 ehci0++                 63766          0
>>> irq19: atapci0                    277001          3
>>> cpu0: timer                    156068748       1961
>>> Total                          156409515       1966
>>>
>>> Could it be related to the fact that there is re0 and ehci0++ on the 
>>> same IRQ?
>>>
>>> Thank you for your help,
>>>
>>> Martin
>>
>> I just tried with a brand new Dell 2708 switch and the problem is 
>> still there. I just confirmed that the UP/DOWN occurs every 10 minutes 
>> (+- a few seconds).
>>
>> Thanks again,
>>
>> Martin
>>

Just to follow-up on my own problem...

I tried to disable some options of the card with :
ifconfig re0 -rxcsum -txcsum -tso -lro -vlanhwtag

but nothing as changed. I just tried to download a big file (FreeBSD 
7.1-REL DVD iso in fact) to see if the deconnection occurs even during a 
transfer. The DVD downloaded successfully and I verified that the MD5 
are OK. BUT, /var/log/messages continue to tell me that:
Feb  4 16:09:29 term003 kernel: re0: link state changed to UP
Feb  4 16:19:26 term003 kernel: re0: link state changed to DOWN
Feb  4 16:19:30 term003 kernel: re0: link state changed to UP
Feb  4 16:19:32 term003 kernel: re0: link state changed to DOWN

during the transfer (which worked OK). I don't know if that can help 
someone to help me ;)

Thanks,

Martin

_______________________________________________
freebsd-questions at freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to "freebsd-questions-unsubscribe at freebsd.org"

I have a solution to this well a work around.

Add -tso to the relevant line in /etc/rc.conf

ifconfig_re0="inet 192.168.1.103  netmask 255.255.255.0 -tso"

Adding -tso stops the link up / link down problem. Now I am understand that this may increase cpu if the traffic on the nic is high. I am sure some one the list will know of any other implications this may have.

It is a known problem and I site I read the bug had been submitted so hopefully it wont exist in 8.0 

Regards

Graeme