Hi Michael,

Considering this is a very obscure kernel bug that has been around for a long 
time, and AstLinux 1.4.x will have it fixed ... I personally would not get 
carried away implementing the workaround.

BTW, the PHYETH_DISABLE_OFFLOAD workaround only applies to physical ethernet 
NICs.

You understand the situation, so do what you think is best.

Lonnie




> On Sep 20, 2020, at 3:45 PM, Michael Knill 
> <michael.kn...@ipcsolutions.com.au> wrote:
> 
> Thanks Lonnie
> 
> One last question; could I just blanket add this to all my systems or are 
> there any that will certainly not be affected by this bug e.g. VM or virtual 
> NIC?
> The plan is to have this command active by default and commented out if 
> desired. 
> 
> Regards
> Michael Knill
> 
> On 20/9/20, 9:58 pm, "Lonnie Abelbeck" <li...@lonnie.abelbeck.com> wrote:
> 
>> Do you think I should remove PHYETH_DISABLE_OFFLOAD and try a BIOS upgrade 
>> to see if it fixes the problem?
> 
>    No ... first make sure setting PHYETH_DISABLE_OFFLOAD solves the issue 
> with your DLS provider and your APU2s.
> 
>    After that, AstLinux 1.4.0 will be the solution without setting 
> PHYETH_DISABLE_OFFLOAD .
> 
>    Ideally, you could test a 1.4-pre-release on an APU2 with the new DLS 
> provider to make certain we have identified the issue.
>    https://www.astlinux-project.org/dev.html
> 
>    Lonnie
> 
> 
>> On Sep 20, 2020, at 5:24 AM, Michael Knill 
>> <michael.kn...@ipcsolutions.com.au> wrote:
>> 
>> Yes my BIOS is quite old:
>> 3999-IPCBuild-CM1 kd # dmesg | grep DMI
>> [    0.000000] DMI: PC Engines APU, BIOS SageBios_PCEngines_APU-45 04/05/2014
>> 
>> Funny as it's a pretty new box.
>> 
>> Do you think I should remove PHYETH_DISABLE_OFFLOAD and try a BIOS upgrade 
>> to see if it fixes the problem?
>> I don't think I will for existing sites as a BIOS upgrade looks pretty hard 
>> to do and I assume cannot be done remotely.
>> 
>> Regards
>> Michael Knill
>> 
>> On 20/9/20, 12:18 pm, "Lonnie Abelbeck" <li...@lonnie.abelbeck.com> wrote:
>> 
>>   Hi Michael,
>> 
>>   Ahhh, very good ... looks like we are on to something.
>> 
>>   Add it to one/some of your APU2s and let us know how it goes.
>> 
>>   As far as my Qotom Q190G4N, I initially had to set PHYETH_DISABLE_OFFLOAD 
>> to keep it from locking-up with sustained high network traffic but then 
>> switched the RAM SO-DIMM with another brand and did not need 
>> PHYETH_DISABLE_OFFLOAD anymore. My comments probably got you started adding 
>> PHYETH_DISABLE_OFFLOAD.
>> 
>>   This is a very obscure kernel bug, as such it never got back-ported to 
>> Linux 3.16.x .
>> 
>>   For the APU2, the BIOS could play a role in how it initializes the NICs 
>> and whether this kernel bug is triggered.
>> 
>>   Lonnie
>> 
>> 
>> 
>>> On Sep 19, 2020, at 7:49 PM, Michael Knill 
>>> <michael.kn...@ipcsolutions.com.au> wrote:
>>> 
>>> Awesome thanks Lonnie.
>>> Yes its all making sense now. I already have this directive in my template 
>>> against the Qotom Q190G4U for some reason (should I have?)
>>> I have two Qotoms connected to the problem provider, one had this directive 
>>> set already and has not failed and one did not (I forgot to change when I 
>>> changed hardware) which fails.
>>> All my APU's don't have this set so all have problems with this provider.
>>> 
>>> I'm thinking we have finally solved this issue.
>>> Thanks so much for your help
>>> 
>>> Regards
>>> Michael Knill
>>> 
>>> On 20/9/20, 9:29 am, "Lonnie Abelbeck" <li...@lonnie.abelbeck.com> wrote:
>>> 
>>>  I would try this first
>>>  --
>>>  PHYETH_DISABLE_OFFLOAD="tso gso gro"
>>>  --
>>>  and see if it fixes the problem.
>>> 
>>>  If by chance it does fix it, then it would not be needed in AstLinux 1.4.x.
>>> 
>>>  The PHYETH_DISABLE_OFFLOAD settings disable some of the "offload" features 
>>> of the NICs in an effort to work around this (somewhat obscure) kernel bug.
>>> 
>>>  It all kind of makes sense that a particular provider is fragmenting 
>>> packets in ways others do not, and hits this kernel bug.
>>> 
>>>  The PHYETH_DISABLE_OFFLOAD setting above is very "safe", only drawback is 
>>> it slightly reduces network performance near the 1 Gbps level.
>>> 
>>>  BTW, if traffic shaping is enabled this PHYETH_DISABLE_OFFLOAD setting is 
>>> already applied to external ethernet NIC(s).
>>> 
>>>  Lonnie
>>> 
>>> 
>>> 
>>>> On Sep 19, 2020, at 6:09 PM, Michael Knill 
>>>> <michael.kn...@ipcsolutions.com.au> wrote:
>>>> 
>>>> Awesome thanks Lonnie. 
>>>> 
>>>> I will give it a try although I have no idea what it does!
>>>> I assume I can remove this when I go to Astlinux 1.4?
>>>> 
>>>> Regards
>>>> Michael Knill
>>>> 
>>>> Sent from my iPhone so please excuse my brevity. 
>>>> 
>>>>> On 19 Sep 2020, at 11:56 pm, Lonnie Abelbeck <li...@lonnie.abelbeck.com> 
>>>>> wrote:
>>>>> 
>>>>> 
>>>>> Hi Michael,
>>>>> 
>>>>> Great info!
>>>>> 
>>>>> Try this in your user.conf, and reboot.
>>>>> --
>>>>> PHYETH_DISABLE_OFFLOAD="tso gso gro"
>>>>> --
>>>>> 
>>>>> If my hunch is correct, this kernel fix added in 4.1.17 may be related ...
>>>>> 
>>>>> net: preserve IP control block during GSO segmentation
>>>>> https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/include/linux/skbuff.h?h=v4.1.17&id=abefd1b4087b9b5e83e7b4e7689f8b8e3cb2899c
>>>>> 
>>>>> Lonnie
>>>>> 
>>>>> 
>>>>> 
>>>>>> On Sep 18, 2020, at 10:56 PM, Michael Knill 
>>>>>> <michael.kn...@ipcsolutions.com.au> wrote:
>>>>>> 
>>>>>> Yay some progress on this problem.
>>>>>> 
>>>>>> I had my 4th site lock up yesterday. It was a site I moved from one 
>>>>>> location to another. There were no changes to the Astlinux box at all 
>>>>>> other than PPPoE credentials but after a couple of hours it locked up. 
>>>>>> So realistically the only change is the internet provider which is a new 
>>>>>> one that I am trialling and is the same provider as two of the other 
>>>>>> sites that are failing. 
>>>>>> 
>>>>>> As we are also using this provider in our home office, I set up another 
>>>>>> box this morning and connected the serial port not expecting anything to 
>>>>>> happen but it locked up and we captured it. Yay! It is attached.
>>>>>> 
>>>>>> I'm hoping it will help the resolution of this problem.
>>>>>> 
>>>>>> Regards
>>>>>> Michael Knill
>>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> _______________________________________________
>>>>> Astlinux-users mailing list
>>>>> Astlinux-users@lists.sourceforge.net
>>>>> https://lists.sourceforge.net/lists/listinfo/astlinux-users
>>>>> 
>>>>> Donations to support AstLinux are graciously accepted via PayPal to 
>>>>> pay...@krisk.org.
>>>>> <APU Crash.log>
>>>> _______________________________________________
>>>> Astlinux-users mailing list
>>>> Astlinux-users@lists.sourceforge.net
>>>> https://lists.sourceforge.net/lists/listinfo/astlinux-users
>>>> 
>>>> Donations to support AstLinux are graciously accepted via PayPal to 
>>>> pay...@krisk.org.
>>> 
>>> 
>>> 
>>>  _______________________________________________
>>>  Astlinux-users mailing list
>>>  Astlinux-users@lists.sourceforge.net
>>>  https://lists.sourceforge.net/lists/listinfo/astlinux-users
>>> 
>>>  Donations to support AstLinux are graciously accepted via PayPal to 
>>> pay...@krisk.org.
>>> 
>>> 
>>> _______________________________________________
>>> Astlinux-users mailing list
>>> Astlinux-users@lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/astlinux-users
>>> 
>>> Donations to support AstLinux are graciously accepted via PayPal to 
>>> pay...@krisk.org.
>> 
>> 
>> 
>>   _______________________________________________
>>   Astlinux-users mailing list
>>   Astlinux-users@lists.sourceforge.net
>>   https://lists.sourceforge.net/lists/listinfo/astlinux-users
>> 
>>   Donations to support AstLinux are graciously accepted via PayPal to 
>> pay...@krisk.org.
>> 
>> 
>> _______________________________________________
>> Astlinux-users mailing list
>> Astlinux-users@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/astlinux-users
>> 
>> Donations to support AstLinux are graciously accepted via PayPal to 
>> pay...@krisk.org.
> 
> 
> 
>    _______________________________________________
>    Astlinux-users mailing list
>    Astlinux-users@lists.sourceforge.net
>    https://lists.sourceforge.net/lists/listinfo/astlinux-users
> 
>    Donations to support AstLinux are graciously accepted via PayPal to 
> pay...@krisk.org.
> 
> 
> _______________________________________________
> Astlinux-users mailing list
> Astlinux-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/astlinux-users
> 
> Donations to support AstLinux are graciously accepted via PayPal to 
> pay...@krisk.org.



_______________________________________________
Astlinux-users mailing list
Astlinux-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/astlinux-users

Donations to support AstLinux are graciously accepted via PayPal to 
pay...@krisk.org.

Reply via email to